From 6aeb937a93622f870660071657c94d8759c63298 Mon Sep 17 00:00:00 2001 From: "Emilia(SleepeeSoftware)" Date: Tue, 3 Mar 2026 00:36:28 +0100 Subject: [PATCH] skeleton for the sterling runtime and base compiler --- Makefile | 31 + build/Sterling.exe | Bin 0 -> 61999 bytes include/stb_c_lexer.h | 941 +++ include/stb_connected_components.h | 1049 ++++ include/stb_divide.h | 433 ++ include/stb_ds.h | 1895 ++++++ include/stb_image.h | 7988 ++++++++++++++++++++++++ include/stb_image_write.h | 1724 +++++ include/stb_include.h | 295 + include/stb_leakcheck.h | 194 + include/stb_sprintf.h | 1906 ++++++ obj/main.o | Bin 0 -> 4689 bytes source/array.h | 66 + source/graphic.h | 7 + source/main.c | 94 + source/sterling_runtime_x86_64elf.nasm | 34 + 16 files changed, 16657 insertions(+) create mode 100644 Makefile create mode 100644 build/Sterling.exe create mode 100644 include/stb_c_lexer.h create mode 100644 include/stb_connected_components.h create mode 100644 include/stb_divide.h create mode 100644 include/stb_ds.h create mode 100644 include/stb_image.h create mode 100644 include/stb_image_write.h create mode 100644 include/stb_include.h create mode 100644 include/stb_leakcheck.h create mode 100644 include/stb_sprintf.h create mode 100644 obj/main.o create mode 100644 source/array.h create mode 100644 source/graphic.h create mode 100644 source/main.c create mode 100644 source/sterling_runtime_x86_64elf.nasm diff --git a/Makefile b/Makefile new file mode 100644 index 0000000..c2da10e --- /dev/null +++ b/Makefile @@ -0,0 +1,31 @@ +NAME := Sterling + +SRC := $(wildcard source/*.c) + +OBJ := $(SRC:source/%.c=obj/%.o) + +CC := gcc +CFLAG := -ggdb -Wall -Wextra -Werror -Wpedantic -I include -O0 -std=c99 +LFLAG := + +all: $(NAME) + +obj/%.o : source/%.c | makedir + $(CC) $(CFLAG) -c $< -o $@ + +$(NAME): $(OBJ) + $(CC) $(OBJ) $(LFLAG) -o build/$(NAME) + +makedir: + mkdir -p obj + mkdir -p build + +clean: + rm -rf obj/ + +fclean: clean + rm -rf build/ + +re: fclean all + +.PHONY: all $(NAME) clean fclean makedir diff --git a/build/Sterling.exe b/build/Sterling.exe new file mode 100644 index 0000000000000000000000000000000000000000..b5e53cfd1d9f09c9a40d95387accb5a26e62d1df GIT binary patch literal 61999 zcmeHw3w&GEndgyZB~F~gPC{q`6ck8k0&(Ib&I?G9V#|@i58+2%gt<{<*;XS$=w-m4X>U% zclI(#I^bih^<(o{;*~E%Sdy`?xfmN~H?rb_f-;;KyB?nc#;QwkrVm7O(0?iePIA-X zq6<0#sGY&sZ4Wr0*FIdhIaszuSrAbts+16cH^|tsS)$1QCm6e5lIGXLjP2uv-2Zw4 zjIGE~b@U0whPmK5Dgj!Z4tJ+<9wEb!K%_hLFduHPUfmi@2LT+W3J*Rc+bVoahX>WG zlLGM^(SS^>2%pvXm<|sptf>$HZpVEn6TOWBrv7_~UP930wt!8z;WsM`aR#1eb+^D9 zAUZwx5aHYKF&!Sl+n-7iXpn3&KqdRpLwqBG9`BPu;LOIygOBO(5WRGi(<9Td)A$ho zJLtkVJlrnfXgnnQa2R!jSBsD7@G!P*&01lY9)geHLwNOcVH_U9TepP+PA$S|6rU^c z3E*QoJcL)VmP20UK7fBjug!#nqKEJ{Y~a902(MJ)b#RdW&qH_{H}UIX`tO!_AK@VV zpNH@^m-Fk7(*JUa7v&)RpNH_a2>Di$Tpr+2zr}PEUhn|V3VHwP|8@(s`1}6R-&a@a z?`v6H+vjsv_0_EyxaUODtnY113e4Z{`cWlgXBixWWwzp9pYQE~fqviH{@$0$ zbD^K)=wzMDyxiyyy)l3a|5@KSVV(7jlB4$fM#Js?e&+A~%oTY+1_buH*1WhD#nf1%^Oe3V^i8KP}*p zKLEtvcLsG`j|m|by@nCv?hcRl{wix*Lu*>gQ;KC%S=7GtdSxgN*RM2i#<``IJI z={e5=%raL4ZEW_l+~6fBXP)3OGud~CmpQUi0 zdC<*CpYSS;O2FtXG$1)@7xK!V$;yjR_actWTY_#cq4m3>68p1&Lb%yOYceIvwu8AQvH_6M+bzbh_U zM{xq$ufa&7rKM6&0nOjP=tVCV;b}Sn@lDXnTsf1&jskX;2iURMXI<++2~=3`18=jg zYa!`w-o!Tqe>e>|h_#%CKMeUFEk*6OJOp8WD0>YqpB9&1o)UW*O|C+vYknEF)q)?&X^<{t_8Sa|h?|U8Kkekr{gMhg*g>>nA z-T%-_G%EBF<%S3mkMc8IM9mj*^TiKuCqaA@P-UXzFId~wx;^MF5MvjRIH+gTjPBx;5LMi;L6g7t z4Zr_*beX5G&W#AQqTlE49d`In_|J?wsn-uuALyfZ_7L^+g?I1|lvju_qC&@ouLDd0 z>xBRK72idfa57{EiZ*95S6>SV5r?SRXyg9l*A$Sjgjm3NI_)@u6nc0jmt(9u8%1y{ zmEbu_h;hH;90}21Hwup-78+#fMO*{stv+|jF(1ZdHZims@t-96-V!X6l1-|M9usN3wHq zQQ{bg^u0(!Bmv;F9QgYH*7nccOLfN=4FTrw`>DUr^{vBz_tlMiFa7}VZz{NN;$q8q zEqP*}t9vD5bs#n$2u}kYG{ZHFc@f2r&%K!x?0sqT-S=F$8D>Br-HlW2ao0Aw>T}(S z#$zwHDf#8Z<@O^!qlRRh1uBa`k0yn=tBC$1ucS-375o>Mf3lyklIwJ%aD8`?z5z^& zZhs&8{6$2H32tb8vU-E^>3-iE@TT63^cj9JT;e~IDTIF(W`c{sYUNA*<5kN*$9Mj$ z?^RCZ_@ckvj2ied^~is@8fCaJ5u>qLH;$U~HGQwt^o=}6{!BM3uylQuTj; zNc$!(d;&OtpZDKPP9Wlg7=ULoe=LlWfrE^~B#3D(IQueneP;w2f4}P^=s4=qmJDh} zF&bXx^fKt}$B1R$3zpySDC7a;`1r^%E6V8LUf`h3LTaS$Tc}OT(*~Lm$K2Ta!vXVw`T>v5n$PhEq zJMizp?%oMkdN=+(80(!VOk-{=e6Y25!kNAq{~p|fq%u8|tX48s50yMv>Ot>`u{vBd zstXS;+SJ8*TpU&xH{#+`>Vk&p*VV-wT%4B|10>Tgt4c~eZeA)W?WRr${IaApQA!1W z|E?mDW8TikBJx$Z(Q((8K>+RMS1l|5Wy$>em#x76^fGr%-%rJ0$(&v!2SdpUm|+E( zVFk?a&`V?=gcO=l%=3kuw6Uxx#|*I9e;ji%)D^s6w}e+T3ycidp`?0Gwp*_w!&W>m6?y_70VkrTHr7>&O`} zLJ^p5N8`sPYfR)xoFX(l?)v6>9)bJ5N5*;O9m;NA@15|J9H9w>mkUddye7(9dvGo} z@_kiPRC446RW6|Q-vknDHZ~Il4+;>ZD4!t0ki#$}=CHZHxHq>+q`l#D+b~|{K9>*8 zajpg)Am-d*Oj4MfygYyN_q&ewl4W3+t8lTIU-Z2I#s9}TZoABn--gbvc96z|-sk!~ z6h!RVr$;VYdqW?Uz0X~jUk+L~mK^B;GUT0|AFryP6uOYeqc;r^e<_^hY#t2Z`iRCY z$T5E4qbZbn5GsBS!|&nD9Tpw8`W7d@dEqOF8Mq|wG%=9ER2F`MK}PMq`WDYcpv>?b zuo}M*7L72HY%*XaXBO)`=-~r0d@F^<9@b-glzkIYQQ^_&c)^`M)H~rWdGu+1Rg|9H zJ5ie6eBpB>J2WZzCPPDz8N$`j@*V<{zsdR`|6Hye>Mwj;HQ@_Jy$M4bG)dn}1b_~F zsv8Ruc}Q&#NBbVayn|zsctrc6e-;^fXly+-5IPNweF%xC$jwA#At8^?{nZ*IOfO{8 zzlP(D6%Np->aQcoP3C#Qk8!$^lZ2;H<$2MS0?$*l{2~}5;qHCl)!PrXBfI7qH$44E z|3K4fWoh`MNOD?+YWwTXQ%Xze`tiAkA*sKw zd6^sHes_ZRv(ZWE5IX2k7y2db#lJ%bdMAn~cPLF?HKu_o`PYz_FEq(d1~G8Pwi|GG z322|m93}V~1FrlrLGTY4@FI!73vdiq-#Co}+F@H<^6+uQXC9p1=uxBZjq-li*P?ut z{|HU~{jNb=@lf_hC}e&CdPr<@O*rsmV~OJZHRzwNvq9JN`92<*$XD4=Zu0J3| zUc@#K(n|`2=H!y{CAA?_q=x-SvzNoW+*>Ovx4U`Z;y_Pf@2CSY3;S_6F5D!d zg23PUY24>1IkEz|y(m8VQ!4%;hkjQtKKddRKgFRt_2Q#nrQ#aSrX( zi;q4;#m5Nx=s_wTr(&;@ipMxEr^}0v(w5;?>^?xUPoP6Z@zL7}3TFZIX1)06N-9F7 zo{}SrsOTbuM;4;UBf_)r8kV^UaRl8pj^IA-=(?JZP*GEg8cd$tGQS|_p{%Rq$WOQ* z`NlcY#Sw+^9e@Q{8XTD)&B;NQ*F1|F)d-p@vJWe=grt)li=i7Kup~|6E7NTueS?sg z*Zdx8X!(|QWe#Azl3d2hxa8>@X)N*`U4#4rg)@6#Y^Dr!e24QoIQFIpq=RE4(Oh&h zm4y#p{Gl9^d`ZZMhwguJPVRTD2Rck7M6bC`3|2aqeqt9;{QE#Qj@0J(bCMi>Y}8nmD<~9Cv+QZ!=TK z-S7G~2Vq564K~PCxc#<>T5@&l2XvE)u)$*C8>9wgg2}mj2uK1zdZB`=`ezWZLTC0p zPCwmU! z=UE7bFQO>=|1Qq(tn0^skRy#fkjY@h)9*S50NrPaLa&!NIy85iWTOy)XT2lT#VbU4rb`d!Z~<8o3*B9NcSED-#p`+kMHFYQpsTV(wHige`uo&i+L z`;cmCyuZKYbwz8N5b8XHIB{qir66yw8fy#B}9CXGF)E5sf5Os6- zVvfWZn@`fTb}Cu5xZ1(`}=6APQveJ!Q5MEv$} zpYKgHph!0{hP#|^-s&6gDj54)Qb!_AD8z9-bN;ONfj3zf>FbuavN-v%`v3t$eq)9F zKR=(Gw9&!0`rhnX#i`H6DiQX9&B@T5>dP<}SF8n-WGRZ7u_7+dpN>5)BC#|MPUid^ zn(}8PGBXjgVHPaq7E@5-I~H_%+xsjrBIkOCFw`7lPy@XkbW_tqyU8>VVE2|4u(^4I z6drodxJbGFn8@_KoIyrU48aTy_9{kvTlSa4?i0Y%b34!7zo$=UVn5cOj~!wmMYns# zGuVn09O^I;WO3p5Bq=@DS zcgXWC@_ds#UnS48<@v2n$zPuTPM*KV&lm3g1?@2E2fF(G<9Dz?VuyUbR??tF9vzCL zxlAx*LwX%u(^xbg+4z`*Ek|ITmEXRs>%pw+fyb=)7Gm#TT($+Gk^SwVke1R~53MU( zySg=774d9_pI5s z-oWl(p1OZ_7B((Vi~cN<=hFk?dN9im-B;!AcW<)2t1}!+yIbR(K|Es~PA20?_ll~9 zI&V$=N~Tw*BC+P8+P~_t7{u}u4I8= zG93v<-O+d|<&L+xQ<3%Jpthl@iBoEhr}JVp>sX96ci?gOaH=bscH>d` zbUf}(bq1qRl=y@4;RGt7PPo;*qNTn{NzoEJ5Q`s@+HtwH&S1I&%C1;>i+g#Bo~~b> zTCu!+m3w*nO1FDOGMqjbjBdq6mdWd2heKtDHdfF>`RPb!*xeS4M7xsVTLhxFwG|Si z+{+X0j$q1-&Wv;h+rwq4a3~##$C%r_Ba%#a1*3O$g_DPQW60VDn=fy5??0Rlr`*A` zJE&6;R>oRGwF@NUY0y$F4@J@)?hw8d;4bf8-c5K)*F-AZ)f$Jf@Vs`Ac+(dP((zC{ z>PDZX2(Y|$wNh`gcKah~^xJ)5Gd@WRu$>8{hwkyNH1#%jidF7(FxehX!*ZxvSK9N;FkagXirC(3h_0g>BfvoG{EPCZx`mii{dX`)5 z;lA2>J(xvrdu8PiJd zwq!Wek#;u)V+fC_Rqo`5mF~40R08$Fuu=@_oC+F3xYxq|Y!$<8Y7|1?kh zpRDcsb?;l?eG5!s0ql)@fc8gv#F@TlSA;U1yYN|sGM!K05sU!JbpGrWjO|64&OgIL zTOBCV`PcZQQKs)eIq>O0xfJJx_#8pmjq_Q2j-ot*^E^Du^(n%^`PcXiqD<%2OBj0^ z<#L?k_za;;=l_b&ODNO16b zoR8u263TR5z6!ddOy@CtUPGDAx0Ye&17$k*>@$sWPit{b&7~6%i2j@C`T2aPYj$!@AqA1h( z1U_k$v4&$y*24}cV_sn2#pf8x?;rpE7H||6xC)()z^o#ttEiDm7&GWA2SmHE6n&2m z`7xby;j{pLK;qEZQC#RMb}V2-$w4&N37UZWik&kaE-24}gAegS{RY(Ux78OHwgCQw z4u4$IGt)N7Zs93)<1ne4($|7_b#cqpM`yAtyJxTk9WFM{ImIJ!9gpXlBdu z%RD%bo8aqAWiy@2DVy~((=(Tom%t|6%Mt>_g z)j8W$*)|AC%`Fc|Mwj_juOzsjTbg<@Kvs z@Sn@;H{|&(c`p2-lv|!x$#aD~?~vy_vB(QrCki5*t#jz*j0eqo)% zhuzF%Hkf_SQJoBjYa{!U!Q^3fv|xKUT^mfLeYAbc*dNn%UA(m`8unvnB^vfJ_ICus z?qNu7Q8M;vXH5*;f>G@4WixowSsM->w1L@#vo;=Vm3%?6Tar=hEuy4uuo6 z*I141op6$U+0_&chZF3tUCq%H*zcg-KK71F?t#j^Ogv_!AbeMpdQ#I8nih^7j3lvX z%09rfR5-20!iO|^0Sa5pU?Tn=iY2TKc3|u(M%%#g5V9<$h2pWcNPD^?84kA6?q@Vi zP$JQl3Mb{xFY;NO-(Vu4r4J{< zY_n(qlXkK#Mp+B98i5&!VIwutswIMS$ANA==-TXH!Ts@Mn(bsl+AyO%_+V66IY{zF z;`_C>u2=|}B-b|-@8o7?1>(9*Tsd)-O1DPhFfb5e+QGI&G7?L-u}`wL5O(;{51(Ug z@kBU=DG&V;Y{j*(rIupsODvU6Mp4P0qEal>nZV)(a!liMr?0WzSG#^4zvBWDSAa$^ z5h?3Tl^u%2%4oaMWYMzd+OoA};0_)ydoO%Y20(N5X)L`Pto+k&M9DMThlfzGrJURO0_>=5p z=}Gs=@{^vEfs^}ACQkO895^|6^7P5!lOrcbPa>QC!mck4e({vdWXwyEU|B z3-VDM^YF>#LBK?|6S5Q>g8t>YV zgu9@y5YL4KL9(=vx0mp_lx=pku_EWv*}E`O=4pXwFcj{HM{xs~Id3Q?Ans2Txu`I+ zx~A60oNJb>#YHJk^gnvNgPlI*~&e*u&}j$hJ0dOwz9294(#WP2CY%($0I|3)RcbU9F6 zD3f953yu;p+>L-a2nD;jGCg))SHP=0R7L;T42_yoQEI~0Eg-?7gh2(N0vEARE_z%q z3E}$*hiZfdg(Bm2dI}bh#(ze&dx!$J1pns)48rgX-Fa)leB%29!s9y9KUPeS0kMTH z(f~J;cre>Q6*wH9ZFORS6%4V3^QgYyFv49h{XV=Kej)lXE1!wALR!mDytSH0GHIqoI;c)?V zbT~fyCW%E)O~tVKVXkY)>JU!IGec)d7;I({X-zEQhHV-bqiD{52AzV(YB>ykIBO=6 zEPQO}!5Pj_98dbN*)vIF=VRY^kX=D;u!s!re9SWg_kB>J7;qi9I(~r*OHf}90MCc>SJ!&Cb4{C?y|uM$ zrh!q{uwxf1GF|)X8X9+Nbu~?O-sVcbCK}8#5UN|it+FnldF!jRZH?Y~Agy1=W>1c> zv4Ry(j?vUWLN>8ES!mwc785HrH>(=*^J<{6d0BPVrdpk^j+IP_qSe&Z;pL&aak;)( zlp1RSux72dxn_rt%{Oqpm6a`ZEr9#9hL+~-4K?-KHE$!7T43M=8XIaGwzv4SmL{r( z{64nOsP+YFY}Hp7)nxoypBHTw8NdKu97Ecn8+}2WXp}(COioXmujgczgoYJJK>FasyDL_88w0| zdR1!(V5HHoujJ0M!oZPz*3{exo1lwU8dcm_joO{{4Lj@Ms{vA_X&bE5(6o)+WZ+?- zbL;!aMmrEkI4#(^K~t{Sgm|i(n9SDX^;ff1R#;_ia}_JI!nQTD)doZ;WDR87>T1}U z$x)j3RMt1+?Yp`JIqB3kv9%_YsykW`G+XL1)@s>0qoM(mK}|Ie(CGZ8dbZxElLoKG zG~if6<b8(s<*Lf11vnNV%u)a3p6fkcEM2Fje6p{-P`PkLDtcX-NgLX z%BD(hGnq@))>vz0(+bh_4r`@qDr@hw*2<=vMblbqrD`f`>x^2V(AssItLwa4r61oG zXkzu*)$2BFG*vfbS8rIm&Qu*Rs&}lbsz$`BYT*-$i29A1AAvPc#qKij$m!uam6e!D zeHx~;`l=nQ(Wv+C*aZ>tjfP7;d)xY z*I37PSrA|-b9T4eQb}}q+@>K`Ui^N_VLNZc z%k~?PJwBSUb~VEdG{LLBiiM1N%xHBD7=SWp)0E6wjp}MICJZq+S=a!Zk_T;PV@gjT+srDWoFPqUxkkEmse)dR2`W?8xn~Y-;x2iQZ04jYymYa@s%^cH!>C zx(u-D|GT7d4jNTFY&SJmvO@-pv~R3zWZed+7B0dL%QYvh*ez?$e{$SaD%Y3ZMy!b_ zS+iaT$eMD2DN#6R^ERx<$>lk47mX3ITIAQ;ns#~vOsoevfQx~j>~PGGdwpW_t<F3JMCfYTA@ryAk_y*x2YyM8n-~16HKN)ZU#AP}_QL9b(?q!zQ1k zLc9$e12~&*Fl&i)k`>%kT85>4;XAO^5^RYSGlw| zS5nJj(GVMgvf0u?f*9>)xw09|zL8TZRt-3BYB1liV5VpB{hZyL`Htm93yYoCU40#a z>7ntt3mwj);$p#)M4C6(B*Erl#!4U@PNh~2g#7bwgeez5Keo`%9&kK}_E$I@q{yPB zkY^3m;5pFJ4}e{XbKceSmd{)0BG?Krxdv3tMk|CQA6`1McxhqHLNcCf%hDMnQ6W*} z))HndfGw8JuUUBelH2eU`-}><9q3L%pDm0%Q>Z!MD1OW_gOnvji%1QUY}O+7A+(!q zwRG`Ph&6``4Y4lkpnXFZ?)_Bb>~J`K0}J8=_lA2?={hGK4Cg;+vJ>Y52R49bJV3Qa z=wCm<{*eBCj{bd-{<&_%u9w>4yz+HdQG2gzPOocbSHYTvu325K*+uQHA`D+g(Jx(B zZgtISa}^Y=De862`j*Sl<@(^k*IlmNE@!)I_CLC=8g!LB>?-<4RQ9^&^}33RUU8M~ zb`?GBnsukkS+vzPqu~05uDMUUoLh_5fPnLG(NA0j&RI{pia>04!3>I2kckw;Gw!(+ zEciNV-Xh6Kxq>T6Z~hNzsEttPo6K?j_3gXMyY6gnFW*{w{q+S7P^*%(Z_7v9c?8t{ z2A{vjhtn@CpzZ28w1c~lO4rdpH2?QH=-w}VHw)@?yXVX-8kE9RNP$#Gs=aFOywZ;q zc;_wdbym$w_~y)T7rgeO_vHmMXGwh8OQs`x>z%i&oIAr=?syq&_&w)beChsTH$KYp z>b@e{KTLdkoUwVO!`^vCFA$gakKfe-t5>fPe`51}jd=8IO=SJ%jcY=O+wm;Y>L%8B z%U%?Kk;3y+!FKG`YxIT4jTL+G-SINb1YC@b9<(oh^`xcO^sNHFVQ#b%<5ovbP0K7^e36IwR3YaBo#Obs!y2 z>}|pmc~SH%5Y}*nH3DIcL|CIE2zrFaJAPxu8oMv7iNr$DE<6S)xyn;0j_=YY)2lo1 za2+1f<7avrO90?MD7=fFIEm-8{~ey6%})%7b+CHlEo*WQuQkT-T4NesxlAgDR~8eN zbxmU(G|i)&8FJ8|hmG6!N6YA*TMA{P@^|*M_X+%;Xo1!Io-)3}2jPB;H@UPsFa*Qh zVYsD!h^L2<%K|bJA14z}2O}@lA$VAoZW7Zg$0QD29l(duXwixwbi16M2{Ir(xT3?s z+*mV;viPrZ3dRNjKpM<(ka?L7+#{98GdNR5Pt`Lx?0NbRp>Zc1lu1^jDbg;sI*KQ> zTMzM{Cp22Uh8ms?GdEH;T<8b#ZyacaNidmiL0PmCbcD7BoiWsRahzS|-Q{X%eoYPa_yab#cm_486Nw0C>j88$Q2;DR~ z1)WvE8JvR7HsB0QfwPNT0_(fUWKIBQcnUi7R7=kkIG+bjVhWtU0M5u1y8j$FCxBxg zJGYaG`7sWh(kXZ?0S)&QI9}kCPk|Ew&LBSc%Q@mmpd+aKEa=!}_HudUmkrp@&=-Vb z9fKT2|Fu-Qf*XO?uYpaY);QEJr4j@t$ciljPyGrhUGa$Ia8uE8s9!WCAjK7@qX++I z<6|7^*HH-qL4ebY|BL8SKUAa*10o~s;dV-r=wbi;;dX^c{YeK*1Uts}e}rFo@ZX4; zUy(RmFJAv0K->_|ZZVQX9IparBqxrjSHD3^Y6`2)!%0LvMZ<1Y{^+2p=4WM9H?D+L z1D1NaP*&|hSyEt{&h|qs)04+um1`FB_Xq%3{=Xxg3X{@UkCQNt|MiL!LlZupTQV4n2hGhukTnqm}=52Q&)`({1NOkP~^KnioAzvk=_dAES zqE$R+t3QzZk!t~G0Gfziw&y>DNWV_mgERLQd_Rf=#2YqpQYL)ESB1=^``-gSO1`a=4~FmQx#@F;{U<4}EZ1|Xq3eNU6VF=!~Se8YVSzTp9#OV=w$f62bM zpv!}wHcxdW3~A`wrQNMs5O2nT;(T-q$bF0r>ku?&e**GC=Is7@g{t}SHQ*$sh!Wog z&ht6@{Vx?6_5#W{Pn0PVh}9om83+Ah{IWfJ z_K?@)#i?v;Ubd^&dqP=Nd{M$6g3@Y%v*+|=I7#s<_C3HC(U?Xod8Ba&I#pi1PxJDs z+N1Y!E*v>7UI%HiuG8Vhf9s)Q=3fDvH)aaGDhDlkS*K>Hd)2AQaHK!~D|m_V)g2RG z&^0_-_YH)Lsba4Obfy|#!@v=~0jJD5)OdG*k&q`u-)6S#dA!^5jOg-Y34=>v!y4~V ziw&)K5@*lAlsJ`Myu`*dg$V3 zktd+bLoKa`$|VeWa>y*@*`~`QUb0%o-{fi-X%tO`Jnk&}e27fO@OD+aL?DzJMVXIf z0nvc%>BocMXv8{bUueBG)&A{Pru(ILM(;niu1vt-? z=AyGk$GHPIioiFlN_$A7Is0O}F73Yp=MHSu9vJU0Fdx0bRf^cX&7>DW$ zq9ghOi6`FeVp{5Pod5GC*i&-|%M-AOQ=coa5#61_t>)?QS-a}+pw?{v~_=#WlN*9-3sC1YILNU=r;BBo}~eIUa7l!#>= zN}dzCJV?W}w#3?ULm*tzTjJA<%ac7)$Rdm>l-xLisa-GYNFprNU;9uNSqed|$0g|{ zS0&FwI?-P_Mx5*~(tWDYG^pE$WT7>aaVWh?FQLEoq1mPDMVQt@_1Cv`c@pTFklB5B zgTI1=LZ0mYI*FFz_Q+^(iXTKB0)=_A&Mu>L?tg_&6OO<>l;uv>V?IpxBM49MVch z^+iVS3;Uxj%0A4k?32yq^5UUk9}vuSKxTR~is<+IVJ7l6>p}faUnCd({&{PqydoR! zLs{6+jS?DU9poGIt$E=aWFKi%6CU*!`IzXh*>)npr8^_1;Z~I6I30JIT8pBx^Eg@y zpBJl9`O-SeYLvMz$IcKg>rnH~0C?pc%QEx)tuBvxPlhp{Du|FLJ3b4hcHf|9gsq2) zKji6Cjh#tlmhq!JF6#Uz|-|U=rO4=S|>@!&D;jlRSp*AaCq3ayn!m z@?HZ67y4?V2W4uL{Zt-6`YP#CA|$;#T!qy#%t?c!CZBoK~;PKT9rwN{ zgh+4kD{r|h?$3jWh4KcPD=m-;%I7VRjo-I%4@`^uu#NkKjr(RB_n+Ii@0u3(mu%c= z>a%EC;jwXl!^VBzw78$MaWArQ_uIJ7Frtm@_l{|C|0^4Jde+*id%(t>p7WGAl6zuW z+(&HO7u&eg{BDtF<6?umj5a?8&Qy`s7&|J`n#PyZqDN_wD~igUrFSdR9z(?7dQB2( z4VhJpGRZ7;_c`CsdHtwS~?iDufL(}5kZR38*#+`PyE!Nn$!^VAhTHNoqaX)S2K5paQX5)Tt zTHJeV+=p!3Cv4o4Htr+S;!aP{Sgg8X*v6fyh^^w`#vU8@^V8yf)W-dsjeC)e`x7?q zqtoJk%*K7h#+~+`E%w>?l#P34THFV0+|S#%FSc=i#^5gJ_a?NODz`CaE0x=rW;8il zjVmqSYuSg|yLh*98@iLltvX3=W5}#>8K&z>I-q1q%yh&5(^Yrizx%*^Kyu4dJ-+Bo?Z^*2C-Xycs{qae3S3dt9 z>QzC{WpHoF@6Q`LR4?bp*|b^t#w0$k>vdUPH%C!c{=C7j5g4Y=`E3Y%jEWPf}C(&Jb-Fv84#enLs{2sL-pn5q!&Zbi3 z8xA})l6Bbhx-74I-PnbYJLH>{Jj%x^x|J3Rvhi`7*S$E2<<%~+K}UI=iI?;`_j~Ad zw_SqQ88R!cGs!Hy?q4R+UG>g;s8>ag>aYABwZWr$IX}*(jPeb~(k!1dc8PVpF3an_ zp(rbl+EA+GQILv&(n3Ktwi#Y0y=8^+mdoNkWaGZr#(kHK`#~Fb`t^CBvCF$B^}LO{ z+s1vLjeC!c`{rqJAGUE{VdLIm<9^iOE`9ztXf@R=WAvaJ@1|asJ?K$dkRh{*yK9sl4tz)T_d;`YXRjZ75Q`oF8Y?G36VM8?(kgW*MpbAL+W&>xl6+Q9LyhGGTjQ z-~q@3QZSqvgO_Iw>LZXa9rkC02*-NZcHPClRWay^`If4ybqgJx+|aGFs|0- z3Q|$7?5rRgtBr9X{dt%2=gZ=ruyN1WxUaBruds37H!bc_8~0HgcejoEjW+Hb)8gJ? z<9^=8eX))EwKndFX>o70aUZd9FST)}hh%sp(LKC-THN>9xSz9eFS2nTFU;oNGcE3W zZQO@#+-V?6O~pvrc;3eS=(M=+vT+}>ai74bwQ_&n#(iK~+?#FOPusYU+qgev<9=dV z+ygf5r)=CaHtwIcaUYx(_gWkGK^yl`8~2adxSyI9cfXDM2^;tGHtyXv?x&~4y~@UY zz{Y*V#(kfS`_Qzwdu-f~+PI&yarYbC_g@J z!>E9ex%WJ?=hIwtMBLzCt>I&NlAPa=S!G`)nWgT3Hi_;k2Y3(ls@SX|Wd6)*!-VSP z{5YHF8(+dc^c=PIP}$e_blverUiGd#yt7kbHxL0FK%=x3$31`uYq)dw;Ctu^B#-5t zat@Nye%!0eOkRhnHY*#X77qgF7_5rhTfoUp=M%t50EZ|z9ro{o5S`BfXBUW&wQ@au zO*r%&A7UhM?BDUh+lN9L-Zj)CacHzK_7o@wERa_KIR_#*@e#$?k5N?pE?WI7AVU@$ zCw-~U0=ZIt&5v4$R?7kDu<)t?gxtd56#_)%NTSt)fY1t>^b(NI1EO|d1mtUgOjx*o zhv--!BY@CBC(-KfQKXn@Q2s3-;*PT$IIvih-q&=(BJ&6FGL*q(qPGQ|n*g!e@WX(N zf~=sdg~HvDG*%opP%ENL$KAl$hf#U)qLL>C2z_17fE*(T4h%Zk>In{Un{en2PvkpB ztMhYjptbWkhZK7`W+2+3?%^%HCLn6Z zLr}f|$T`c1`!yg#7DzF^5IGEV!x}dN(gR*ZL(sVe5Op_MKpFt?=n@9^;|IxsEuQgk z%T@{CoQ4X97M}sc;I3(S#~ywyhz?`KJp~*q67?-33h)sXouKvaGxB*YKj zuzf%?v{(g*n_EoLxdo5`T?^t`0|>=BL+0Ip1W;q}dIXRYKrkSW0b=#KF9G5P(BS?Z zK-3+~0w#Y5oQYj@CC@l;deF+?{uUsH#ZblOz`Ip`C^TIR$Ucia%K@>Di%LKk`W*vP z_71)2M;zepz>6<Mm4;;@!--=JDGMZ@6)5Y(z)aPK8L2%17>O}yAFJPe@W ztDgc6K@4kr9uTX9Uk9YrqQ#4VtUxQn&i@^dK?@xxeN`8Sf%8E?PVin)EM zi6S(24_HRQExnJ_FpBp+St$Pta6A^sJ49J;g#x<*UybV4HO2RQV9<)%t4iNRki{Ksrg!;m0bq^rweu}{P2q2?6ueM}3*a}$^7Tx=RLmg$rpU(oK z_9X#Q7vF? zp=dl6#^|=tX#|dufdG;11Ekb4x<3v`k1h|-zToqxbx+VJA(63{C!2O+&=^y zzpfWSJ`4y2U!wWPThF^(_)1^Pd1x4qUntkgotTgexOzy#&as za2~^deg()W;DESnRg9mpP%V!w*yiN zc?_I80oiP!6Q)#!A8zC=AWFkH>-c@(^jP%zI3U)T@EJgM-JnJsW(SH0tek5B%ADde zz)??P2*^JGGJ-nc8^6NU>TrkFmJD`=asMa0zpGsfCWEo|a7tZ9Vr_9%4({Kd3?Eci z_&q7Tk%a=f&y0g#TS?+2`mtbpxHXtau#R9#3(@Zr(c8-D*T!&P?qDz)X$^pNXBfX! z>)cHPH{YbS3oPZaKun`wN$?PKMj?o1Uu`DKOB|hZEwbsZ0$u zElqDmH$iL!@J8{-LAc%T6;9U zKNy7tg3&Ok*adgU(ov+Lxu8KCt_+ekcxzm}HQiPvcBt&+Lzt~?(AL8-zDYs%cQsm? z><^;vLLK4I0leszJDx_KXKSe3FOrU?Q?1E(T5F5r*VLv$1iMSYOLfKb;nj3vG;9kZ zIcn`nL?a=1k(NkcpkO?tFfM$4t-8@$ho2PpZLg_EqSG2qhT9`43=T0`wYC`0@Ph=J z6)u6<=t`se?z_)E;`0yjtXNsgf%2sk+w*&~mTZAY+Nn|h3;kgOJQ);8#wo1Q7o9~OfI$e>=HCIwYq^8128&=K zqIIS;{QRRvujWGc9HdwtU8}9t$}v|ph0`su4xZ1n`cz6#9f^WHy26^fkXY7vNJvM1 zPcg<2Rip$a0J64r!r;>&w=WMLihjiU^dNNIl>MIwVhEQRwPQUkja@fWm& z%K63`Uq~VH*%M&;sjylRq`e@c2bru0Ks}%je?9D&!iz4tltf>#WQk^EkJ8Hm$WM{5 zu*oBkjAEp$E(XLEF4Zcx&D3=wt6yuw!Grn6hh-MgaydY@W;VAn5b}GaaugwOLKhg_ zhS45RS%x{yT=bJhlk|tksdZDx9`VHxlx%@uvb7VBmRQJ8Zb%CR66_2M=KNoi*Fjl45lSOhw}as}wNIUXnm2+F4MBRBJ>z`-;#9 zH*0Qa6e%&;fm@0KwcZhv4+b7(^to6CXj2lCJ(!G)fg#@1@wrIsM$nk2K&S+hU2-rK zief%7xV)p#kPV016qYv9)U^b$BuN>0{97bjt$ISh)+M$h$_9 z_i)zmHj%c*d40()Xxlp1=qOVTpEt;>Wu9^^HNLomLL}O~lszkQH=TUGl@>iKrdbC) zIeU#hnc0HXk>~o0V42N zR?T{D%EOE~H5EZ1t&>Gu*V9xL$E-P47H!>M)Ha{62@elhAR5#dxb#?2{;VWdR$y`3 zYHT3O>;`1yyh=eLkHJ(JB->`^C*%zJ@=85D#M5#Uwk2m)Ex?aZM@ z8f+J9B`ly0q|(^Lq`io&a4n|+bd}){=0zHM2#rMAWO$!>ktFKuVo(=QCIY`wyXmrD z_&sP{N!p_gs#Q)Gw#{U|Ei*DAV@aA#4e1dEK1ihJ>Q$4aU}hhkbSWp+T{ej(jds}~ z9k!b~N#)MUZc`T2H_eJPaBO|#}$<*3<=" CLEX_greatereq +#define STB_C_LEX_C_LOGICAL Y // "&&" CLEX_andand "||" CLEX_oror +#define STB_C_LEX_C_SHIFTS Y // "<<" CLEX_shl ">>" CLEX_shr +#define STB_C_LEX_C_INCREMENTS Y // "++" CLEX_plusplus "--" CLEX_minusminus +#define STB_C_LEX_C_ARROW Y // "->" CLEX_arrow +#define STB_C_LEX_EQUAL_ARROW N // "=>" CLEX_eqarrow +#define STB_C_LEX_C_BITWISEEQ Y // "&=" CLEX_andeq "|=" CLEX_oreq "^=" CLEX_xoreq +#define STB_C_LEX_C_ARITHEQ Y // "+=" CLEX_pluseq "-=" CLEX_minuseq + // "*=" CLEX_muleq "/=" CLEX_diveq "%=" CLEX_modeq + // if both STB_C_LEX_SHIFTS & STB_C_LEX_ARITHEQ: + // "<<=" CLEX_shleq ">>=" CLEX_shreq + +#define STB_C_LEX_PARSE_SUFFIXES N // letters after numbers are parsed as part of those numbers, and must be in suffix list below +#define STB_C_LEX_DECIMAL_SUFFIXES "" // decimal integer suffixes e.g. "uUlL" -- these are returned as-is in string storage +#define STB_C_LEX_HEX_SUFFIXES "" // e.g. "uUlL" +#define STB_C_LEX_OCTAL_SUFFIXES "" // e.g. "uUlL" +#define STB_C_LEX_FLOAT_SUFFIXES "" // + +#define STB_C_LEX_0_IS_EOF N // if Y, ends parsing at '\0'; if N, returns '\0' as token +#define STB_C_LEX_INTEGERS_AS_DOUBLES N // parses integers as doubles so they can be larger than 'int', but only if STB_C_LEX_STDLIB==N +#define STB_C_LEX_MULTILINE_DSTRINGS N // allow newlines in double-quoted strings +#define STB_C_LEX_MULTILINE_SSTRINGS N // allow newlines in single-quoted strings +#define STB_C_LEX_USE_STDLIB Y // use strtod,strtol for parsing #s; otherwise inaccurate hack +#define STB_C_LEX_DOLLAR_IDENTIFIER Y // allow $ as an identifier character +#define STB_C_LEX_FLOAT_NO_DECIMAL Y // allow floats that have no decimal point if they have an exponent + +#define STB_C_LEX_DEFINE_ALL_TOKEN_NAMES N // if Y, all CLEX_ token names are defined, even if never returned + // leaving it as N should help you catch config bugs + +#define STB_C_LEX_DISCARD_PREPROCESSOR Y // discard C-preprocessor directives (e.g. after prepocess + // still have #line, #pragma, etc) + +//#define STB_C_LEX_ISWHITE(str) ... // return length in bytes of whitespace characters if first char is whitespace + +#define STB_C_LEXER_DEFINITIONS // This line prevents the header file from replacing your definitions +// --END-- +#endif +#endif + +#ifndef INCLUDE_STB_C_LEXER_H +#define INCLUDE_STB_C_LEXER_H + +typedef struct +{ + // lexer variables + char *input_stream; + char *eof; + char *parse_point; + char *string_storage; + int string_storage_len; + + // lexer parse location for error messages + char *where_firstchar; + char *where_lastchar; + + // lexer token variables + long token; + double real_number; + long int_number; + char *string; + int string_len; +} stb_lexer; + +typedef struct +{ + int line_number; + int line_offset; +} stb_lex_location; + +#ifdef __cplusplus +extern "C" { +#endif + +extern void stb_c_lexer_init(stb_lexer *lexer, const char *input_stream, const char *input_stream_end, char *string_store, int store_length); +// this function initialize the 'lexer' structure +// Input: +// - input_stream points to the file to parse, loaded into memory +// - input_stream_end points to the end of the file, or NULL if you use 0-for-EOF +// - string_store is storage the lexer can use for storing parsed strings and identifiers +// - store_length is the length of that storage + +extern int stb_c_lexer_get_token(stb_lexer *lexer); +// this function returns non-zero if a token is parsed, or 0 if at EOF +// Output: +// - lexer->token is the token ID, which is unicode code point for a single-char token, < 0 for a multichar or eof or error +// - lexer->real_number is a double constant value for CLEX_floatlit, or CLEX_intlit if STB_C_LEX_INTEGERS_AS_DOUBLES +// - lexer->int_number is an integer constant for CLEX_intlit if !STB_C_LEX_INTEGERS_AS_DOUBLES, or character for CLEX_charlit +// - lexer->string is a 0-terminated string for CLEX_dqstring or CLEX_sqstring or CLEX_identifier +// - lexer->string_len is the byte length of lexer->string + +extern void stb_c_lexer_get_location(const stb_lexer *lexer, const char *where, stb_lex_location *loc); +// this inefficient function returns the line number and character offset of a +// given location in the file as returned by stb_lex_token. Because it's inefficient, +// you should only call it for errors, not for every token. +// For error messages of invalid tokens, you typically want the location of the start +// of the token (which caused the token to be invalid). For bugs involving legit +// tokens, you can report the first or the range. +// Output: +// - loc->line_number is the line number in the file, counting from 1, of the location +// - loc->line_offset is the char-offset in the line, counting from 0, of the location + + +#ifdef __cplusplus +} +#endif + +enum +{ + CLEX_eof = 256, + CLEX_parse_error, + CLEX_intlit , + CLEX_floatlit , + CLEX_id , + CLEX_dqstring , + CLEX_sqstring , + CLEX_charlit , + CLEX_eq , + CLEX_noteq , + CLEX_lesseq , + CLEX_greatereq , + CLEX_andand , + CLEX_oror , + CLEX_shl , + CLEX_shr , + CLEX_plusplus , + CLEX_minusminus , + CLEX_pluseq , + CLEX_minuseq , + CLEX_muleq , + CLEX_diveq , + CLEX_modeq , + CLEX_andeq , + CLEX_oreq , + CLEX_xoreq , + CLEX_arrow , + CLEX_eqarrow , + CLEX_shleq, CLEX_shreq, + + CLEX_first_unused_token + +}; +#endif // INCLUDE_STB_C_LEXER_H + +#ifdef STB_C_LEXER_IMPLEMENTATION + +// Hacky definitions so we can easily #if on them +#define Y(x) 1 +#define N(x) 0 + +#if STB_C_LEX_INTEGERS_AS_DOUBLES(x) +typedef double stb__clex_int; +#define intfield real_number +#define STB__clex_int_as_double +#else +typedef long stb__clex_int; +#define intfield int_number +#endif + +// Convert these config options to simple conditional #defines so we can more +// easily test them once we've change the meaning of Y/N + +#if STB_C_LEX_PARSE_SUFFIXES(x) +#define STB__clex_parse_suffixes +#endif + +#if STB_C_LEX_C99_HEX_FLOATS(x) +#define STB__clex_hex_floats +#endif + +#if STB_C_LEX_C_HEX_INTS(x) +#define STB__clex_hex_ints +#endif + +#if STB_C_LEX_C_DECIMAL_INTS(x) +#define STB__clex_decimal_ints +#endif + +#if STB_C_LEX_C_OCTAL_INTS(x) +#define STB__clex_octal_ints +#endif + +#if STB_C_LEX_C_DECIMAL_FLOATS(x) +#define STB__clex_decimal_floats +#endif + +#if STB_C_LEX_DISCARD_PREPROCESSOR(x) +#define STB__clex_discard_preprocessor +#endif + +#if STB_C_LEX_USE_STDLIB(x) && (!defined(STB__clex_hex_floats) || __STDC_VERSION__ >= 199901L) +#define STB__CLEX_use_stdlib +#include +#endif + +// Now for the rest of the file we'll use the basic definition where +// where Y expands to its contents and N expands to nothing +#undef Y +#define Y(a) a +#undef N +#define N(a) + +// API function +void stb_c_lexer_init(stb_lexer *lexer, const char *input_stream, const char *input_stream_end, char *string_store, int store_length) +{ + lexer->input_stream = (char *) input_stream; + lexer->eof = (char *) input_stream_end; + lexer->parse_point = (char *) input_stream; + lexer->string_storage = string_store; + lexer->string_storage_len = store_length; +} + +// API function +void stb_c_lexer_get_location(const stb_lexer *lexer, const char *where, stb_lex_location *loc) +{ + char *p = lexer->input_stream; + int line_number = 1; + int char_offset = 0; + while (*p && p < where) { + if (*p == '\n' || *p == '\r') { + p += (p[0]+p[1] == '\r'+'\n' ? 2 : 1); // skip newline + line_number += 1; + char_offset = 0; + } else { + ++p; + ++char_offset; + } + } + loc->line_number = line_number; + loc->line_offset = char_offset; +} + +// main helper function for returning a parsed token +static int stb__clex_token(stb_lexer *lexer, int token, char *start, char *end) +{ + lexer->token = token; + lexer->where_firstchar = start; + lexer->where_lastchar = end; + lexer->parse_point = end+1; + return 1; +} + +// helper function for returning eof +static int stb__clex_eof(stb_lexer *lexer) +{ + lexer->token = CLEX_eof; + return 0; +} + +static int stb__clex_iswhite(int x) +{ + return x == ' ' || x == '\t' || x == '\r' || x == '\n' || x == '\f'; +} + +static const char *stb__strchr(const char *str, int ch) +{ + for (; *str; ++str) + if (*str == ch) + return str; + return 0; +} + +// parse suffixes at the end of a number +static int stb__clex_parse_suffixes(stb_lexer *lexer, long tokenid, char *start, char *cur, const char *suffixes) +{ + #ifdef STB__clex_parse_suffixes + lexer->string = lexer->string_storage; + lexer->string_len = 0; + + while ((*cur >= 'a' && *cur <= 'z') || (*cur >= 'A' && *cur <= 'Z')) { + if (stb__strchr(suffixes, *cur) == 0) + return stb__clex_token(lexer, CLEX_parse_error, start, cur); + if (lexer->string_len+1 >= lexer->string_storage_len) + return stb__clex_token(lexer, CLEX_parse_error, start, cur); + lexer->string[lexer->string_len++] = *cur++; + } + #else + suffixes = suffixes; // attempt to suppress warnings + #endif + return stb__clex_token(lexer, tokenid, start, cur-1); +} + +#ifndef STB__CLEX_use_stdlib +static double stb__clex_pow(double base, unsigned int exponent) +{ + double value=1; + for ( ; exponent; exponent >>= 1) { + if (exponent & 1) + value *= base; + base *= base; + } + return value; +} + +static double stb__clex_parse_float(char *p, char **q) +{ + char *s = p; + double value=0; + int base=10; + int exponent=0; + +#ifdef STB__clex_hex_floats + if (*p == '0') { + if (p[1] == 'x' || p[1] == 'X') { + base=16; + p += 2; + } + } +#endif + + for (;;) { + if (*p >= '0' && *p <= '9') + value = value*base + (*p++ - '0'); +#ifdef STB__clex_hex_floats + else if (base == 16 && *p >= 'a' && *p <= 'f') + value = value*base + 10 + (*p++ - 'a'); + else if (base == 16 && *p >= 'A' && *p <= 'F') + value = value*base + 10 + (*p++ - 'A'); +#endif + else + break; + } + + if (*p == '.') { + double pow, addend = 0; + ++p; + for (pow=1; ; pow*=base) { + if (*p >= '0' && *p <= '9') + addend = addend*base + (*p++ - '0'); +#ifdef STB__clex_hex_floats + else if (base == 16 && *p >= 'a' && *p <= 'f') + addend = addend*base + 10 + (*p++ - 'a'); + else if (base == 16 && *p >= 'A' && *p <= 'F') + addend = addend*base + 10 + (*p++ - 'A'); +#endif + else + break; + } + value += addend / pow; + } +#ifdef STB__clex_hex_floats + if (base == 16) { + // exponent required for hex float literal + if (*p != 'p' && *p != 'P') { + *q = s; + return 0; + } + exponent = 1; + } else +#endif + exponent = (*p == 'e' || *p == 'E'); + + if (exponent) { + int sign = p[1] == '-'; + unsigned int exponent=0; + double power=1; + ++p; + if (*p == '-' || *p == '+') + ++p; + while (*p >= '0' && *p <= '9') + exponent = exponent*10 + (*p++ - '0'); + +#ifdef STB__clex_hex_floats + if (base == 16) + power = stb__clex_pow(2, exponent); + else +#endif + power = stb__clex_pow(10, exponent); + if (sign) + value /= power; + else + value *= power; + } + *q = p; + return value; +} +#endif + +static int stb__clex_parse_char(char *p, char **q) +{ + if (*p == '\\') { + *q = p+2; // tentatively guess we'll parse two characters + switch(p[1]) { + case '\\': return '\\'; + case '\'': return '\''; + case '"': return '"'; + case 't': return '\t'; + case 'f': return '\f'; + case 'n': return '\n'; + case 'r': return '\r'; + case '0': return '\0'; // @TODO ocatal constants + case 'x': case 'X': return -1; // @TODO hex constants + case 'u': return -1; // @TODO unicode constants + } + } + *q = p+1; + return (unsigned char) *p; +} + +static int stb__clex_parse_string(stb_lexer *lexer, char *p, int type) +{ + char *start = p; + char delim = *p++; // grab the " or ' for later matching + char *out = lexer->string_storage; + char *outend = lexer->string_storage + lexer->string_storage_len; + while (*p != delim) { + int n; + if (*p == '\\') { + char *q; + n = stb__clex_parse_char(p, &q); + if (n < 0) + return stb__clex_token(lexer, CLEX_parse_error, start, q); + p = q; + } else { + // @OPTIMIZE: could speed this up by looping-while-not-backslash + n = (unsigned char) *p++; + } + if (out+1 > outend) + return stb__clex_token(lexer, CLEX_parse_error, start, p); + // @TODO expand unicode escapes to UTF8 + *out++ = (char) n; + } + *out = 0; + lexer->string = lexer->string_storage; + lexer->string_len = (int) (out - lexer->string_storage); + return stb__clex_token(lexer, type, start, p); +} + +int stb_c_lexer_get_token(stb_lexer *lexer) +{ + char *p = lexer->parse_point; + + // skip whitespace and comments + for (;;) { + #ifdef STB_C_LEX_ISWHITE + while (p != lexer->stream_end) { + int n; + n = STB_C_LEX_ISWHITE(p); + if (n == 0) break; + if (lexer->eof && lexer->eof - lexer->parse_point < n) + return stb__clex_token(tok, CLEX_parse_error, p,lexer->eof-1); + p += n; + } + #else + while (p != lexer->eof && stb__clex_iswhite(*p)) + ++p; + #endif + + STB_C_LEX_CPP_COMMENTS( + if (p != lexer->eof && p[0] == '/' && p[1] == '/') { + while (p != lexer->eof && *p != '\r' && *p != '\n') + ++p; + continue; + } + ) + + STB_C_LEX_C_COMMENTS( + if (p != lexer->eof && p[0] == '/' && p[1] == '*') { + char *start = p; + p += 2; + while (p != lexer->eof && (p[0] != '*' || p[1] != '/')) + ++p; + if (p == lexer->eof) + return stb__clex_token(lexer, CLEX_parse_error, start, p-1); + p += 2; + continue; + } + ) + + #ifdef STB__clex_discard_preprocessor + // @TODO this discards everything after a '#', regardless + // of where in the line the # is, rather than requiring it + // be at the start. (because this parser doesn't otherwise + // check for line breaks!) + if (p != lexer->eof && p[0] == '#') { + while (p != lexer->eof && *p != '\r' && *p != '\n') + ++p; + continue; + } + #endif + + break; + } + + if (p == lexer->eof) + return stb__clex_eof(lexer); + + switch (*p) { + default: + if ( (*p >= 'a' && *p <= 'z') + || (*p >= 'A' && *p <= 'Z') + || *p == '_' || (unsigned char) *p >= 128 // >= 128 is UTF8 char + STB_C_LEX_DOLLAR_IDENTIFIER( || *p == '$' ) ) + { + int n = 0; + lexer->string = lexer->string_storage; + do { + if (n+1 >= lexer->string_storage_len) + return stb__clex_token(lexer, CLEX_parse_error, p, p+n); + lexer->string[n] = p[n]; + ++n; + } while ( + (p[n] >= 'a' && p[n] <= 'z') + || (p[n] >= 'A' && p[n] <= 'Z') + || (p[n] >= '0' && p[n] <= '9') // allow digits in middle of identifier + || p[n] == '_' || (unsigned char) p[n] >= 128 + STB_C_LEX_DOLLAR_IDENTIFIER( || p[n] == '$' ) + ); + lexer->string[n] = 0; + lexer->string_len = n; + return stb__clex_token(lexer, CLEX_id, p, p+n-1); + } + + // check for EOF + STB_C_LEX_0_IS_EOF( + if (*p == 0) + return stb__clex_eof(lexer); + ) + + single_char: + // not an identifier, return the character as itself + return stb__clex_token(lexer, *p, p, p); + + case '+': + if (p+1 != lexer->eof) { + STB_C_LEX_C_INCREMENTS(if (p[1] == '+') return stb__clex_token(lexer, CLEX_plusplus, p,p+1);) + STB_C_LEX_C_ARITHEQ( if (p[1] == '=') return stb__clex_token(lexer, CLEX_pluseq , p,p+1);) + } + goto single_char; + case '-': + if (p+1 != lexer->eof) { + STB_C_LEX_C_INCREMENTS(if (p[1] == '-') return stb__clex_token(lexer, CLEX_minusminus, p,p+1);) + STB_C_LEX_C_ARITHEQ( if (p[1] == '=') return stb__clex_token(lexer, CLEX_minuseq , p,p+1);) + STB_C_LEX_C_ARROW( if (p[1] == '>') return stb__clex_token(lexer, CLEX_arrow , p,p+1);) + } + goto single_char; + case '&': + if (p+1 != lexer->eof) { + STB_C_LEX_C_LOGICAL( if (p[1] == '&') return stb__clex_token(lexer, CLEX_andand, p,p+1);) + STB_C_LEX_C_BITWISEEQ(if (p[1] == '=') return stb__clex_token(lexer, CLEX_andeq , p,p+1);) + } + goto single_char; + case '|': + if (p+1 != lexer->eof) { + STB_C_LEX_C_LOGICAL( if (p[1] == '|') return stb__clex_token(lexer, CLEX_oror, p,p+1);) + STB_C_LEX_C_BITWISEEQ(if (p[1] == '=') return stb__clex_token(lexer, CLEX_oreq, p,p+1);) + } + goto single_char; + case '=': + if (p+1 != lexer->eof) { + STB_C_LEX_C_COMPARISONS(if (p[1] == '=') return stb__clex_token(lexer, CLEX_eq, p,p+1);) + STB_C_LEX_EQUAL_ARROW( if (p[1] == '>') return stb__clex_token(lexer, CLEX_eqarrow, p,p+1);) + } + goto single_char; + case '!': + STB_C_LEX_C_COMPARISONS(if (p+1 != lexer->eof && p[1] == '=') return stb__clex_token(lexer, CLEX_noteq, p,p+1);) + goto single_char; + case '^': + STB_C_LEX_C_BITWISEEQ(if (p+1 != lexer->eof && p[1] == '=') return stb__clex_token(lexer, CLEX_xoreq, p,p+1)); + goto single_char; + case '%': + STB_C_LEX_C_ARITHEQ(if (p+1 != lexer->eof && p[1] == '=') return stb__clex_token(lexer, CLEX_modeq, p,p+1)); + goto single_char; + case '*': + STB_C_LEX_C_ARITHEQ(if (p+1 != lexer->eof && p[1] == '=') return stb__clex_token(lexer, CLEX_muleq, p,p+1)); + goto single_char; + case '/': + STB_C_LEX_C_ARITHEQ(if (p+1 != lexer->eof && p[1] == '=') return stb__clex_token(lexer, CLEX_diveq, p,p+1)); + goto single_char; + case '<': + if (p+1 != lexer->eof) { + STB_C_LEX_C_COMPARISONS(if (p[1] == '=') return stb__clex_token(lexer, CLEX_lesseq, p,p+1);) + STB_C_LEX_C_SHIFTS( if (p[1] == '<') { + STB_C_LEX_C_ARITHEQ(if (p+2 != lexer->eof && p[2] == '=') + return stb__clex_token(lexer, CLEX_shleq, p,p+2);) + return stb__clex_token(lexer, CLEX_shl, p,p+1); + } + ) + } + goto single_char; + case '>': + if (p+1 != lexer->eof) { + STB_C_LEX_C_COMPARISONS(if (p[1] == '=') return stb__clex_token(lexer, CLEX_greatereq, p,p+1);) + STB_C_LEX_C_SHIFTS( if (p[1] == '>') { + STB_C_LEX_C_ARITHEQ(if (p+2 != lexer->eof && p[2] == '=') + return stb__clex_token(lexer, CLEX_shreq, p,p+2);) + return stb__clex_token(lexer, CLEX_shr, p,p+1); + } + ) + } + goto single_char; + + case '"': + STB_C_LEX_C_DQ_STRINGS(return stb__clex_parse_string(lexer, p, CLEX_dqstring);) + goto single_char; + case '\'': + STB_C_LEX_C_SQ_STRINGS(return stb__clex_parse_string(lexer, p, CLEX_sqstring);) + STB_C_LEX_C_CHARS( + { + char *start = p; + lexer->int_number = stb__clex_parse_char(p+1, &p); + if (lexer->int_number < 0) + return stb__clex_token(lexer, CLEX_parse_error, start,start); + if (p == lexer->eof || *p != '\'') + return stb__clex_token(lexer, CLEX_parse_error, start,p); + return stb__clex_token(lexer, CLEX_charlit, start, p+1); + }) + goto single_char; + + case '0': + #if defined(STB__clex_hex_ints) || defined(STB__clex_hex_floats) + if (p+1 != lexer->eof) { + if (p[1] == 'x' || p[1] == 'X') { + char *q; + + #ifdef STB__clex_hex_floats + for (q=p+2; + q != lexer->eof && ((*q >= '0' && *q <= '9') || (*q >= 'a' && *q <= 'f') || (*q >= 'A' && *q <= 'F')); + ++q); + if (q != lexer->eof) { + if (*q == '.' STB_C_LEX_FLOAT_NO_DECIMAL(|| *q == 'p' || *q == 'P')) { + #ifdef STB__CLEX_use_stdlib + lexer->real_number = strtod((char *) p, (char**) &q); + #else + lexer->real_number = stb__clex_parse_float(p, &q); + #endif + + if (p == q) + return stb__clex_token(lexer, CLEX_parse_error, p,q); + return stb__clex_parse_suffixes(lexer, CLEX_floatlit, p,q, STB_C_LEX_FLOAT_SUFFIXES); + + } + } + #endif // STB__CLEX_hex_floats + + #ifdef STB__clex_hex_ints + #ifdef STB__CLEX_use_stdlib + lexer->int_number = strtol((char *) p, (char **) &q, 16); + #else + { + stb__clex_int n=0; + for (q=p+2; q != lexer->eof; ++q) { + if (*q >= '0' && *q <= '9') + n = n*16 + (*q - '0'); + else if (*q >= 'a' && *q <= 'f') + n = n*16 + (*q - 'a') + 10; + else if (*q >= 'A' && *q <= 'F') + n = n*16 + (*q - 'A') + 10; + else + break; + } + lexer->int_number = n; + } + #endif + if (q == p+2) + return stb__clex_token(lexer, CLEX_parse_error, p-2,p-1); + return stb__clex_parse_suffixes(lexer, CLEX_intlit, p,q, STB_C_LEX_HEX_SUFFIXES); + #endif + } + } + #endif // defined(STB__clex_hex_ints) || defined(STB__clex_hex_floats) + // can't test for octal because we might parse '0.0' as float or as '0' '.' '0', + // so have to do float first + + /* FALL THROUGH */ + case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': + + #ifdef STB__clex_decimal_floats + { + char *q = p; + while (q != lexer->eof && (*q >= '0' && *q <= '9')) + ++q; + if (q != lexer->eof) { + if (*q == '.' STB_C_LEX_FLOAT_NO_DECIMAL(|| *q == 'e' || *q == 'E')) { + #ifdef STB__CLEX_use_stdlib + lexer->real_number = strtod((char *) p, (char**) &q); + #else + lexer->real_number = stb__clex_parse_float(p, &q); + #endif + + return stb__clex_parse_suffixes(lexer, CLEX_floatlit, p,q, STB_C_LEX_FLOAT_SUFFIXES); + + } + } + } + #endif // STB__clex_decimal_floats + + #ifdef STB__clex_octal_ints + if (p[0] == '0') { + char *q = p; + #ifdef STB__CLEX_use_stdlib + lexer->int_number = strtol((char *) p, (char **) &q, 8); + #else + stb__clex_int n=0; + while (q != lexer->eof) { + if (*q >= '0' && *q <= '7') + n = n*8 + (*q - '0'); + else + break; + ++q; + } + if (q != lexer->eof && (*q == '8' || *q=='9')) + return stb__clex_token(lexer, CLEX_parse_error, p, q); + lexer->int_number = n; + #endif + return stb__clex_parse_suffixes(lexer, CLEX_intlit, p,q, STB_C_LEX_OCTAL_SUFFIXES); + } + #endif // STB__clex_octal_ints + + #ifdef STB__clex_decimal_ints + { + char *q = p; + #ifdef STB__CLEX_use_stdlib + lexer->int_number = strtol((char *) p, (char **) &q, 10); + #else + stb__clex_int n=0; + while (q != lexer->eof) { + if (*q >= '0' && *q <= '9') + n = n*10 + (*q - '0'); + else + break; + ++q; + } + lexer->int_number = n; + #endif + return stb__clex_parse_suffixes(lexer, CLEX_intlit, p,q, STB_C_LEX_OCTAL_SUFFIXES); + } + #endif // STB__clex_decimal_ints + goto single_char; + } +} +#endif // STB_C_LEXER_IMPLEMENTATION + +#ifdef STB_C_LEXER_SELF_TEST +#define _CRT_SECURE_NO_WARNINGS +#include +#include + +static void print_token(stb_lexer *lexer) +{ + switch (lexer->token) { + case CLEX_id : printf("_%s", lexer->string); break; + case CLEX_eq : printf("=="); break; + case CLEX_noteq : printf("!="); break; + case CLEX_lesseq : printf("<="); break; + case CLEX_greatereq : printf(">="); break; + case CLEX_andand : printf("&&"); break; + case CLEX_oror : printf("||"); break; + case CLEX_shl : printf("<<"); break; + case CLEX_shr : printf(">>"); break; + case CLEX_plusplus : printf("++"); break; + case CLEX_minusminus: printf("--"); break; + case CLEX_arrow : printf("->"); break; + case CLEX_andeq : printf("&="); break; + case CLEX_oreq : printf("|="); break; + case CLEX_xoreq : printf("^="); break; + case CLEX_pluseq : printf("+="); break; + case CLEX_minuseq : printf("-="); break; + case CLEX_muleq : printf("*="); break; + case CLEX_diveq : printf("/="); break; + case CLEX_modeq : printf("%%="); break; + case CLEX_shleq : printf("<<="); break; + case CLEX_shreq : printf(">>="); break; + case CLEX_eqarrow : printf("=>"); break; + case CLEX_dqstring : printf("\"%s\"", lexer->string); break; + case CLEX_sqstring : printf("'\"%s\"'", lexer->string); break; + case CLEX_charlit : printf("'%s'", lexer->string); break; + #if defined(STB__clex_int_as_double) && !defined(STB__CLEX_use_stdlib) + case CLEX_intlit : printf("#%g", lexer->real_number); break; + #else + case CLEX_intlit : printf("#%ld", lexer->int_number); break; + #endif + case CLEX_floatlit : printf("%g", lexer->real_number); break; + default: + if (lexer->token >= 0 && lexer->token < 256) + printf("%c", (int) lexer->token); + else { + printf("<<>>\n", lexer->token); + } + break; + } +} + +/* Force a test +of parsing +multiline comments */ + +/*/ comment /*/ +/**/ extern /**/ + +void dummy(void) +{ + double some_floats[] = { + 1.0501, -10.4e12, 5E+10, +#if 0 // not supported in C++ or C-pre-99, so don't try to compile it, but let our parser test it + 0x1.0p+24, 0xff.FP-8, 0x1p-23, +#endif + 4. + }; + (void) sizeof(some_floats); + (void) some_floats[1]; + + printf("test %d",1); // https://github.com/nothings/stb/issues/13 +} + +int main(int argc, char **argv) +{ + FILE *f = fopen("stb_c_lexer.h","rb"); + char *text = (char *) malloc(1 << 20); + int len = f ? (int) fread(text, 1, 1<<20, f) : -1; + stb_lexer lex; + if (len < 0) { + fprintf(stderr, "Error opening file\n"); + free(text); + fclose(f); + return 1; + } + fclose(f); + + stb_c_lexer_init(&lex, text, text+len, (char *) malloc(0x10000), 0x10000); + while (stb_c_lexer_get_token(&lex)) { + if (lex.token == CLEX_parse_error) { + printf("\n<<>>\n"); + break; + } + print_token(&lex); + printf(" "); + } + return 0; +} +#endif +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_connected_components.h b/include/stb_connected_components.h new file mode 100644 index 0000000..f762f65 --- /dev/null +++ b/include/stb_connected_components.h @@ -0,0 +1,1049 @@ +// stb_connected_components - v0.96 - public domain connected components on grids +// http://github.com/nothings/stb +// +// Finds connected components on 2D grids for testing reachability between +// two points, with fast updates when changing reachability (e.g. on one machine +// it was typically 0.2ms w/ 1024x1024 grid). Each grid square must be "open" or +// "closed" (traversable or untraversable), and grid squares are only connected +// to their orthogonal neighbors, not diagonally. +// +// In one source file, create the implementation by doing something like this: +// +// #define STBCC_GRID_COUNT_X_LOG2 10 +// #define STBCC_GRID_COUNT_Y_LOG2 10 +// #define STB_CONNECTED_COMPONENTS_IMPLEMENTATION +// #include "stb_connected_components.h" +// +// The above creates an implementation that can run on maps up to 1024x1024. +// Map sizes must be a multiple of (1<<(LOG2/2)) on each axis (e.g. 32 if LOG2=10, +// 16 if LOG2=8, etc.) (You can just pad your map with untraversable space.) +// +// MEMORY USAGE +// +// Uses about 6-7 bytes per grid square (e.g. 7MB for a 1024x1024 grid). +// Uses a single worst-case allocation which you pass in. +// +// PERFORMANCE +// +// On a core i7-2700K at 3.5 Ghz, for a particular 1024x1024 map (map_03.png): +// +// Creating map : 44.85 ms +// Making one square traversable: 0.27 ms (average over 29,448 calls) +// Making one square untraversable: 0.23 ms (average over 30,123 calls) +// Reachability query: 0.00001 ms (average over 4,000,000 calls) +// +// On non-degenerate maps update time is O(N^0.5), but on degenerate maps like +// checkerboards or 50% random, update time is O(N^0.75) (~2ms on above machine). +// +// CHANGELOG +// +// 0.96 (2019-03-04) Fix warnings +// 0.95 (2016-10-16) Bugfix if multiple clumps in one cluster connect to same clump in another +// 0.94 (2016-04-17) Bugfix & optimize worst case (checkerboard & random) +// 0.93 (2016-04-16) Reduce memory by 10x for 1Kx1K map; small speedup +// 0.92 (2016-04-16) Compute sqrt(N) cluster size by default +// 0.91 (2016-04-15) Initial release +// +// TODO: +// - better API documentation +// - more comments +// - try re-integrating naive algorithm & compare performance +// - more optimized batching (current approach still recomputes local clumps many times) +// - function for setting a grid of squares at once (just use batching) +// +// LICENSE +// +// See end of file for license information. +// +// ALGORITHM +// +// The NxN grid map is split into sqrt(N) x sqrt(N) blocks called +// "clusters". Each cluster independently computes a set of connected +// components within that cluster (ignoring all connectivity out of +// that cluster) using a union-find disjoint set forest. This produces a bunch +// of locally connected components called "clumps". Each clump is (a) connected +// within its cluster, (b) does not directly connect to any other clumps in the +// cluster (though it may connect to them by paths that lead outside the cluster, +// but those are ignored at this step), and (c) maintains an adjacency list of +// all clumps in adjacent clusters that it _is_ connected to. Then a second +// union-find disjoint set forest is used to compute connected clumps +// globally, across the whole map. Reachability is then computed by +// finding which clump each input point belongs to, and checking whether +// those clumps are in the same "global" connected component. +// +// The above data structure can be updated efficiently; on a change +// of a single grid square on the map, only one cluster changes its +// purely-local state, so only one cluster needs its clumps fully +// recomputed. Clumps in adjacent clusters need their adjacency lists +// updated: first to remove all references to the old clumps in the +// rebuilt cluster, then to add new references to the new clumps. Both +// of these operations can use the existing "find which clump each input +// point belongs to" query to compute that adjacency information rapidly. + +#ifndef INCLUDE_STB_CONNECTED_COMPONENTS_H +#define INCLUDE_STB_CONNECTED_COMPONENTS_H + +#include + +typedef struct st_stbcc_grid stbcc_grid; + +#ifdef __cplusplus +extern "C" { +#endif + +////////////////////////////////////////////////////////////////////////////////////////// +// +// initialization +// + +// you allocate the grid data structure to this size (note that it will be very big!!!) +extern size_t stbcc_grid_sizeof(void); + +// initialize the grid, value of map[] is 0 = traversable, non-0 is solid +extern void stbcc_init_grid(stbcc_grid *g, unsigned char *map, int w, int h); + + +////////////////////////////////////////////////////////////////////////////////////////// +// +// main functionality +// + +// update a grid square state, 0 = traversable, non-0 is solid +// i can add a batch-update if it's needed +extern void stbcc_update_grid(stbcc_grid *g, int x, int y, int solid); + +// query if two grid squares are reachable from each other +extern int stbcc_query_grid_node_connection(stbcc_grid *g, int x1, int y1, int x2, int y2); + + +////////////////////////////////////////////////////////////////////////////////////////// +// +// bonus functions +// + +// wrap multiple stbcc_update_grid calls in these function to compute +// multiple updates more efficiently; cannot make queries inside batch +extern void stbcc_update_batch_begin(stbcc_grid *g); +extern void stbcc_update_batch_end(stbcc_grid *g); + +// query the grid data structure for whether a given square is open or not +extern int stbcc_query_grid_open(stbcc_grid *g, int x, int y); + +// get a unique id for the connected component this is in; it's not necessarily +// small, you'll need a hash table or something to remap it (or just use +extern unsigned int stbcc_get_unique_id(stbcc_grid *g, int x, int y); +#define STBCC_NULL_UNIQUE_ID 0xffffffff // returned for closed map squares + +#ifdef __cplusplus +} +#endif + +#endif // INCLUDE_STB_CONNECTED_COMPONENTS_H + +#ifdef STB_CONNECTED_COMPONENTS_IMPLEMENTATION + +#include +#include // memset + +#if !defined(STBCC_GRID_COUNT_X_LOG2) || !defined(STBCC_GRID_COUNT_Y_LOG2) + #error "You must define STBCC_GRID_COUNT_X_LOG2 and STBCC_GRID_COUNT_Y_LOG2 to define the max grid supported." +#endif + +#define STBCC__GRID_COUNT_X (1 << STBCC_GRID_COUNT_X_LOG2) +#define STBCC__GRID_COUNT_Y (1 << STBCC_GRID_COUNT_Y_LOG2) + +#define STBCC__MAP_STRIDE (1 << (STBCC_GRID_COUNT_X_LOG2-3)) + +#ifndef STBCC_CLUSTER_SIZE_X_LOG2 + #define STBCC_CLUSTER_SIZE_X_LOG2 (STBCC_GRID_COUNT_X_LOG2/2) // log2(sqrt(2^N)) = 1/2 * log2(2^N)) = 1/2 * N + #if STBCC_CLUSTER_SIZE_X_LOG2 > 6 + #undef STBCC_CLUSTER_SIZE_X_LOG2 + #define STBCC_CLUSTER_SIZE_X_LOG2 6 + #endif +#endif + +#ifndef STBCC_CLUSTER_SIZE_Y_LOG2 + #define STBCC_CLUSTER_SIZE_Y_LOG2 (STBCC_GRID_COUNT_Y_LOG2/2) + #if STBCC_CLUSTER_SIZE_Y_LOG2 > 6 + #undef STBCC_CLUSTER_SIZE_Y_LOG2 + #define STBCC_CLUSTER_SIZE_Y_LOG2 6 + #endif +#endif + +#define STBCC__CLUSTER_SIZE_X (1 << STBCC_CLUSTER_SIZE_X_LOG2) +#define STBCC__CLUSTER_SIZE_Y (1 << STBCC_CLUSTER_SIZE_Y_LOG2) + +#define STBCC__CLUSTER_COUNT_X_LOG2 (STBCC_GRID_COUNT_X_LOG2 - STBCC_CLUSTER_SIZE_X_LOG2) +#define STBCC__CLUSTER_COUNT_Y_LOG2 (STBCC_GRID_COUNT_Y_LOG2 - STBCC_CLUSTER_SIZE_Y_LOG2) + +#define STBCC__CLUSTER_COUNT_X (1 << STBCC__CLUSTER_COUNT_X_LOG2) +#define STBCC__CLUSTER_COUNT_Y (1 << STBCC__CLUSTER_COUNT_Y_LOG2) + +#if STBCC__CLUSTER_SIZE_X >= STBCC__GRID_COUNT_X || STBCC__CLUSTER_SIZE_Y >= STBCC__GRID_COUNT_Y + #error "STBCC_CLUSTER_SIZE_X/Y_LOG2 must be smaller than STBCC_GRID_COUNT_X/Y_LOG2" +#endif + +// worst case # of clumps per cluster +#define STBCC__MAX_CLUMPS_PER_CLUSTER_LOG2 (STBCC_CLUSTER_SIZE_X_LOG2 + STBCC_CLUSTER_SIZE_Y_LOG2-1) +#define STBCC__MAX_CLUMPS_PER_CLUSTER (1 << STBCC__MAX_CLUMPS_PER_CLUSTER_LOG2) +#define STBCC__MAX_CLUMPS (STBCC__MAX_CLUMPS_PER_CLUSTER * STBCC__CLUSTER_COUNT_X * STBCC__CLUSTER_COUNT_Y) +#define STBCC__NULL_CLUMPID STBCC__MAX_CLUMPS_PER_CLUSTER + +#define STBCC__CLUSTER_X_FOR_COORD_X(x) ((x) >> STBCC_CLUSTER_SIZE_X_LOG2) +#define STBCC__CLUSTER_Y_FOR_COORD_Y(y) ((y) >> STBCC_CLUSTER_SIZE_Y_LOG2) + +#define STBCC__MAP_BYTE_MASK(x,y) (1 << ((x) & 7)) +#define STBCC__MAP_BYTE(g,x,y) ((g)->map[y][(x) >> 3]) +#define STBCC__MAP_OPEN(g,x,y) (STBCC__MAP_BYTE(g,x,y) & STBCC__MAP_BYTE_MASK(x,y)) + +typedef unsigned short stbcc__clumpid; +typedef unsigned char stbcc__verify_max_clumps[STBCC__MAX_CLUMPS_PER_CLUSTER < (1 << (8*sizeof(stbcc__clumpid))) ? 1 : -1]; + +#define STBCC__MAX_EXITS_PER_CLUSTER (STBCC__CLUSTER_SIZE_X + STBCC__CLUSTER_SIZE_Y) // 64 for 32x32 +#define STBCC__MAX_EXITS_PER_CLUMP (STBCC__CLUSTER_SIZE_X + STBCC__CLUSTER_SIZE_Y) // 64 for 32x32 +#define STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER (STBCC__MAX_EXITS_PER_CLUMP) + +// 2^19 * 2^6 => 2^25 exits => 2^26 => 64MB for 1024x1024 + +// Logic for above on 4x4 grid: +// +// Many clumps: One clump: +// + + + + +// +X.X. +XX.X+ +// .X.X+ .XXX +// +X.X. XXX. +// .X.X+ +X.XX+ +// + + + + +// +// 8 exits either way + +typedef unsigned char stbcc__verify_max_exits[STBCC__MAX_EXITS_PER_CLUMP <= 256]; + +typedef struct +{ + unsigned short clump_index:12; + signed short cluster_dx:2; + signed short cluster_dy:2; +} stbcc__relative_clumpid; + +typedef union +{ + struct { + unsigned int clump_index:12; + unsigned int cluster_x:10; + unsigned int cluster_y:10; + } f; + unsigned int c; +} stbcc__global_clumpid; + +// rebuilt cluster 3,4 + +// what changes in cluster 2,4 + +typedef struct +{ + stbcc__global_clumpid global_label; // 4 + unsigned char num_adjacent; // 1 + unsigned char max_adjacent; // 1 + unsigned char adjacent_clump_list_index; // 1 + unsigned char reserved; +} stbcc__clump; // 8 + +#define STBCC__CLUSTER_ADJACENCY_COUNT (STBCC__MAX_EXITS_PER_CLUSTER*2) +typedef struct +{ + short num_clumps; + unsigned char num_edge_clumps; + unsigned char rebuild_adjacency; + stbcc__clump clump[STBCC__MAX_CLUMPS_PER_CLUSTER]; // 8 * 2^9 = 4KB + stbcc__relative_clumpid adjacency_storage[STBCC__CLUSTER_ADJACENCY_COUNT]; // 256 bytes +} stbcc__cluster; + +struct st_stbcc_grid +{ + int w,h,cw,ch; + int in_batched_update; + //unsigned char cluster_dirty[STBCC__CLUSTER_COUNT_Y][STBCC__CLUSTER_COUNT_X]; // could bitpack, but: 1K x 1K => 1KB + unsigned char map[STBCC__GRID_COUNT_Y][STBCC__MAP_STRIDE]; // 1K x 1K => 1K x 128 => 128KB + stbcc__clumpid clump_for_node[STBCC__GRID_COUNT_Y][STBCC__GRID_COUNT_X]; // 1K x 1K x 2 = 2MB + stbcc__cluster cluster[STBCC__CLUSTER_COUNT_Y][STBCC__CLUSTER_COUNT_X]; // 1K x 4.5KB = 4.5MB +}; + +int stbcc_query_grid_node_connection(stbcc_grid *g, int x1, int y1, int x2, int y2) +{ + stbcc__global_clumpid label1, label2; + stbcc__clumpid c1 = g->clump_for_node[y1][x1]; + stbcc__clumpid c2 = g->clump_for_node[y2][x2]; + int cx1 = STBCC__CLUSTER_X_FOR_COORD_X(x1); + int cy1 = STBCC__CLUSTER_Y_FOR_COORD_Y(y1); + int cx2 = STBCC__CLUSTER_X_FOR_COORD_X(x2); + int cy2 = STBCC__CLUSTER_Y_FOR_COORD_Y(y2); + assert(!g->in_batched_update); + if (c1 == STBCC__NULL_CLUMPID || c2 == STBCC__NULL_CLUMPID) + return 0; + label1 = g->cluster[cy1][cx1].clump[c1].global_label; + label2 = g->cluster[cy2][cx2].clump[c2].global_label; + if (label1.c == label2.c) + return 1; + return 0; +} + +int stbcc_query_grid_open(stbcc_grid *g, int x, int y) +{ + return STBCC__MAP_OPEN(g, x, y) != 0; +} + +unsigned int stbcc_get_unique_id(stbcc_grid *g, int x, int y) +{ + stbcc__clumpid c = g->clump_for_node[y][x]; + int cx = STBCC__CLUSTER_X_FOR_COORD_X(x); + int cy = STBCC__CLUSTER_Y_FOR_COORD_Y(y); + assert(!g->in_batched_update); + if (c == STBCC__NULL_CLUMPID) return STBCC_NULL_UNIQUE_ID; + return g->cluster[cy][cx].clump[c].global_label.c; +} + +typedef struct +{ + unsigned char x,y; +} stbcc__tinypoint; + +typedef struct +{ + stbcc__tinypoint parent[STBCC__CLUSTER_SIZE_Y][STBCC__CLUSTER_SIZE_X]; // 32x32 => 2KB + stbcc__clumpid label[STBCC__CLUSTER_SIZE_Y][STBCC__CLUSTER_SIZE_X]; +} stbcc__cluster_build_info; + +static void stbcc__build_clumps_for_cluster(stbcc_grid *g, int cx, int cy); +static void stbcc__remove_connections_to_adjacent_cluster(stbcc_grid *g, int cx, int cy, int dx, int dy); +static void stbcc__add_connections_to_adjacent_cluster(stbcc_grid *g, int cx, int cy, int dx, int dy); + +static stbcc__global_clumpid stbcc__clump_find(stbcc_grid *g, stbcc__global_clumpid n) +{ + stbcc__global_clumpid q; + stbcc__clump *c = &g->cluster[n.f.cluster_y][n.f.cluster_x].clump[n.f.clump_index]; + + if (c->global_label.c == n.c) + return n; + + q = stbcc__clump_find(g, c->global_label); + c->global_label = q; + return q; +} + +typedef struct +{ + unsigned int cluster_x; + unsigned int cluster_y; + unsigned int clump_index; +} stbcc__unpacked_clumpid; + +static void stbcc__clump_union(stbcc_grid *g, stbcc__unpacked_clumpid m, int x, int y, int idx) +{ + stbcc__clump *mc = &g->cluster[m.cluster_y][m.cluster_x].clump[m.clump_index]; + stbcc__clump *nc = &g->cluster[y][x].clump[idx]; + stbcc__global_clumpid mp = stbcc__clump_find(g, mc->global_label); + stbcc__global_clumpid np = stbcc__clump_find(g, nc->global_label); + + if (mp.c == np.c) + return; + + g->cluster[mp.f.cluster_y][mp.f.cluster_x].clump[mp.f.clump_index].global_label = np; +} + +static void stbcc__build_connected_components_for_clumps(stbcc_grid *g) +{ + int i,j,k,h; + + for (j=0; j < STBCC__CLUSTER_COUNT_Y; ++j) { + for (i=0; i < STBCC__CLUSTER_COUNT_X; ++i) { + stbcc__cluster *cluster = &g->cluster[j][i]; + for (k=0; k < (int) cluster->num_edge_clumps; ++k) { + stbcc__global_clumpid m; + m.f.clump_index = k; + m.f.cluster_x = i; + m.f.cluster_y = j; + assert((int) m.f.clump_index == k && (int) m.f.cluster_x == i && (int) m.f.cluster_y == j); + cluster->clump[k].global_label = m; + } + } + } + + for (j=0; j < STBCC__CLUSTER_COUNT_Y; ++j) { + for (i=0; i < STBCC__CLUSTER_COUNT_X; ++i) { + stbcc__cluster *cluster = &g->cluster[j][i]; + for (k=0; k < (int) cluster->num_edge_clumps; ++k) { + stbcc__clump *clump = &cluster->clump[k]; + stbcc__unpacked_clumpid m; + stbcc__relative_clumpid *adj; + m.clump_index = k; + m.cluster_x = i; + m.cluster_y = j; + adj = &cluster->adjacency_storage[clump->adjacent_clump_list_index]; + for (h=0; h < clump->num_adjacent; ++h) { + unsigned int clump_index = adj[h].clump_index; + unsigned int x = adj[h].cluster_dx + i; + unsigned int y = adj[h].cluster_dy + j; + stbcc__clump_union(g, m, x, y, clump_index); + } + } + } + } + + for (j=0; j < STBCC__CLUSTER_COUNT_Y; ++j) { + for (i=0; i < STBCC__CLUSTER_COUNT_X; ++i) { + stbcc__cluster *cluster = &g->cluster[j][i]; + for (k=0; k < (int) cluster->num_edge_clumps; ++k) { + stbcc__global_clumpid m; + m.f.clump_index = k; + m.f.cluster_x = i; + m.f.cluster_y = j; + stbcc__clump_find(g, m); + } + } + } +} + +static void stbcc__build_all_connections_for_cluster(stbcc_grid *g, int cx, int cy) +{ + // in this particular case, we are fully non-incremental. that means we + // can discover the correct sizes for the arrays, but requires we build + // the data into temporary data structures, or just count the sizes, so + // for simplicity we do the latter + stbcc__cluster *cluster = &g->cluster[cy][cx]; + unsigned char connected[STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER][STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER/8]; // 64 x 8 => 1KB + unsigned char num_adj[STBCC__MAX_CLUMPS_PER_CLUSTER] = { 0 }; + int x = cx * STBCC__CLUSTER_SIZE_X; + int y = cy * STBCC__CLUSTER_SIZE_Y; + int step_x, step_y=0, i, j, k, n, m, dx, dy, total; + int extra; + + g->cluster[cy][cx].rebuild_adjacency = 0; + + total = 0; + for (m=0; m < 4; ++m) { + switch (m) { + case 0: + dx = 1, dy = 0; + step_x = 0, step_y = 1; + i = STBCC__CLUSTER_SIZE_X-1; + j = 0; + n = STBCC__CLUSTER_SIZE_Y; + break; + case 1: + dx = -1, dy = 0; + i = 0; + j = 0; + step_x = 0; + step_y = 1; + n = STBCC__CLUSTER_SIZE_Y; + break; + case 2: + dy = -1, dx = 0; + i = 0; + j = 0; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + break; + case 3: + dy = 1, dx = 0; + i = 0; + j = STBCC__CLUSTER_SIZE_Y-1; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + break; + } + + if (cx+dx < 0 || cx+dx >= g->cw || cy+dy < 0 || cy+dy >= g->ch) + continue; + + memset(connected, 0, sizeof(connected)); + for (k=0; k < n; ++k) { + if (STBCC__MAP_OPEN(g, x+i, y+j) && STBCC__MAP_OPEN(g, x+i+dx, y+j+dy)) { + stbcc__clumpid src = g->clump_for_node[y+j][x+i]; + stbcc__clumpid dest = g->clump_for_node[y+j+dy][x+i+dx]; + if (0 == (connected[src][dest>>3] & (1 << (dest & 7)))) { + connected[src][dest>>3] |= 1 << (dest & 7); + ++num_adj[src]; + ++total; + } + } + i += step_x; + j += step_y; + } + } + + assert(total <= STBCC__CLUSTER_ADJACENCY_COUNT); + + // decide how to apportion unused adjacency slots; only clumps that lie + // on the edges of the cluster need adjacency slots, so divide them up + // evenly between those clumps + + // we want: + // extra = (STBCC__CLUSTER_ADJACENCY_COUNT - total) / cluster->num_edge_clumps; + // but we efficiently approximate this without a divide, because + // ignoring edge-vs-non-edge with 'num_adj[i]*2' was faster than + // 'num_adj[i]+extra' with the divide + if (total + (cluster->num_edge_clumps<<2) <= STBCC__CLUSTER_ADJACENCY_COUNT) + extra = 4; + else if (total + (cluster->num_edge_clumps<<1) <= STBCC__CLUSTER_ADJACENCY_COUNT) + extra = 2; + else if (total + (cluster->num_edge_clumps<<0) <= STBCC__CLUSTER_ADJACENCY_COUNT) + extra = 1; + else + extra = 0; + + total = 0; + for (i=0; i < (int) cluster->num_edge_clumps; ++i) { + int alloc = num_adj[i]+extra; + if (alloc > STBCC__MAX_EXITS_PER_CLUSTER) + alloc = STBCC__MAX_EXITS_PER_CLUSTER; + assert(total < 256); // must fit in byte + cluster->clump[i].adjacent_clump_list_index = (unsigned char) total; + cluster->clump[i].max_adjacent = alloc; + cluster->clump[i].num_adjacent = 0; + total += alloc; + } + assert(total <= STBCC__CLUSTER_ADJACENCY_COUNT); + + stbcc__add_connections_to_adjacent_cluster(g, cx, cy, -1, 0); + stbcc__add_connections_to_adjacent_cluster(g, cx, cy, 1, 0); + stbcc__add_connections_to_adjacent_cluster(g, cx, cy, 0,-1); + stbcc__add_connections_to_adjacent_cluster(g, cx, cy, 0, 1); + // make sure all of the above succeeded. + assert(g->cluster[cy][cx].rebuild_adjacency == 0); +} + +static void stbcc__add_connections_to_adjacent_cluster_with_rebuild(stbcc_grid *g, int cx, int cy, int dx, int dy) +{ + if (cx >= 0 && cx < g->cw && cy >= 0 && cy < g->ch) { + stbcc__add_connections_to_adjacent_cluster(g, cx, cy, dx, dy); + if (g->cluster[cy][cx].rebuild_adjacency) + stbcc__build_all_connections_for_cluster(g, cx, cy); + } +} + +void stbcc_update_grid(stbcc_grid *g, int x, int y, int solid) +{ + int cx,cy; + + if (!solid) { + if (STBCC__MAP_OPEN(g,x,y)) + return; + } else { + if (!STBCC__MAP_OPEN(g,x,y)) + return; + } + + cx = STBCC__CLUSTER_X_FOR_COORD_X(x); + cy = STBCC__CLUSTER_Y_FOR_COORD_Y(y); + + stbcc__remove_connections_to_adjacent_cluster(g, cx-1, cy, 1, 0); + stbcc__remove_connections_to_adjacent_cluster(g, cx+1, cy, -1, 0); + stbcc__remove_connections_to_adjacent_cluster(g, cx, cy-1, 0, 1); + stbcc__remove_connections_to_adjacent_cluster(g, cx, cy+1, 0,-1); + + if (!solid) + STBCC__MAP_BYTE(g,x,y) |= STBCC__MAP_BYTE_MASK(x,y); + else + STBCC__MAP_BYTE(g,x,y) &= ~STBCC__MAP_BYTE_MASK(x,y); + + stbcc__build_clumps_for_cluster(g, cx, cy); + stbcc__build_all_connections_for_cluster(g, cx, cy); + + stbcc__add_connections_to_adjacent_cluster_with_rebuild(g, cx-1, cy, 1, 0); + stbcc__add_connections_to_adjacent_cluster_with_rebuild(g, cx+1, cy, -1, 0); + stbcc__add_connections_to_adjacent_cluster_with_rebuild(g, cx, cy-1, 0, 1); + stbcc__add_connections_to_adjacent_cluster_with_rebuild(g, cx, cy+1, 0,-1); + + if (!g->in_batched_update) + stbcc__build_connected_components_for_clumps(g); + #if 0 + else + g->cluster_dirty[cy][cx] = 1; + #endif +} + +void stbcc_update_batch_begin(stbcc_grid *g) +{ + assert(!g->in_batched_update); + g->in_batched_update = 1; +} + +void stbcc_update_batch_end(stbcc_grid *g) +{ + assert(g->in_batched_update); + g->in_batched_update = 0; + stbcc__build_connected_components_for_clumps(g); // @OPTIMIZE: only do this if update was non-empty +} + +size_t stbcc_grid_sizeof(void) +{ + return sizeof(stbcc_grid); +} + +void stbcc_init_grid(stbcc_grid *g, unsigned char *map, int w, int h) +{ + int i,j,k; + assert(w % STBCC__CLUSTER_SIZE_X == 0); + assert(h % STBCC__CLUSTER_SIZE_Y == 0); + assert(w % 8 == 0); + + g->w = w; + g->h = h; + g->cw = w >> STBCC_CLUSTER_SIZE_X_LOG2; + g->ch = h >> STBCC_CLUSTER_SIZE_Y_LOG2; + g->in_batched_update = 0; + + #if 0 + for (j=0; j < STBCC__CLUSTER_COUNT_Y; ++j) + for (i=0; i < STBCC__CLUSTER_COUNT_X; ++i) + g->cluster_dirty[j][i] = 0; + #endif + + for (j=0; j < h; ++j) { + for (i=0; i < w; i += 8) { + unsigned char c = 0; + for (k=0; k < 8; ++k) + if (map[j*w + (i+k)] == 0) + c |= (1 << k); + g->map[j][i>>3] = c; + } + } + + for (j=0; j < g->ch; ++j) + for (i=0; i < g->cw; ++i) + stbcc__build_clumps_for_cluster(g, i, j); + + for (j=0; j < g->ch; ++j) + for (i=0; i < g->cw; ++i) + stbcc__build_all_connections_for_cluster(g, i, j); + + stbcc__build_connected_components_for_clumps(g); + + for (j=0; j < g->h; ++j) + for (i=0; i < g->w; ++i) + assert(g->clump_for_node[j][i] <= STBCC__NULL_CLUMPID); +} + + +static void stbcc__add_clump_connection(stbcc_grid *g, int x1, int y1, int x2, int y2) +{ + stbcc__cluster *cluster; + stbcc__clump *clump; + + int cx1 = STBCC__CLUSTER_X_FOR_COORD_X(x1); + int cy1 = STBCC__CLUSTER_Y_FOR_COORD_Y(y1); + int cx2 = STBCC__CLUSTER_X_FOR_COORD_X(x2); + int cy2 = STBCC__CLUSTER_Y_FOR_COORD_Y(y2); + + stbcc__clumpid c1 = g->clump_for_node[y1][x1]; + stbcc__clumpid c2 = g->clump_for_node[y2][x2]; + + stbcc__relative_clumpid rc; + + assert(cx1 != cx2 || cy1 != cy2); + assert(abs(cx1-cx2) + abs(cy1-cy2) == 1); + + // add connection to c2 in c1 + + rc.clump_index = c2; + rc.cluster_dx = x2-x1; + rc.cluster_dy = y2-y1; + + cluster = &g->cluster[cy1][cx1]; + clump = &cluster->clump[c1]; + assert(clump->num_adjacent <= clump->max_adjacent); + if (clump->num_adjacent == clump->max_adjacent) + g->cluster[cy1][cx1].rebuild_adjacency = 1; + else { + stbcc__relative_clumpid *adj = &cluster->adjacency_storage[clump->adjacent_clump_list_index]; + assert(clump->num_adjacent < STBCC__MAX_EXITS_PER_CLUMP); + assert(clump->adjacent_clump_list_index + clump->num_adjacent <= STBCC__CLUSTER_ADJACENCY_COUNT); + adj[clump->num_adjacent++] = rc; + } +} + +static void stbcc__remove_clump_connection(stbcc_grid *g, int x1, int y1, int x2, int y2) +{ + stbcc__cluster *cluster; + stbcc__clump *clump; + stbcc__relative_clumpid *adj; + int i; + + int cx1 = STBCC__CLUSTER_X_FOR_COORD_X(x1); + int cy1 = STBCC__CLUSTER_Y_FOR_COORD_Y(y1); + int cx2 = STBCC__CLUSTER_X_FOR_COORD_X(x2); + int cy2 = STBCC__CLUSTER_Y_FOR_COORD_Y(y2); + + stbcc__clumpid c1 = g->clump_for_node[y1][x1]; + stbcc__clumpid c2 = g->clump_for_node[y2][x2]; + + stbcc__relative_clumpid rc; + + assert(cx1 != cx2 || cy1 != cy2); + assert(abs(cx1-cx2) + abs(cy1-cy2) == 1); + + // add connection to c2 in c1 + + rc.clump_index = c2; + rc.cluster_dx = x2-x1; + rc.cluster_dy = y2-y1; + + cluster = &g->cluster[cy1][cx1]; + clump = &cluster->clump[c1]; + adj = &cluster->adjacency_storage[clump->adjacent_clump_list_index]; + + for (i=0; i < clump->num_adjacent; ++i) + if (rc.clump_index == adj[i].clump_index && + rc.cluster_dx == adj[i].cluster_dx && + rc.cluster_dy == adj[i].cluster_dy) + break; + + if (i < clump->num_adjacent) + adj[i] = adj[--clump->num_adjacent]; + else + assert(0); +} + +static void stbcc__add_connections_to_adjacent_cluster(stbcc_grid *g, int cx, int cy, int dx, int dy) +{ + unsigned char connected[STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER][STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER/8] = { { 0 } }; + int x = cx * STBCC__CLUSTER_SIZE_X; + int y = cy * STBCC__CLUSTER_SIZE_Y; + int step_x, step_y=0, i, j, k, n; + + if (cx < 0 || cx >= g->cw || cy < 0 || cy >= g->ch) + return; + + if (cx+dx < 0 || cx+dx >= g->cw || cy+dy < 0 || cy+dy >= g->ch) + return; + + if (g->cluster[cy][cx].rebuild_adjacency) + return; + + assert(abs(dx) + abs(dy) == 1); + + if (dx == 1) { + i = STBCC__CLUSTER_SIZE_X-1; + j = 0; + step_x = 0; + step_y = 1; + n = STBCC__CLUSTER_SIZE_Y; + } else if (dx == -1) { + i = 0; + j = 0; + step_x = 0; + step_y = 1; + n = STBCC__CLUSTER_SIZE_Y; + } else if (dy == -1) { + i = 0; + j = 0; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + } else if (dy == 1) { + i = 0; + j = STBCC__CLUSTER_SIZE_Y-1; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + } else { + assert(0); + return; + } + + for (k=0; k < n; ++k) { + if (STBCC__MAP_OPEN(g, x+i, y+j) && STBCC__MAP_OPEN(g, x+i+dx, y+j+dy)) { + stbcc__clumpid src = g->clump_for_node[y+j][x+i]; + stbcc__clumpid dest = g->clump_for_node[y+j+dy][x+i+dx]; + if (0 == (connected[src][dest>>3] & (1 << (dest & 7)))) { + assert((dest>>3) < sizeof(connected)); + connected[src][dest>>3] |= 1 << (dest & 7); + stbcc__add_clump_connection(g, x+i, y+j, x+i+dx, y+j+dy); + if (g->cluster[cy][cx].rebuild_adjacency) + break; + } + } + i += step_x; + j += step_y; + } +} + +static void stbcc__remove_connections_to_adjacent_cluster(stbcc_grid *g, int cx, int cy, int dx, int dy) +{ + unsigned char disconnected[STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER][STBCC__MAX_EDGE_CLUMPS_PER_CLUSTER/8] = { { 0 } }; + int x = cx * STBCC__CLUSTER_SIZE_X; + int y = cy * STBCC__CLUSTER_SIZE_Y; + int step_x, step_y=0, i, j, k, n; + + if (cx < 0 || cx >= g->cw || cy < 0 || cy >= g->ch) + return; + + if (cx+dx < 0 || cx+dx >= g->cw || cy+dy < 0 || cy+dy >= g->ch) + return; + + assert(abs(dx) + abs(dy) == 1); + + if (dx == 1) { + i = STBCC__CLUSTER_SIZE_X-1; + j = 0; + step_x = 0; + step_y = 1; + n = STBCC__CLUSTER_SIZE_Y; + } else if (dx == -1) { + i = 0; + j = 0; + step_x = 0; + step_y = 1; + n = STBCC__CLUSTER_SIZE_Y; + } else if (dy == -1) { + i = 0; + j = 0; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + } else if (dy == 1) { + i = 0; + j = STBCC__CLUSTER_SIZE_Y-1; + step_x = 1; + step_y = 0; + n = STBCC__CLUSTER_SIZE_X; + } else { + assert(0); + return; + } + + for (k=0; k < n; ++k) { + if (STBCC__MAP_OPEN(g, x+i, y+j) && STBCC__MAP_OPEN(g, x+i+dx, y+j+dy)) { + stbcc__clumpid src = g->clump_for_node[y+j][x+i]; + stbcc__clumpid dest = g->clump_for_node[y+j+dy][x+i+dx]; + if (0 == (disconnected[src][dest>>3] & (1 << (dest & 7)))) { + disconnected[src][dest>>3] |= 1 << (dest & 7); + stbcc__remove_clump_connection(g, x+i, y+j, x+i+dx, y+j+dy); + } + } + i += step_x; + j += step_y; + } +} + +static stbcc__tinypoint stbcc__incluster_find(stbcc__cluster_build_info *cbi, int x, int y) +{ + stbcc__tinypoint p,q; + p = cbi->parent[y][x]; + if (p.x == x && p.y == y) + return p; + q = stbcc__incluster_find(cbi, p.x, p.y); + cbi->parent[y][x] = q; + return q; +} + +static void stbcc__incluster_union(stbcc__cluster_build_info *cbi, int x1, int y1, int x2, int y2) +{ + stbcc__tinypoint p = stbcc__incluster_find(cbi, x1,y1); + stbcc__tinypoint q = stbcc__incluster_find(cbi, x2,y2); + + if (p.x == q.x && p.y == q.y) + return; + + cbi->parent[p.y][p.x] = q; +} + +static void stbcc__switch_root(stbcc__cluster_build_info *cbi, int x, int y, stbcc__tinypoint p) +{ + cbi->parent[p.y][p.x].x = x; + cbi->parent[p.y][p.x].y = y; + cbi->parent[y][x].x = x; + cbi->parent[y][x].y = y; +} + +static void stbcc__build_clumps_for_cluster(stbcc_grid *g, int cx, int cy) +{ + stbcc__cluster *c; + stbcc__cluster_build_info cbi; + int label=0; + int i,j; + int x = cx * STBCC__CLUSTER_SIZE_X; + int y = cy * STBCC__CLUSTER_SIZE_Y; + + // set initial disjoint set forest state + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) { + for (i=0; i < STBCC__CLUSTER_SIZE_X; ++i) { + cbi.parent[j][i].x = i; + cbi.parent[j][i].y = j; + } + } + + // join all sets that are connected + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) { + // check down only if not on bottom row + if (j < STBCC__CLUSTER_SIZE_Y-1) + for (i=0; i < STBCC__CLUSTER_SIZE_X; ++i) + if (STBCC__MAP_OPEN(g,x+i,y+j) && STBCC__MAP_OPEN(g,x+i ,y+j+1)) + stbcc__incluster_union(&cbi, i,j, i,j+1); + // check right for everything but rightmost column + for (i=0; i < STBCC__CLUSTER_SIZE_X-1; ++i) + if (STBCC__MAP_OPEN(g,x+i,y+j) && STBCC__MAP_OPEN(g,x+i+1,y+j )) + stbcc__incluster_union(&cbi, i,j, i+1,j); + } + + // label all non-empty clumps along edges so that all edge clumps are first + // in list; this means in degenerate case we can skip traversing non-edge clumps. + // because in the first pass we only label leaders, we swap the leader to the + // edge first + + // first put solid labels on all the edges; these will get overwritten if they're open + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) + cbi.label[j][0] = cbi.label[j][STBCC__CLUSTER_SIZE_X-1] = STBCC__NULL_CLUMPID; + for (i=0; i < STBCC__CLUSTER_SIZE_X; ++i) + cbi.label[0][i] = cbi.label[STBCC__CLUSTER_SIZE_Y-1][i] = STBCC__NULL_CLUMPID; + + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) { + i = 0; + if (STBCC__MAP_OPEN(g, x+i, y+j)) { + stbcc__tinypoint p = stbcc__incluster_find(&cbi, i,j); + if (p.x == i && p.y == j) + // if this is the leader, give it a label + cbi.label[j][i] = label++; + else if (!(p.x == 0 || p.x == STBCC__CLUSTER_SIZE_X-1 || p.y == 0 || p.y == STBCC__CLUSTER_SIZE_Y-1)) { + // if leader is in interior, promote this edge node to leader and label + stbcc__switch_root(&cbi, i, j, p); + cbi.label[j][i] = label++; + } + // else if leader is on edge, do nothing (it'll get labelled when we reach it) + } + i = STBCC__CLUSTER_SIZE_X-1; + if (STBCC__MAP_OPEN(g, x+i, y+j)) { + stbcc__tinypoint p = stbcc__incluster_find(&cbi, i,j); + if (p.x == i && p.y == j) + cbi.label[j][i] = label++; + else if (!(p.x == 0 || p.x == STBCC__CLUSTER_SIZE_X-1 || p.y == 0 || p.y == STBCC__CLUSTER_SIZE_Y-1)) { + stbcc__switch_root(&cbi, i, j, p); + cbi.label[j][i] = label++; + } + } + } + + for (i=1; i < STBCC__CLUSTER_SIZE_Y-1; ++i) { + j = 0; + if (STBCC__MAP_OPEN(g, x+i, y+j)) { + stbcc__tinypoint p = stbcc__incluster_find(&cbi, i,j); + if (p.x == i && p.y == j) + cbi.label[j][i] = label++; + else if (!(p.x == 0 || p.x == STBCC__CLUSTER_SIZE_X-1 || p.y == 0 || p.y == STBCC__CLUSTER_SIZE_Y-1)) { + stbcc__switch_root(&cbi, i, j, p); + cbi.label[j][i] = label++; + } + } + j = STBCC__CLUSTER_SIZE_Y-1; + if (STBCC__MAP_OPEN(g, x+i, y+j)) { + stbcc__tinypoint p = stbcc__incluster_find(&cbi, i,j); + if (p.x == i && p.y == j) + cbi.label[j][i] = label++; + else if (!(p.x == 0 || p.x == STBCC__CLUSTER_SIZE_X-1 || p.y == 0 || p.y == STBCC__CLUSTER_SIZE_Y-1)) { + stbcc__switch_root(&cbi, i, j, p); + cbi.label[j][i] = label++; + } + } + } + + c = &g->cluster[cy][cx]; + c->num_edge_clumps = label; + + // label any internal clusters + for (j=1; j < STBCC__CLUSTER_SIZE_Y-1; ++j) { + for (i=1; i < STBCC__CLUSTER_SIZE_X-1; ++i) { + stbcc__tinypoint p = cbi.parent[j][i]; + if (p.x == i && p.y == j) { + if (STBCC__MAP_OPEN(g,x+i,y+j)) + cbi.label[j][i] = label++; + else + cbi.label[j][i] = STBCC__NULL_CLUMPID; + } + } + } + + // label all other nodes + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) { + for (i=0; i < STBCC__CLUSTER_SIZE_X; ++i) { + stbcc__tinypoint p = stbcc__incluster_find(&cbi, i,j); + if (p.x != i || p.y != j) { + if (STBCC__MAP_OPEN(g,x+i,y+j)) + cbi.label[j][i] = cbi.label[p.y][p.x]; + } + if (STBCC__MAP_OPEN(g,x+i,y+j)) + assert(cbi.label[j][i] != STBCC__NULL_CLUMPID); + } + } + + c->num_clumps = label; + + for (i=0; i < label; ++i) { + c->clump[i].num_adjacent = 0; + c->clump[i].max_adjacent = 0; + } + + for (j=0; j < STBCC__CLUSTER_SIZE_Y; ++j) + for (i=0; i < STBCC__CLUSTER_SIZE_X; ++i) { + g->clump_for_node[y+j][x+i] = cbi.label[j][i]; // @OPTIMIZE: remove cbi.label entirely + assert(g->clump_for_node[y+j][x+i] <= STBCC__NULL_CLUMPID); + } + + // set the global label for all interior clumps since they can't have connections, + // so we don't have to do this on the global pass (brings from O(N) to O(N^0.75)) + for (i=(int) c->num_edge_clumps; i < (int) c->num_clumps; ++i) { + stbcc__global_clumpid gc; + gc.f.cluster_x = cx; + gc.f.cluster_y = cy; + gc.f.clump_index = i; + c->clump[i].global_label = gc; + } + + c->rebuild_adjacency = 1; // flag that it has no valid adjacency data +} + +#endif // STB_CONNECTED_COMPONENTS_IMPLEMENTATION +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_divide.h b/include/stb_divide.h new file mode 100644 index 0000000..6a51e3f --- /dev/null +++ b/include/stb_divide.h @@ -0,0 +1,433 @@ +// stb_divide.h - v0.94 - public domain - Sean Barrett, Feb 2010 +// Three kinds of divide/modulus of signed integers. +// +// HISTORY +// +// v0.94 Fix integer overflow issues +// v0.93 2020-02-02 Write useful exit() value from main() +// v0.92 2019-02-25 Fix warning +// v0.91 2010-02-27 Fix euclidean division by INT_MIN for non-truncating C +// Check result with 64-bit math to catch such cases +// v0.90 2010-02-24 First public release +// +// USAGE +// +// In *ONE* source file, put: +// +// #define STB_DIVIDE_IMPLEMENTATION +// // #define C_INTEGER_DIVISION_TRUNCATES // see Note 1 +// // #define C_INTEGER_DIVISION_FLOORS // see Note 2 +// #include "stb_divide.h" +// +// Other source files should just include stb_divide.h +// +// Note 1: On platforms/compilers that you know signed C division +// truncates, you can #define C_INTEGER_DIVISION_TRUNCATES. +// +// Note 2: On platforms/compilers that you know signed C division +// floors (rounds to negative infinity), you can #define +// C_INTEGER_DIVISION_FLOORS. +// +// You can #define STB_DIVIDE_TEST in which case the implementation +// will generate a main() and compiling the result will create a +// program that tests the implementation. Run it with no arguments +// and any output indicates an error; run it with any argument and +// it will also print the test results. Define STB_DIVIDE_TEST_64 +// to a 64-bit integer type to avoid overflows in the result-checking +// which give false negatives. +// +// ABOUT +// +// This file provides three different consistent divide/mod pairs +// implemented on top of arbitrary C/C++ division, including correct +// handling of overflow of intermediate calculations: +// +// trunc: a/b truncates to 0, a%b has same sign as a +// floor: a/b truncates to -inf, a%b has same sign as b +// eucl: a/b truncates to sign(b)*inf, a%b is non-negative +// +// Not necessarily optimal; I tried to keep it generally efficient, +// but there may be better ways. +// +// Briefly, for those who are not familiar with the problem, we note +// the reason these divides exist and are interesting: +// +// 'trunc' is easy to implement in hardware (strip the signs, +// compute, reapply the signs), thus is commonly defined +// by many languages (including C99) +// +// 'floor' is simple to define and better behaved than trunc; +// for example it divides integers into fixed-size buckets +// without an extra-wide bucket at 0, and for a fixed +// divisor N there are only |N| possible moduli. +// +// 'eucl' guarantees fixed-sized buckets *and* a non-negative +// modulus and defines division to be whatever is needed +// to achieve that result. +// +// See "The Euclidean definition of the functions div and mod" +// by Raymond Boute (1992), or "Division and Modulus for Computer +// Scientists" by Daan Leijen (2001) +// +// We assume of the built-in C division: +// (a) modulus is the remainder for the corresponding division +// (b) a/b truncates if a and b are the same sign +// +// Property (a) requires (a/b)*b + (a%b)==a, and is required by C. +// Property (b) seems to be true of all hardware but is *not* satisfied +// by the euclidean division operator we define, so it's possibly not +// always true. If any such platform turns up, we can add more cases. +// (Possibly only stb_div_trunc currently relies on property (b).) +// +// LICENSE +// +// See end of file for license information. + + +#ifndef INCLUDE_STB_DIVIDE_H +#define INCLUDE_STB_DIVIDE_H + +#ifdef __cplusplus +extern "C" { +#endif + +extern int stb_div_trunc(int value_to_be_divided, int value_to_divide_by); +extern int stb_div_floor(int value_to_be_divided, int value_to_divide_by); +extern int stb_div_eucl (int value_to_be_divided, int value_to_divide_by); +extern int stb_mod_trunc(int value_to_be_divided, int value_to_divide_by); +extern int stb_mod_floor(int value_to_be_divided, int value_to_divide_by); +extern int stb_mod_eucl (int value_to_be_divided, int value_to_divide_by); + +#ifdef __cplusplus +} +#endif + +#ifdef STB_DIVIDE_IMPLEMENTATION + +#if defined(__STDC_VERSION) && __STDC_VERSION__ >= 19901 + #ifndef C_INTEGER_DIVISION_TRUNCATES + #define C_INTEGER_DIVISION_TRUNCATES + #endif +#endif + +#ifndef INT_MIN +#include // if you have no limits.h, #define INT_MIN yourself +#endif + +// the following macros are designed to allow testing +// other platforms by simulating them +#ifndef STB_DIVIDE_TEST_FLOOR + #define stb__div(a,b) ((a)/(b)) + #define stb__mod(a,b) ((a)%(b)) +#else + // implement floor-style divide on trunc platform + #ifndef C_INTEGER_DIVISION_TRUNCATES + #error "floor test requires truncating division" + #endif + #undef C_INTEGER_DIVISION_TRUNCATES + int stb__div(int v1, int v2) + { + int q = v1/v2, r = v1%v2; + if ((r > 0 && v2 < 0) || (r < 0 && v2 > 0)) + return q-1; + else + return q; + } + + int stb__mod(int v1, int v2) + { + int r = v1%v2; + if ((r > 0 && v2 < 0) || (r < 0 && v2 > 0)) + return r+v2; + else + return r; + } +#endif + +int stb_div_trunc(int v1, int v2) +{ + #ifdef C_INTEGER_DIVISION_TRUNCATES + return v1/v2; + #else + if (v1 >= 0 && v2 <= 0) + return -stb__div(-v1,v2); // both negative to avoid overflow + if (v1 <= 0 && v2 >= 0) + if (v1 != INT_MIN) + return -stb__div(v1,-v2); // both negative to avoid overflow + else + return -stb__div(v1+v2,-v2)-1; // push v1 away from wrap point + else + return v1/v2; // same sign, so expect truncation + #endif +} + +int stb_div_floor(int v1, int v2) +{ + #ifdef C_INTEGER_DIVISION_FLOORS + return v1/v2; + #else + if (v1 >= 0 && v2 < 0) { + if (v2 + 1 >= INT_MIN + v1) // check if increasing v1's magnitude overflows + return -stb__div((v2+1)-v1,v2); // nope, so just compute it + else + return -stb__div(-v1,v2) + ((-v1)%v2 ? -1 : 0); + } + if (v1 < 0 && v2 >= 0) { + if (v1 != INT_MIN) { + if (v1 + 1 >= INT_MIN + v2) // check if increasing v1's magnitude overflows + return -stb__div((v1+1)-v2,-v2); // nope, so just compute it + else + return -stb__div(-v1,v2) + (stb__mod(v1,-v2) ? -1 : 0); + } else // it must be possible to compute -(v1+v2) without overflowing + return -stb__div(-(v1+v2),v2) + (stb__mod(-(v1+v2),v2) ? -2 : -1); + } else + return v1/v2; // same sign, so expect truncation + #endif +} + +int stb_div_eucl(int v1, int v2) +{ + int q,r; + #ifdef C_INTEGER_DIVISION_TRUNCATES + q = v1/v2; + r = v1%v2; + #else + // handle every quadrant separately, since we can't rely on q and r flor + if (v1 >= 0) + if (v2 >= 0) + return stb__div(v1,v2); + else if (v2 != INT_MIN) + q = -stb__div(v1,-v2), r = stb__mod(v1,-v2); + else + q = 0, r = v1; + else if (v1 != INT_MIN) + if (v2 >= 0) + q = -stb__div(-v1,v2), r = -stb__mod(-v1,v2); + else if (v2 != INT_MIN) + q = stb__div(-v1,-v2), r = -stb__mod(-v1,-v2); + else // if v2 is INT_MIN, then we can't use -v2, but we can't divide by v2 + q = 1, r = v1-q*v2; + else // if v1 is INT_MIN, we have to move away from overflow place + if (v2 >= 0) + q = -stb__div(-(v1+v2),v2)-1, r = -stb__mod(-(v1+v2),v2); + else if (v2 != INT_MIN) + q = stb__div(-(v1-v2),-v2)+1, r = -stb__mod(-(v1-v2),-v2); + else // for INT_MIN / INT_MIN, we need to be extra-careful to avoid overflow + q = 1, r = 0; + #endif + if (r >= 0) + return q; + else + return q + (v2 > 0 ? -1 : 1); +} + +int stb_mod_trunc(int v1, int v2) +{ + #ifdef C_INTEGER_DIVISION_TRUNCATES + return v1%v2; + #else + if (v1 >= 0) { // modulus result should always be positive + int r = stb__mod(v1,v2); + if (r >= 0) + return r; + else + return r - (v2 < 0 ? v2 : -v2); + } else { // modulus result should always be negative + int r = stb__mod(v1,v2); + if (r <= 0) + return r; + else + return r + (v2 < 0 ? v2 : -v2); + } + #endif +} + +int stb_mod_floor(int v1, int v2) +{ + #ifdef C_INTEGER_DIVISION_FLOORS + return v1%v2; + #else + if (v2 >= 0) { // result should always be positive + int r = stb__mod(v1,v2); + if (r >= 0) + return r; + else + return r + v2; + } else { // result should always be negative + int r = stb__mod(v1,v2); + if (r <= 0) + return r; + else + return r + v2; + } + #endif +} + +int stb_mod_eucl(int v1, int v2) +{ + int r = stb__mod(v1,v2); + + if (r >= 0) + return r; + else + return r - (v2 < 0 ? v2 : -v2); // negative abs() [to avoid overflow] +} + +#ifdef STB_DIVIDE_TEST +#include +#include +#include + +int show=0; +int err=0; + +void stbdiv_check(int q, int r, int a, int b, char *type, int dir) +{ + if ((dir > 0 && r < 0) || (dir < 0 && r > 0)) { + fprintf(stderr, "FAILED: %s(%d,%d) remainder %d in wrong direction\n", type,a,b,r); + err++; + } else + if (b != INT_MIN) // can't compute abs(), but if b==INT_MIN all remainders are valid + if (r <= -abs(b) || r >= abs(b)) { + fprintf(stderr, "FAILED: %s(%d,%d) remainder %d out of range\n", type,a,b,r); + err++; + } + #ifdef STB_DIVIDE_TEST_64 + { + STB_DIVIDE_TEST_64 q64 = q, r64=r, a64=a, b64=b; + if (q64*b64+r64 != a64) { + fprintf(stderr, "FAILED: %s(%d,%d) remainder %d doesn't match quotient %d\n", type,a,b,r,q); + err++; + } + } + #else + if (q*b+r != a) { + fprintf(stderr, "FAILED: %s(%d,%d) remainder %d doesn't match quotient %d\n", type,a,b,r,q); + err++; + } + #endif +} + +void test(int a, int b) +{ + int q,r; + if (show) printf("(%+11d,%+d) | ", a,b); + q = stb_div_trunc(a,b), r = stb_mod_trunc(a,b); + if (show) printf("(%+11d,%+2d) ", q,r); stbdiv_check(q,r,a,b, "trunc",a); + q = stb_div_floor(a,b), r = stb_mod_floor(a,b); + if (show) printf("(%+11d,%+2d) ", q,r); stbdiv_check(q,r,a,b, "floor",b); + q = stb_div_eucl (a,b), r = stb_mod_eucl (a,b); + if (show) printf("(%+11d,%+2d)\n", q,r); stbdiv_check(q,r,a,b, "euclidean",1); +} + +void testh(int a, int b) +{ + int q,r; + if (show) printf("(%08x,%08x) |\n", a,b); + q = stb_div_trunc(a,b), r = stb_mod_trunc(a,b); stbdiv_check(q,r,a,b, "trunc",a); + if (show) printf(" (%08x,%08x)", q,r); + q = stb_div_floor(a,b), r = stb_mod_floor(a,b); stbdiv_check(q,r,a,b, "floor",b); + if (show) printf(" (%08x,%08x)", q,r); + q = stb_div_eucl (a,b), r = stb_mod_eucl (a,b); stbdiv_check(q,r,a,b, "euclidean",1); + if (show) printf(" (%08x,%08x)\n ", q,r); +} + +int main(int argc, char **argv) +{ + if (argc > 1) show=1; + + test(8,3); + test(8,-3); + test(-8,3); + test(-8,-3); + test(1,2); + test(1,-2); + test(-1,2); + test(-1,-2); + test(8,4); + test(8,-4); + test(-8,4); + test(-8,-4); + + test(INT_MAX,1); + test(INT_MIN,1); + test(INT_MIN+1,1); + test(INT_MAX,-1); + //test(INT_MIN,-1); // this traps in MSVC, so we leave it untested + test(INT_MIN+1,-1); + test(INT_MIN,-2); + test(INT_MIN+1,2); + test(INT_MIN+1,-2); + test(INT_MAX,2); + test(INT_MAX,-2); + test(INT_MIN+1,2); + test(INT_MIN+1,-2); + test(INT_MIN,2); + test(INT_MIN,-2); + test(INT_MIN,7); + test(INT_MIN,-7); + test(INT_MIN+1,4); + test(INT_MIN+1,-4); + + testh(-7, INT_MIN); + testh(-1, INT_MIN); + testh(1, INT_MIN); + testh(7, INT_MIN); + + testh(INT_MAX-1, INT_MIN); + testh(INT_MAX, INT_MIN); + testh(INT_MIN, INT_MIN); + testh(INT_MIN+1, INT_MIN); + + testh(INT_MAX-1, INT_MAX); + testh(INT_MAX , INT_MAX); + testh(INT_MIN , INT_MAX); + testh(INT_MIN+1, INT_MAX); + + return err > 0 ? 1 : 0; +} +#endif // STB_DIVIDE_TEST +#endif // STB_DIVIDE_IMPLEMENTATION +#endif // INCLUDE_STB_DIVIDE_H + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_ds.h b/include/stb_ds.h new file mode 100644 index 0000000..e84c82d --- /dev/null +++ b/include/stb_ds.h @@ -0,0 +1,1895 @@ +/* stb_ds.h - v0.67 - public domain data structures - Sean Barrett 2019 + + This is a single-header-file library that provides easy-to-use + dynamic arrays and hash tables for C (also works in C++). + + For a gentle introduction: + http://nothings.org/stb_ds + + To use this library, do this in *one* C or C++ file: + #define STB_DS_IMPLEMENTATION + #include "stb_ds.h" + +TABLE OF CONTENTS + + Table of Contents + Compile-time options + License + Documentation + Notes + Notes - Dynamic arrays + Notes - Hash maps + Credits + +COMPILE-TIME OPTIONS + + #define STBDS_NO_SHORT_NAMES + + This flag needs to be set globally. + + By default stb_ds exposes shorter function names that are not qualified + with the "stbds_" prefix. If these names conflict with the names in your + code, define this flag. + + #define STBDS_SIPHASH_2_4 + + This flag only needs to be set in the file containing #define STB_DS_IMPLEMENTATION. + + By default stb_ds.h hashes using a weaker variant of SipHash and a custom hash for + 4- and 8-byte keys. On 64-bit platforms, you can define the above flag to force + stb_ds.h to use specification-compliant SipHash-2-4 for all keys. Doing so makes + hash table insertion about 20% slower on 4- and 8-byte keys, 5% slower on + 64-byte keys, and 10% slower on 256-byte keys on my test computer. + + #define STBDS_REALLOC(context,ptr,size) better_realloc + #define STBDS_FREE(context,ptr) better_free + + These defines only need to be set in the file containing #define STB_DS_IMPLEMENTATION. + + By default stb_ds uses stdlib realloc() and free() for memory management. You can + substitute your own functions instead by defining these symbols. You must either + define both, or neither. Note that at the moment, 'context' will always be NULL. + @TODO add an array/hash initialization function that takes a memory context pointer. + + #define STBDS_UNIT_TESTS + + Defines a function stbds_unit_tests() that checks the functioning of the data structures. + + Note that on older versions of gcc (e.g. 5.x.x) you may need to build with '-std=c++0x' + (or equivalentally '-std=c++11') when using anonymous structures as seen on the web + page or in STBDS_UNIT_TESTS. + +LICENSE + + Placed in the public domain and also MIT licensed. + See end of file for detailed license information. + +DOCUMENTATION + + Dynamic Arrays + + Non-function interface: + + Declare an empty dynamic array of type T + T* foo = NULL; + + Access the i'th item of a dynamic array 'foo' of type T, T* foo: + foo[i] + + Functions (actually macros) + + arrfree: + void arrfree(T*); + Frees the array. + + arrlen: + ptrdiff_t arrlen(T*); + Returns the number of elements in the array. + + arrlenu: + size_t arrlenu(T*); + Returns the number of elements in the array as an unsigned type. + + arrpop: + T arrpop(T* a) + Removes the final element of the array and returns it. + + arrput: + T arrput(T* a, T b); + Appends the item b to the end of array a. Returns b. + + arrins: + T arrins(T* a, int p, T b); + Inserts the item b into the middle of array a, into a[p], + moving the rest of the array over. Returns b. + + arrinsn: + void arrinsn(T* a, int p, int n); + Inserts n uninitialized items into array a starting at a[p], + moving the rest of the array over. + + arraddnptr: + T* arraddnptr(T* a, int n) + Appends n uninitialized items onto array at the end. + Returns a pointer to the first uninitialized item added. + + arraddnindex: + size_t arraddnindex(T* a, int n) + Appends n uninitialized items onto array at the end. + Returns the index of the first uninitialized item added. + + arrdel: + void arrdel(T* a, int p); + Deletes the element at a[p], moving the rest of the array over. + + arrdeln: + void arrdeln(T* a, int p, int n); + Deletes n elements starting at a[p], moving the rest of the array over. + + arrdelswap: + void arrdelswap(T* a, int p); + Deletes the element at a[p], replacing it with the element from + the end of the array. O(1) performance. + + arrsetlen: + void arrsetlen(T* a, int n); + Changes the length of the array to n. Allocates uninitialized + slots at the end if necessary. + + arrsetcap: + size_t arrsetcap(T* a, int n); + Sets the length of allocated storage to at least n. It will not + change the length of the array. + + arrcap: + size_t arrcap(T* a); + Returns the number of total elements the array can contain without + needing to be reallocated. + + Hash maps & String hash maps + + Given T is a structure type: struct { TK key; TV value; }. Note that some + functions do not require TV value and can have other fields. For string + hash maps, TK must be 'char *'. + + Special interface: + + stbds_rand_seed: + void stbds_rand_seed(size_t seed); + For security against adversarially chosen data, you should seed the + library with a strong random number. Or at least seed it with time(). + + stbds_hash_string: + size_t stbds_hash_string(char *str, size_t seed); + Returns a hash value for a string. + + stbds_hash_bytes: + size_t stbds_hash_bytes(void *p, size_t len, size_t seed); + These functions hash an arbitrary number of bytes. The function + uses a custom hash for 4- and 8-byte data, and a weakened version + of SipHash for everything else. On 64-bit platforms you can get + specification-compliant SipHash-2-4 on all data by defining + STBDS_SIPHASH_2_4, at a significant cost in speed. + + Non-function interface: + + Declare an empty hash map of type T + T* foo = NULL; + + Access the i'th entry in a hash table T* foo: + foo[i] + + Function interface (actually macros): + + hmfree + shfree + void hmfree(T*); + void shfree(T*); + Frees the hashmap and sets the pointer to NULL. + + hmlen + shlen + ptrdiff_t hmlen(T*) + ptrdiff_t shlen(T*) + Returns the number of elements in the hashmap. + + hmlenu + shlenu + size_t hmlenu(T*) + size_t shlenu(T*) + Returns the number of elements in the hashmap. + + hmgeti + shgeti + hmgeti_ts + ptrdiff_t hmgeti(T*, TK key) + ptrdiff_t shgeti(T*, char* key) + ptrdiff_t hmgeti_ts(T*, TK key, ptrdiff_t tempvar) + Returns the index in the hashmap which has the key 'key', or -1 + if the key is not present. + + hmget + hmget_ts + shget + TV hmget(T*, TK key) + TV shget(T*, char* key) + TV hmget_ts(T*, TK key, ptrdiff_t tempvar) + Returns the value corresponding to 'key' in the hashmap. + The structure must have a 'value' field + + hmgets + shgets + T hmgets(T*, TK key) + T shgets(T*, char* key) + Returns the structure corresponding to 'key' in the hashmap. + + hmgetp + shgetp + hmgetp_ts + hmgetp_null + shgetp_null + T* hmgetp(T*, TK key) + T* shgetp(T*, char* key) + T* hmgetp_ts(T*, TK key, ptrdiff_t tempvar) + T* hmgetp_null(T*, TK key) + T* shgetp_null(T*, char *key) + Returns a pointer to the structure corresponding to 'key' in + the hashmap. Functions ending in "_null" return NULL if the key + is not present in the hashmap; the others return a pointer to a + structure holding the default value (but not the searched-for key). + + hmdefault + shdefault + TV hmdefault(T*, TV value) + TV shdefault(T*, TV value) + Sets the default value for the hashmap, the value which will be + returned by hmget/shget if the key is not present. + + hmdefaults + shdefaults + TV hmdefaults(T*, T item) + TV shdefaults(T*, T item) + Sets the default struct for the hashmap, the contents which will be + returned by hmgets/shgets if the key is not present. + + hmput + shput + TV hmput(T*, TK key, TV value) + TV shput(T*, char* key, TV value) + Inserts a pair into the hashmap. If the key is already + present in the hashmap, updates its value. + + hmputs + shputs + T hmputs(T*, T item) + T shputs(T*, T item) + Inserts a struct with T.key into the hashmap. If the struct is already + present in the hashmap, updates it. + + hmdel + shdel + int hmdel(T*, TK key) + int shdel(T*, char* key) + If 'key' is in the hashmap, deletes its entry and returns 1. + Otherwise returns 0. + + Function interface (actually macros) for strings only: + + sh_new_strdup + void sh_new_strdup(T*); + Overwrites the existing pointer with a newly allocated + string hashmap which will automatically allocate and free + each string key using realloc/free + + sh_new_arena + void sh_new_arena(T*); + Overwrites the existing pointer with a newly allocated + string hashmap which will automatically allocate each string + key to a string arena. Every string key ever used by this + hash table remains in the arena until the arena is freed. + Additionally, any key which is deleted and reinserted will + be allocated multiple times in the string arena. + +NOTES + + * These data structures are realloc'd when they grow, and the macro + "functions" write to the provided pointer. This means: (a) the pointer + must be an lvalue, and (b) the pointer to the data structure is not + stable, and you must maintain it the same as you would a realloc'd + pointer. For example, if you pass a pointer to a dynamic array to a + function which updates it, the function must return back the new + pointer to the caller. This is the price of trying to do this in C. + + * The following are the only functions that are thread-safe on a single data + structure, i.e. can be run in multiple threads simultaneously on the same + data structure + hmlen shlen + hmlenu shlenu + hmget_ts shget_ts + hmgeti_ts shgeti_ts + hmgets_ts shgets_ts + + * You iterate over the contents of a dynamic array and a hashmap in exactly + the same way, using arrlen/hmlen/shlen: + + for (i=0; i < arrlen(foo); ++i) + ... foo[i] ... + + * All operations except arrins/arrdel are O(1) amortized, but individual + operations can be slow, so these data structures may not be suitable + for real time use. Dynamic arrays double in capacity as needed, so + elements are copied an average of once. Hash tables double/halve + their size as needed, with appropriate hysteresis to maintain O(1) + performance. + +NOTES - DYNAMIC ARRAY + + * If you know how long a dynamic array is going to be in advance, you can avoid + extra memory allocations by using arrsetlen to allocate it to that length in + advance and use foo[n] while filling it out, or arrsetcap to allocate the memory + for that length and use arrput/arrpush as normal. + + * Unlike some other versions of the dynamic array, this version should + be safe to use with strict-aliasing optimizations. + +NOTES - HASH MAP + + * For compilers other than GCC and clang (e.g. Visual Studio), for hmput/hmget/hmdel + and variants, the key must be an lvalue (so the macro can take the address of it). + Extensions are used that eliminate this requirement if you're using C99 and later + in GCC or clang, or if you're using C++ in GCC. But note that this can make your + code less portable. + + * To test for presence of a key in a hashmap, just do 'hmgeti(foo,key) >= 0'. + + * The iteration order of your data in the hashmap is determined solely by the + order of insertions and deletions. In particular, if you never delete, new + keys are always added at the end of the array. This will be consistent + across all platforms and versions of the library. However, you should not + attempt to serialize the internal hash table, as the hash is not consistent + between different platforms, and may change with future versions of the library. + + * Use sh_new_arena() for string hashmaps that you never delete from. Initialize + with NULL if you're managing the memory for your strings, or your strings are + never freed (at least until the hashmap is freed). Otherwise, use sh_new_strdup(). + @TODO: make an arena variant that garbage collects the strings with a trivial + copy collector into a new arena whenever the table shrinks / rebuilds. Since + current arena recommendation is to only use arena if it never deletes, then + this can just replace current arena implementation. + + * If adversarial input is a serious concern and you're on a 64-bit platform, + enable STBDS_SIPHASH_2_4 (see the 'Compile-time options' section), and pass + a strong random number to stbds_rand_seed. + + * The default value for the hash table is stored in foo[-1], so if you + use code like 'hmget(T,k)->value = 5' you can accidentally overwrite + the value stored by hmdefault if 'k' is not present. + +CREDITS + + Sean Barrett -- library, idea for dynamic array API/implementation + Per Vognsen -- idea for hash table API/implementation + Rafael Sachetto -- arrpop() + github:HeroicKatora -- arraddn() reworking + + Bugfixes: + Andy Durdin + Shane Liesegang + Vinh Truong + Andreas Molzer + github:hashitaku + github:srdjanstipic + Macoy Madson + Andreas Vennstrom + Tobias Mansfield-Williams +*/ + +#ifdef STBDS_UNIT_TESTS +#define _CRT_SECURE_NO_WARNINGS +#endif + +#ifndef INCLUDE_STB_DS_H +#define INCLUDE_STB_DS_H + +#include +#include + +#ifndef STBDS_NO_SHORT_NAMES +#define arrlen stbds_arrlen +#define arrlenu stbds_arrlenu +#define arrput stbds_arrput +#define arrpush stbds_arrput +#define arrpop stbds_arrpop +#define arrfree stbds_arrfree +#define arraddn stbds_arraddn // deprecated, use one of the following instead: +#define arraddnptr stbds_arraddnptr +#define arraddnindex stbds_arraddnindex +#define arrsetlen stbds_arrsetlen +#define arrlast stbds_arrlast +#define arrins stbds_arrins +#define arrinsn stbds_arrinsn +#define arrdel stbds_arrdel +#define arrdeln stbds_arrdeln +#define arrdelswap stbds_arrdelswap +#define arrcap stbds_arrcap +#define arrsetcap stbds_arrsetcap + +#define hmput stbds_hmput +#define hmputs stbds_hmputs +#define hmget stbds_hmget +#define hmget_ts stbds_hmget_ts +#define hmgets stbds_hmgets +#define hmgetp stbds_hmgetp +#define hmgetp_ts stbds_hmgetp_ts +#define hmgetp_null stbds_hmgetp_null +#define hmgeti stbds_hmgeti +#define hmgeti_ts stbds_hmgeti_ts +#define hmdel stbds_hmdel +#define hmlen stbds_hmlen +#define hmlenu stbds_hmlenu +#define hmfree stbds_hmfree +#define hmdefault stbds_hmdefault +#define hmdefaults stbds_hmdefaults + +#define shput stbds_shput +#define shputi stbds_shputi +#define shputs stbds_shputs +#define shget stbds_shget +#define shgeti stbds_shgeti +#define shgets stbds_shgets +#define shgetp stbds_shgetp +#define shgetp_null stbds_shgetp_null +#define shdel stbds_shdel +#define shlen stbds_shlen +#define shlenu stbds_shlenu +#define shfree stbds_shfree +#define shdefault stbds_shdefault +#define shdefaults stbds_shdefaults +#define sh_new_arena stbds_sh_new_arena +#define sh_new_strdup stbds_sh_new_strdup + +#define stralloc stbds_stralloc +#define strreset stbds_strreset +#endif + +#if defined(STBDS_REALLOC) && !defined(STBDS_FREE) || !defined(STBDS_REALLOC) && defined(STBDS_FREE) +#error "You must define both STBDS_REALLOC and STBDS_FREE, or neither." +#endif +#if !defined(STBDS_REALLOC) && !defined(STBDS_FREE) +#include +#define STBDS_REALLOC(c,p,s) realloc(p,s) +#define STBDS_FREE(c,p) free(p) +#endif + +#ifdef _MSC_VER +#define STBDS_NOTUSED(v) (void)(v) +#else +#define STBDS_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +// for security against attackers, seed the library with a random number, at least time() but stronger is better +extern void stbds_rand_seed(size_t seed); + +// these are the hash functions used internally if you want to test them or use them for other purposes +extern size_t stbds_hash_bytes(void *p, size_t len, size_t seed); +extern size_t stbds_hash_string(char *str, size_t seed); + +// this is a simple string arena allocator, initialize with e.g. 'stbds_string_arena my_arena={0}'. +typedef struct stbds_string_arena stbds_string_arena; +extern char * stbds_stralloc(stbds_string_arena *a, char *str); +extern void stbds_strreset(stbds_string_arena *a); + +// have to #define STBDS_UNIT_TESTS to call this +extern void stbds_unit_tests(void); + +/////////////// +// +// Everything below here is implementation details +// + +extern void * stbds_arrgrowf(void *a, size_t elemsize, size_t addlen, size_t min_cap); +extern void stbds_arrfreef(void *a); +extern void stbds_hmfree_func(void *p, size_t elemsize); +extern void * stbds_hmget_key(void *a, size_t elemsize, void *key, size_t keysize, int mode); +extern void * stbds_hmget_key_ts(void *a, size_t elemsize, void *key, size_t keysize, ptrdiff_t *temp, int mode); +extern void * stbds_hmput_default(void *a, size_t elemsize); +extern void * stbds_hmput_key(void *a, size_t elemsize, void *key, size_t keysize, int mode); +extern void * stbds_hmdel_key(void *a, size_t elemsize, void *key, size_t keysize, size_t keyoffset, int mode); +extern void * stbds_shmode_func(size_t elemsize, int mode); + +#ifdef __cplusplus +} +#endif + +#if defined(__GNUC__) || defined(__clang__) +#define STBDS_HAS_TYPEOF +#ifdef __cplusplus +//#define STBDS_HAS_LITERAL_ARRAY // this is currently broken for clang +#endif +#endif + +#if !defined(__cplusplus) +#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L +#define STBDS_HAS_LITERAL_ARRAY +#endif +#endif + +// this macro takes the address of the argument, but on gcc/clang can accept rvalues +#if defined(STBDS_HAS_LITERAL_ARRAY) && defined(STBDS_HAS_TYPEOF) + #if __clang__ + #define STBDS_ADDRESSOF(typevar, value) ((__typeof__(typevar)[1]){value}) // literal array decays to pointer to value + #else + #define STBDS_ADDRESSOF(typevar, value) ((typeof(typevar)[1]){value}) // literal array decays to pointer to value + #endif +#else +#define STBDS_ADDRESSOF(typevar, value) &(value) +#endif + +#define STBDS_OFFSETOF(var,field) ((char *) &(var)->field - (char *) (var)) + +#define stbds_header(t) ((stbds_array_header *) (t) - 1) +#define stbds_temp(t) stbds_header(t)->temp +#define stbds_temp_key(t) (*(char **) stbds_header(t)->hash_table) + +#define stbds_arrsetcap(a,n) (stbds_arrgrow(a,0,n)) +#define stbds_arrsetlen(a,n) ((stbds_arrcap(a) < (size_t) (n) ? stbds_arrsetcap((a),(size_t)(n)),0 : 0), (a) ? stbds_header(a)->length = (size_t) (n) : 0) +#define stbds_arrcap(a) ((a) ? stbds_header(a)->capacity : 0) +#define stbds_arrlen(a) ((a) ? (ptrdiff_t) stbds_header(a)->length : 0) +#define stbds_arrlenu(a) ((a) ? stbds_header(a)->length : 0) +#define stbds_arrput(a,v) (stbds_arrmaybegrow(a,1), (a)[stbds_header(a)->length++] = (v)) +#define stbds_arrpush stbds_arrput // synonym +#define stbds_arrpop(a) (stbds_header(a)->length--, (a)[stbds_header(a)->length]) +#define stbds_arraddn(a,n) ((void)(stbds_arraddnindex(a, n))) // deprecated, use one of the following instead: +#define stbds_arraddnptr(a,n) (stbds_arrmaybegrow(a,n), (n) ? (stbds_header(a)->length += (n), &(a)[stbds_header(a)->length-(n)]) : (a)) +#define stbds_arraddnindex(a,n)(stbds_arrmaybegrow(a,n), (n) ? (stbds_header(a)->length += (n), stbds_header(a)->length-(n)) : stbds_arrlen(a)) +#define stbds_arraddnoff stbds_arraddnindex +#define stbds_arrlast(a) ((a)[stbds_header(a)->length-1]) +#define stbds_arrfree(a) ((void) ((a) ? STBDS_FREE(NULL,stbds_header(a)) : (void)0), (a)=NULL) +#define stbds_arrdel(a,i) stbds_arrdeln(a,i,1) +#define stbds_arrdeln(a,i,n) (memmove(&(a)[i], &(a)[(i)+(n)], sizeof *(a) * (stbds_header(a)->length-(n)-(i))), stbds_header(a)->length -= (n)) +#define stbds_arrdelswap(a,i) ((a)[i] = stbds_arrlast(a), stbds_header(a)->length -= 1) +#define stbds_arrinsn(a,i,n) (stbds_arraddn((a),(n)), memmove(&(a)[(i)+(n)], &(a)[i], sizeof *(a) * (stbds_header(a)->length-(n)-(i)))) +#define stbds_arrins(a,i,v) (stbds_arrinsn((a),(i),1), (a)[i]=(v)) + +#define stbds_arrmaybegrow(a,n) ((!(a) || stbds_header(a)->length + (n) > stbds_header(a)->capacity) \ + ? (stbds_arrgrow(a,n,0),0) : 0) + +#define stbds_arrgrow(a,b,c) ((a) = stbds_arrgrowf_wrapper((a), sizeof *(a), (b), (c))) + +#define stbds_hmput(t, k, v) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), (void*) STBDS_ADDRESSOF((t)->key, (k)), sizeof (t)->key, 0), \ + (t)[stbds_temp((t)-1)].key = (k), \ + (t)[stbds_temp((t)-1)].value = (v)) + +#define stbds_hmputs(t, s) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), &(s).key, sizeof (s).key, STBDS_HM_BINARY), \ + (t)[stbds_temp((t)-1)] = (s)) + +#define stbds_hmgeti(t,k) \ + ((t) = stbds_hmget_key_wrapper((t), sizeof *(t), (void*) STBDS_ADDRESSOF((t)->key, (k)), sizeof (t)->key, STBDS_HM_BINARY), \ + stbds_temp((t)-1)) + +#define stbds_hmgeti_ts(t,k,temp) \ + ((t) = stbds_hmget_key_ts_wrapper((t), sizeof *(t), (void*) STBDS_ADDRESSOF((t)->key, (k)), sizeof (t)->key, &(temp), STBDS_HM_BINARY), \ + (temp)) + +#define stbds_hmgetp(t, k) \ + ((void) stbds_hmgeti(t,k), &(t)[stbds_temp((t)-1)]) + +#define stbds_hmgetp_ts(t, k, temp) \ + ((void) stbds_hmgeti_ts(t,k,temp), &(t)[temp]) + +#define stbds_hmdel(t,k) \ + (((t) = stbds_hmdel_key_wrapper((t),sizeof *(t), (void*) STBDS_ADDRESSOF((t)->key, (k)), sizeof (t)->key, STBDS_OFFSETOF((t),key), STBDS_HM_BINARY)),(t)?stbds_temp((t)-1):0) + +#define stbds_hmdefault(t, v) \ + ((t) = stbds_hmput_default_wrapper((t), sizeof *(t)), (t)[-1].value = (v)) + +#define stbds_hmdefaults(t, s) \ + ((t) = stbds_hmput_default_wrapper((t), sizeof *(t)), (t)[-1] = (s)) + +#define stbds_hmfree(p) \ + ((void) ((p) != NULL ? stbds_hmfree_func((p)-1,sizeof*(p)),0 : 0),(p)=NULL) + +#define stbds_hmgets(t, k) (*stbds_hmgetp(t,k)) +#define stbds_hmget(t, k) (stbds_hmgetp(t,k)->value) +#define stbds_hmget_ts(t, k, temp) (stbds_hmgetp_ts(t,k,temp)->value) +#define stbds_hmlen(t) ((t) ? (ptrdiff_t) stbds_header((t)-1)->length-1 : 0) +#define stbds_hmlenu(t) ((t) ? stbds_header((t)-1)->length-1 : 0) +#define stbds_hmgetp_null(t,k) (stbds_hmgeti(t,k) == -1 ? NULL : &(t)[stbds_temp((t)-1)]) + +#define stbds_shput(t, k, v) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), (void*) (k), sizeof (t)->key, STBDS_HM_STRING), \ + (t)[stbds_temp((t)-1)].value = (v)) + +#define stbds_shputi(t, k, v) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), (void*) (k), sizeof (t)->key, STBDS_HM_STRING), \ + (t)[stbds_temp((t)-1)].value = (v), stbds_temp((t)-1)) + +#define stbds_shputs(t, s) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), (void*) (s).key, sizeof (s).key, STBDS_HM_STRING), \ + (t)[stbds_temp((t)-1)] = (s), \ + (t)[stbds_temp((t)-1)].key = stbds_temp_key((t)-1)) // above line overwrites whole structure, so must rewrite key here if it was allocated internally + +#define stbds_pshput(t, p) \ + ((t) = stbds_hmput_key_wrapper((t), sizeof *(t), (void*) (p)->key, sizeof (p)->key, STBDS_HM_PTR_TO_STRING), \ + (t)[stbds_temp((t)-1)] = (p)) + +#define stbds_shgeti(t,k) \ + ((t) = stbds_hmget_key_wrapper((t), sizeof *(t), (void*) (k), sizeof (t)->key, STBDS_HM_STRING), \ + stbds_temp((t)-1)) + +#define stbds_pshgeti(t,k) \ + ((t) = stbds_hmget_key_wrapper((t), sizeof *(t), (void*) (k), sizeof (*(t))->key, STBDS_HM_PTR_TO_STRING), \ + stbds_temp((t)-1)) + +#define stbds_shgetp(t, k) \ + ((void) stbds_shgeti(t,k), &(t)[stbds_temp((t)-1)]) + +#define stbds_pshget(t, k) \ + ((void) stbds_pshgeti(t,k), (t)[stbds_temp((t)-1)]) + +#define stbds_shdel(t,k) \ + (((t) = stbds_hmdel_key_wrapper((t),sizeof *(t), (void*) (k), sizeof (t)->key, STBDS_OFFSETOF((t),key), STBDS_HM_STRING)),(t)?stbds_temp((t)-1):0) +#define stbds_pshdel(t,k) \ + (((t) = stbds_hmdel_key_wrapper((t),sizeof *(t), (void*) (k), sizeof (*(t))->key, STBDS_OFFSETOF(*(t),key), STBDS_HM_PTR_TO_STRING)),(t)?stbds_temp((t)-1):0) + +#define stbds_sh_new_arena(t) \ + ((t) = stbds_shmode_func_wrapper(t, sizeof *(t), STBDS_SH_ARENA)) +#define stbds_sh_new_strdup(t) \ + ((t) = stbds_shmode_func_wrapper(t, sizeof *(t), STBDS_SH_STRDUP)) + +#define stbds_shdefault(t, v) stbds_hmdefault(t,v) +#define stbds_shdefaults(t, s) stbds_hmdefaults(t,s) + +#define stbds_shfree stbds_hmfree +#define stbds_shlenu stbds_hmlenu + +#define stbds_shgets(t, k) (*stbds_shgetp(t,k)) +#define stbds_shget(t, k) (stbds_shgetp(t,k)->value) +#define stbds_shgetp_null(t,k) (stbds_shgeti(t,k) == -1 ? NULL : &(t)[stbds_temp((t)-1)]) +#define stbds_shlen stbds_hmlen + +typedef struct +{ + size_t length; + size_t capacity; + void * hash_table; + ptrdiff_t temp; +} stbds_array_header; + +typedef struct stbds_string_block +{ + struct stbds_string_block *next; + char storage[8]; +} stbds_string_block; + +struct stbds_string_arena +{ + stbds_string_block *storage; + size_t remaining; + unsigned char block; + unsigned char mode; // this isn't used by the string arena itself +}; + +#define STBDS_HM_BINARY 0 +#define STBDS_HM_STRING 1 + +enum +{ + STBDS_SH_NONE, + STBDS_SH_DEFAULT, + STBDS_SH_STRDUP, + STBDS_SH_ARENA +}; + +#ifdef __cplusplus +// in C we use implicit assignment from these void*-returning functions to T*. +// in C++ these templates make the same code work +template static T * stbds_arrgrowf_wrapper(T *a, size_t elemsize, size_t addlen, size_t min_cap) { + return (T*)stbds_arrgrowf((void *)a, elemsize, addlen, min_cap); +} +template static T * stbds_hmget_key_wrapper(T *a, size_t elemsize, void *key, size_t keysize, int mode) { + return (T*)stbds_hmget_key((void*)a, elemsize, key, keysize, mode); +} +template static T * stbds_hmget_key_ts_wrapper(T *a, size_t elemsize, void *key, size_t keysize, ptrdiff_t *temp, int mode) { + return (T*)stbds_hmget_key_ts((void*)a, elemsize, key, keysize, temp, mode); +} +template static T * stbds_hmput_default_wrapper(T *a, size_t elemsize) { + return (T*)stbds_hmput_default((void *)a, elemsize); +} +template static T * stbds_hmput_key_wrapper(T *a, size_t elemsize, void *key, size_t keysize, int mode) { + return (T*)stbds_hmput_key((void*)a, elemsize, key, keysize, mode); +} +template static T * stbds_hmdel_key_wrapper(T *a, size_t elemsize, void *key, size_t keysize, size_t keyoffset, int mode){ + return (T*)stbds_hmdel_key((void*)a, elemsize, key, keysize, keyoffset, mode); +} +template static T * stbds_shmode_func_wrapper(T *, size_t elemsize, int mode) { + return (T*)stbds_shmode_func(elemsize, mode); +} +#else +#define stbds_arrgrowf_wrapper stbds_arrgrowf +#define stbds_hmget_key_wrapper stbds_hmget_key +#define stbds_hmget_key_ts_wrapper stbds_hmget_key_ts +#define stbds_hmput_default_wrapper stbds_hmput_default +#define stbds_hmput_key_wrapper stbds_hmput_key +#define stbds_hmdel_key_wrapper stbds_hmdel_key +#define stbds_shmode_func_wrapper(t,e,m) stbds_shmode_func(e,m) +#endif + +#endif // INCLUDE_STB_DS_H + + +////////////////////////////////////////////////////////////////////////////// +// +// IMPLEMENTATION +// + +#ifdef STB_DS_IMPLEMENTATION +#include +#include + +#ifndef STBDS_ASSERT +#define STBDS_ASSERT_WAS_UNDEFINED +#define STBDS_ASSERT(x) ((void) 0) +#endif + +#ifdef STBDS_STATISTICS +#define STBDS_STATS(x) x +size_t stbds_array_grow; +size_t stbds_hash_grow; +size_t stbds_hash_shrink; +size_t stbds_hash_rebuild; +size_t stbds_hash_probes; +size_t stbds_hash_alloc; +size_t stbds_rehash_probes; +size_t stbds_rehash_items; +#else +#define STBDS_STATS(x) +#endif + +// +// stbds_arr implementation +// + +//int *prev_allocs[65536]; +//int num_prev; + +void *stbds_arrgrowf(void *a, size_t elemsize, size_t addlen, size_t min_cap) +{ + stbds_array_header temp={0}; // force debugging + void *b; + size_t min_len = stbds_arrlen(a) + addlen; + (void) sizeof(temp); + + // compute the minimum capacity needed + if (min_len > min_cap) + min_cap = min_len; + + if (min_cap <= stbds_arrcap(a)) + return a; + + // increase needed capacity to guarantee O(1) amortized + if (min_cap < 2 * stbds_arrcap(a)) + min_cap = 2 * stbds_arrcap(a); + else if (min_cap < 4) + min_cap = 4; + + //if (num_prev < 65536) if (a) prev_allocs[num_prev++] = (int *) ((char *) a+1); + //if (num_prev == 2201) + // num_prev = num_prev; + b = STBDS_REALLOC(NULL, (a) ? stbds_header(a) : 0, elemsize * min_cap + sizeof(stbds_array_header)); + //if (num_prev < 65536) prev_allocs[num_prev++] = (int *) (char *) b; + b = (char *) b + sizeof(stbds_array_header); + if (a == NULL) { + stbds_header(b)->length = 0; + stbds_header(b)->hash_table = 0; + stbds_header(b)->temp = 0; + } else { + STBDS_STATS(++stbds_array_grow); + } + stbds_header(b)->capacity = min_cap; + + return b; +} + +void stbds_arrfreef(void *a) +{ + STBDS_FREE(NULL, stbds_header(a)); +} + +// +// stbds_hm hash table implementation +// + +#ifdef STBDS_INTERNAL_SMALL_BUCKET +#define STBDS_BUCKET_LENGTH 4 +#else +#define STBDS_BUCKET_LENGTH 8 +#endif + +#define STBDS_BUCKET_SHIFT (STBDS_BUCKET_LENGTH == 8 ? 3 : 2) +#define STBDS_BUCKET_MASK (STBDS_BUCKET_LENGTH-1) +#define STBDS_CACHE_LINE_SIZE 64 + +#define STBDS_ALIGN_FWD(n,a) (((n) + (a) - 1) & ~((a)-1)) + +typedef struct +{ + size_t hash [STBDS_BUCKET_LENGTH]; + ptrdiff_t index[STBDS_BUCKET_LENGTH]; +} stbds_hash_bucket; // in 32-bit, this is one 64-byte cache line; in 64-bit, each array is one 64-byte cache line + +typedef struct +{ + char * temp_key; // this MUST be the first field of the hash table + size_t slot_count; + size_t used_count; + size_t used_count_threshold; + size_t used_count_shrink_threshold; + size_t tombstone_count; + size_t tombstone_count_threshold; + size_t seed; + size_t slot_count_log2; + stbds_string_arena string; + stbds_hash_bucket *storage; // not a separate allocation, just 64-byte aligned storage after this struct +} stbds_hash_index; + +#define STBDS_INDEX_EMPTY -1 +#define STBDS_INDEX_DELETED -2 +#define STBDS_INDEX_IN_USE(x) ((x) >= 0) + +#define STBDS_HASH_EMPTY 0 +#define STBDS_HASH_DELETED 1 + +static size_t stbds_hash_seed=0x31415926; + +void stbds_rand_seed(size_t seed) +{ + stbds_hash_seed = seed; +} + +#define stbds_load_32_or_64(var, temp, v32, v64_hi, v64_lo) \ + temp = v64_lo ^ v32, temp <<= 16, temp <<= 16, temp >>= 16, temp >>= 16, /* discard if 32-bit */ \ + var = v64_hi, var <<= 16, var <<= 16, /* discard if 32-bit */ \ + var ^= temp ^ v32 + +#define STBDS_SIZE_T_BITS ((sizeof (size_t)) * 8) + +static size_t stbds_probe_position(size_t hash, size_t slot_count, size_t slot_log2) +{ + size_t pos; + STBDS_NOTUSED(slot_log2); + pos = hash & (slot_count-1); + #ifdef STBDS_INTERNAL_BUCKET_START + pos &= ~STBDS_BUCKET_MASK; + #endif + return pos; +} + +static size_t stbds_log2(size_t slot_count) +{ + size_t n=0; + while (slot_count > 1) { + slot_count >>= 1; + ++n; + } + return n; +} + +static stbds_hash_index *stbds_make_hash_index(size_t slot_count, stbds_hash_index *ot) +{ + stbds_hash_index *t; + t = (stbds_hash_index *) STBDS_REALLOC(NULL,0,(slot_count >> STBDS_BUCKET_SHIFT) * sizeof(stbds_hash_bucket) + sizeof(stbds_hash_index) + STBDS_CACHE_LINE_SIZE-1); + t->storage = (stbds_hash_bucket *) STBDS_ALIGN_FWD((size_t) (t+1), STBDS_CACHE_LINE_SIZE); + t->slot_count = slot_count; + t->slot_count_log2 = stbds_log2(slot_count); + t->tombstone_count = 0; + t->used_count = 0; + + #if 0 // A1 + t->used_count_threshold = slot_count*12/16; // if 12/16th of table is occupied, grow + t->tombstone_count_threshold = slot_count* 2/16; // if tombstones are 2/16th of table, rebuild + t->used_count_shrink_threshold = slot_count* 4/16; // if table is only 4/16th full, shrink + #elif 1 // A2 + //t->used_count_threshold = slot_count*12/16; // if 12/16th of table is occupied, grow + //t->tombstone_count_threshold = slot_count* 3/16; // if tombstones are 3/16th of table, rebuild + //t->used_count_shrink_threshold = slot_count* 4/16; // if table is only 4/16th full, shrink + + // compute without overflowing + t->used_count_threshold = slot_count - (slot_count>>2); + t->tombstone_count_threshold = (slot_count>>3) + (slot_count>>4); + t->used_count_shrink_threshold = slot_count >> 2; + + #elif 0 // B1 + t->used_count_threshold = slot_count*13/16; // if 13/16th of table is occupied, grow + t->tombstone_count_threshold = slot_count* 2/16; // if tombstones are 2/16th of table, rebuild + t->used_count_shrink_threshold = slot_count* 5/16; // if table is only 5/16th full, shrink + #else // C1 + t->used_count_threshold = slot_count*14/16; // if 14/16th of table is occupied, grow + t->tombstone_count_threshold = slot_count* 2/16; // if tombstones are 2/16th of table, rebuild + t->used_count_shrink_threshold = slot_count* 6/16; // if table is only 6/16th full, shrink + #endif + // Following statistics were measured on a Core i7-6700 @ 4.00Ghz, compiled with clang 7.0.1 -O2 + // Note that the larger tables have high variance as they were run fewer times + // A1 A2 B1 C1 + // 0.10ms : 0.10ms : 0.10ms : 0.11ms : 2,000 inserts creating 2K table + // 0.96ms : 0.95ms : 0.97ms : 1.04ms : 20,000 inserts creating 20K table + // 14.48ms : 14.46ms : 10.63ms : 11.00ms : 200,000 inserts creating 200K table + // 195.74ms : 196.35ms : 203.69ms : 214.92ms : 2,000,000 inserts creating 2M table + // 2193.88ms : 2209.22ms : 2285.54ms : 2437.17ms : 20,000,000 inserts creating 20M table + // 65.27ms : 53.77ms : 65.33ms : 65.47ms : 500,000 inserts & deletes in 2K table + // 72.78ms : 62.45ms : 71.95ms : 72.85ms : 500,000 inserts & deletes in 20K table + // 89.47ms : 77.72ms : 96.49ms : 96.75ms : 500,000 inserts & deletes in 200K table + // 97.58ms : 98.14ms : 97.18ms : 97.53ms : 500,000 inserts & deletes in 2M table + // 118.61ms : 119.62ms : 120.16ms : 118.86ms : 500,000 inserts & deletes in 20M table + // 192.11ms : 194.39ms : 196.38ms : 195.73ms : 500,000 inserts & deletes in 200M table + + if (slot_count <= STBDS_BUCKET_LENGTH) + t->used_count_shrink_threshold = 0; + // to avoid infinite loop, we need to guarantee that at least one slot is empty and will terminate probes + STBDS_ASSERT(t->used_count_threshold + t->tombstone_count_threshold < t->slot_count); + STBDS_STATS(++stbds_hash_alloc); + if (ot) { + t->string = ot->string; + // reuse old seed so we can reuse old hashes so below "copy out old data" doesn't do any hashing + t->seed = ot->seed; + } else { + size_t a,b,temp; + memset(&t->string, 0, sizeof(t->string)); + t->seed = stbds_hash_seed; + // LCG + // in 32-bit, a = 2147001325 b = 715136305 + // in 64-bit, a = 2862933555777941757 b = 3037000493 + stbds_load_32_or_64(a,temp, 2147001325, 0x27bb2ee6, 0x87b0b0fd); + stbds_load_32_or_64(b,temp, 715136305, 0, 0xb504f32d); + stbds_hash_seed = stbds_hash_seed * a + b; + } + + { + size_t i,j; + for (i=0; i < slot_count >> STBDS_BUCKET_SHIFT; ++i) { + stbds_hash_bucket *b = &t->storage[i]; + for (j=0; j < STBDS_BUCKET_LENGTH; ++j) + b->hash[j] = STBDS_HASH_EMPTY; + for (j=0; j < STBDS_BUCKET_LENGTH; ++j) + b->index[j] = STBDS_INDEX_EMPTY; + } + } + + // copy out the old data, if any + if (ot) { + size_t i,j; + t->used_count = ot->used_count; + for (i=0; i < ot->slot_count >> STBDS_BUCKET_SHIFT; ++i) { + stbds_hash_bucket *ob = &ot->storage[i]; + for (j=0; j < STBDS_BUCKET_LENGTH; ++j) { + if (STBDS_INDEX_IN_USE(ob->index[j])) { + size_t hash = ob->hash[j]; + size_t pos = stbds_probe_position(hash, t->slot_count, t->slot_count_log2); + size_t step = STBDS_BUCKET_LENGTH; + STBDS_STATS(++stbds_rehash_items); + for (;;) { + size_t limit,z; + stbds_hash_bucket *bucket; + bucket = &t->storage[pos >> STBDS_BUCKET_SHIFT]; + STBDS_STATS(++stbds_rehash_probes); + + for (z=pos & STBDS_BUCKET_MASK; z < STBDS_BUCKET_LENGTH; ++z) { + if (bucket->hash[z] == 0) { + bucket->hash[z] = hash; + bucket->index[z] = ob->index[j]; + goto done; + } + } + + limit = pos & STBDS_BUCKET_MASK; + for (z = 0; z < limit; ++z) { + if (bucket->hash[z] == 0) { + bucket->hash[z] = hash; + bucket->index[z] = ob->index[j]; + goto done; + } + } + + pos += step; // quadratic probing + step += STBDS_BUCKET_LENGTH; + pos &= (t->slot_count-1); + } + } + done: + ; + } + } + } + + return t; +} + +#define STBDS_ROTATE_LEFT(val, n) (((val) << (n)) | ((val) >> (STBDS_SIZE_T_BITS - (n)))) +#define STBDS_ROTATE_RIGHT(val, n) (((val) >> (n)) | ((val) << (STBDS_SIZE_T_BITS - (n)))) + +size_t stbds_hash_string(char *str, size_t seed) +{ + size_t hash = seed; + while (*str) + hash = STBDS_ROTATE_LEFT(hash, 9) + (unsigned char) *str++; + + // Thomas Wang 64-to-32 bit mix function, hopefully also works in 32 bits + hash ^= seed; + hash = (~hash) + (hash << 18); + hash ^= hash ^ STBDS_ROTATE_RIGHT(hash,31); + hash = hash * 21; + hash ^= hash ^ STBDS_ROTATE_RIGHT(hash,11); + hash += (hash << 6); + hash ^= STBDS_ROTATE_RIGHT(hash,22); + return hash+seed; +} + +#ifdef STBDS_SIPHASH_2_4 +#define STBDS_SIPHASH_C_ROUNDS 2 +#define STBDS_SIPHASH_D_ROUNDS 4 +typedef int STBDS_SIPHASH_2_4_can_only_be_used_in_64_bit_builds[sizeof(size_t) == 8 ? 1 : -1]; +#endif + +#ifndef STBDS_SIPHASH_C_ROUNDS +#define STBDS_SIPHASH_C_ROUNDS 1 +#endif +#ifndef STBDS_SIPHASH_D_ROUNDS +#define STBDS_SIPHASH_D_ROUNDS 1 +#endif + +#ifdef _MSC_VER +#pragma warning(push) +#pragma warning(disable:4127) // conditional expression is constant, for do..while(0) and sizeof()== +#endif + +static size_t stbds_siphash_bytes(void *p, size_t len, size_t seed) +{ + unsigned char *d = (unsigned char *) p; + size_t i,j; + size_t v0,v1,v2,v3, data; + + // hash that works on 32- or 64-bit registers without knowing which we have + // (computes different results on 32-bit and 64-bit platform) + // derived from siphash, but on 32-bit platforms very different as it uses 4 32-bit state not 4 64-bit + v0 = ((((size_t) 0x736f6d65 << 16) << 16) + 0x70736575) ^ seed; + v1 = ((((size_t) 0x646f7261 << 16) << 16) + 0x6e646f6d) ^ ~seed; + v2 = ((((size_t) 0x6c796765 << 16) << 16) + 0x6e657261) ^ seed; + v3 = ((((size_t) 0x74656462 << 16) << 16) + 0x79746573) ^ ~seed; + + #ifdef STBDS_TEST_SIPHASH_2_4 + // hardcoded with key material in the siphash test vectors + v0 ^= 0x0706050403020100ull ^ seed; + v1 ^= 0x0f0e0d0c0b0a0908ull ^ ~seed; + v2 ^= 0x0706050403020100ull ^ seed; + v3 ^= 0x0f0e0d0c0b0a0908ull ^ ~seed; + #endif + + #define STBDS_SIPROUND() \ + do { \ + v0 += v1; v1 = STBDS_ROTATE_LEFT(v1, 13); v1 ^= v0; v0 = STBDS_ROTATE_LEFT(v0,STBDS_SIZE_T_BITS/2); \ + v2 += v3; v3 = STBDS_ROTATE_LEFT(v3, 16); v3 ^= v2; \ + v2 += v1; v1 = STBDS_ROTATE_LEFT(v1, 17); v1 ^= v2; v2 = STBDS_ROTATE_LEFT(v2,STBDS_SIZE_T_BITS/2); \ + v0 += v3; v3 = STBDS_ROTATE_LEFT(v3, 21); v3 ^= v0; \ + } while (0) + + for (i=0; i+sizeof(size_t) <= len; i += sizeof(size_t), d += sizeof(size_t)) { + data = d[0] | (d[1] << 8) | (d[2] << 16) | (d[3] << 24); + data |= (size_t) (d[4] | (d[5] << 8) | (d[6] << 16) | (d[7] << 24)) << 16 << 16; // discarded if size_t == 4 + + v3 ^= data; + for (j=0; j < STBDS_SIPHASH_C_ROUNDS; ++j) + STBDS_SIPROUND(); + v0 ^= data; + } + data = len << (STBDS_SIZE_T_BITS-8); + switch (len - i) { + case 7: data |= ((size_t) d[6] << 24) << 24; // fall through + case 6: data |= ((size_t) d[5] << 20) << 20; // fall through + case 5: data |= ((size_t) d[4] << 16) << 16; // fall through + case 4: data |= (d[3] << 24); // fall through + case 3: data |= (d[2] << 16); // fall through + case 2: data |= (d[1] << 8); // fall through + case 1: data |= d[0]; // fall through + case 0: break; + } + v3 ^= data; + for (j=0; j < STBDS_SIPHASH_C_ROUNDS; ++j) + STBDS_SIPROUND(); + v0 ^= data; + v2 ^= 0xff; + for (j=0; j < STBDS_SIPHASH_D_ROUNDS; ++j) + STBDS_SIPROUND(); + +#ifdef STBDS_SIPHASH_2_4 + return v0^v1^v2^v3; +#else + return v1^v2^v3; // slightly stronger since v0^v3 in above cancels out final round operation? I tweeted at the authors of SipHash about this but they didn't reply +#endif +} + +size_t stbds_hash_bytes(void *p, size_t len, size_t seed) +{ +#ifdef STBDS_SIPHASH_2_4 + return stbds_siphash_bytes(p,len,seed); +#else + unsigned char *d = (unsigned char *) p; + + if (len == 4) { + unsigned int hash = d[0] | (d[1] << 8) | (d[2] << 16) | (d[3] << 24); + #if 0 + // HASH32-A Bob Jenkin's hash function w/o large constants + hash ^= seed; + hash -= (hash<<6); + hash ^= (hash>>17); + hash -= (hash<<9); + hash ^= seed; + hash ^= (hash<<4); + hash -= (hash<<3); + hash ^= (hash<<10); + hash ^= (hash>>15); + #elif 1 + // HASH32-BB Bob Jenkin's presumably-accidental version of Thomas Wang hash with rotates turned into shifts. + // Note that converting these back to rotates makes it run a lot slower, presumably due to collisions, so I'm + // not really sure what's going on. + hash ^= seed; + hash = (hash ^ 61) ^ (hash >> 16); + hash = hash + (hash << 3); + hash = hash ^ (hash >> 4); + hash = hash * 0x27d4eb2d; + hash ^= seed; + hash = hash ^ (hash >> 15); + #else // HASH32-C - Murmur3 + hash ^= seed; + hash *= 0xcc9e2d51; + hash = (hash << 17) | (hash >> 15); + hash *= 0x1b873593; + hash ^= seed; + hash = (hash << 19) | (hash >> 13); + hash = hash*5 + 0xe6546b64; + hash ^= hash >> 16; + hash *= 0x85ebca6b; + hash ^= seed; + hash ^= hash >> 13; + hash *= 0xc2b2ae35; + hash ^= hash >> 16; + #endif + // Following statistics were measured on a Core i7-6700 @ 4.00Ghz, compiled with clang 7.0.1 -O2 + // Note that the larger tables have high variance as they were run fewer times + // HASH32-A // HASH32-BB // HASH32-C + // 0.10ms // 0.10ms // 0.10ms : 2,000 inserts creating 2K table + // 0.96ms // 0.95ms // 0.99ms : 20,000 inserts creating 20K table + // 14.69ms // 14.43ms // 14.97ms : 200,000 inserts creating 200K table + // 199.99ms // 195.36ms // 202.05ms : 2,000,000 inserts creating 2M table + // 2234.84ms // 2187.74ms // 2240.38ms : 20,000,000 inserts creating 20M table + // 55.68ms // 53.72ms // 57.31ms : 500,000 inserts & deletes in 2K table + // 63.43ms // 61.99ms // 65.73ms : 500,000 inserts & deletes in 20K table + // 80.04ms // 77.96ms // 81.83ms : 500,000 inserts & deletes in 200K table + // 100.42ms // 97.40ms // 102.39ms : 500,000 inserts & deletes in 2M table + // 119.71ms // 120.59ms // 121.63ms : 500,000 inserts & deletes in 20M table + // 185.28ms // 195.15ms // 187.74ms : 500,000 inserts & deletes in 200M table + // 15.58ms // 14.79ms // 15.52ms : 200,000 inserts creating 200K table with varying key spacing + + return (((size_t) hash << 16 << 16) | hash) ^ seed; + } else if (len == 8 && sizeof(size_t) == 8) { + size_t hash = d[0] | (d[1] << 8) | (d[2] << 16) | (d[3] << 24); + hash |= (size_t) (d[4] | (d[5] << 8) | (d[6] << 16) | (d[7] << 24)) << 16 << 16; // avoid warning if size_t == 4 + hash ^= seed; + hash = (~hash) + (hash << 21); + hash ^= STBDS_ROTATE_RIGHT(hash,24); + hash *= 265; + hash ^= STBDS_ROTATE_RIGHT(hash,14); + hash ^= seed; + hash *= 21; + hash ^= STBDS_ROTATE_RIGHT(hash,28); + hash += (hash << 31); + hash = (~hash) + (hash << 18); + return hash; + } else { + return stbds_siphash_bytes(p,len,seed); + } +#endif +} +#ifdef _MSC_VER +#pragma warning(pop) +#endif + + +static int stbds_is_key_equal(void *a, size_t elemsize, void *key, size_t keysize, size_t keyoffset, int mode, size_t i) +{ + if (mode >= STBDS_HM_STRING) + return 0==strcmp((char *) key, * (char **) ((char *) a + elemsize*i + keyoffset)); + else + return 0==memcmp(key, (char *) a + elemsize*i + keyoffset, keysize); +} + +#define STBDS_HASH_TO_ARR(x,elemsize) ((char*) (x) - (elemsize)) +#define STBDS_ARR_TO_HASH(x,elemsize) ((char*) (x) + (elemsize)) + +#define stbds_hash_table(a) ((stbds_hash_index *) stbds_header(a)->hash_table) + +void stbds_hmfree_func(void *a, size_t elemsize) +{ + if (a == NULL) return; + if (stbds_hash_table(a) != NULL) { + if (stbds_hash_table(a)->string.mode == STBDS_SH_STRDUP) { + size_t i; + // skip 0th element, which is default + for (i=1; i < stbds_header(a)->length; ++i) + STBDS_FREE(NULL, *(char**) ((char *) a + elemsize*i)); + } + stbds_strreset(&stbds_hash_table(a)->string); + } + STBDS_FREE(NULL, stbds_header(a)->hash_table); + STBDS_FREE(NULL, stbds_header(a)); +} + +static ptrdiff_t stbds_hm_find_slot(void *a, size_t elemsize, void *key, size_t keysize, size_t keyoffset, int mode) +{ + void *raw_a = STBDS_HASH_TO_ARR(a,elemsize); + stbds_hash_index *table = stbds_hash_table(raw_a); + size_t hash = mode >= STBDS_HM_STRING ? stbds_hash_string((char*)key,table->seed) : stbds_hash_bytes(key, keysize,table->seed); + size_t step = STBDS_BUCKET_LENGTH; + size_t limit,i; + size_t pos; + stbds_hash_bucket *bucket; + + if (hash < 2) hash += 2; // stored hash values are forbidden from being 0, so we can detect empty slots + + pos = stbds_probe_position(hash, table->slot_count, table->slot_count_log2); + + for (;;) { + STBDS_STATS(++stbds_hash_probes); + bucket = &table->storage[pos >> STBDS_BUCKET_SHIFT]; + + // start searching from pos to end of bucket, this should help performance on small hash tables that fit in cache + for (i=pos & STBDS_BUCKET_MASK; i < STBDS_BUCKET_LENGTH; ++i) { + if (bucket->hash[i] == hash) { + if (stbds_is_key_equal(a, elemsize, key, keysize, keyoffset, mode, bucket->index[i])) { + return (pos & ~STBDS_BUCKET_MASK)+i; + } + } else if (bucket->hash[i] == STBDS_HASH_EMPTY) { + return -1; + } + } + + // search from beginning of bucket to pos + limit = pos & STBDS_BUCKET_MASK; + for (i = 0; i < limit; ++i) { + if (bucket->hash[i] == hash) { + if (stbds_is_key_equal(a, elemsize, key, keysize, keyoffset, mode, bucket->index[i])) { + return (pos & ~STBDS_BUCKET_MASK)+i; + } + } else if (bucket->hash[i] == STBDS_HASH_EMPTY) { + return -1; + } + } + + // quadratic probing + pos += step; + step += STBDS_BUCKET_LENGTH; + pos &= (table->slot_count-1); + } + /* NOTREACHED */ +} + +void * stbds_hmget_key_ts(void *a, size_t elemsize, void *key, size_t keysize, ptrdiff_t *temp, int mode) +{ + size_t keyoffset = 0; + if (a == NULL) { + // make it non-empty so we can return a temp + a = stbds_arrgrowf(0, elemsize, 0, 1); + stbds_header(a)->length += 1; + memset(a, 0, elemsize); + *temp = STBDS_INDEX_EMPTY; + // adjust a to point after the default element + return STBDS_ARR_TO_HASH(a,elemsize); + } else { + stbds_hash_index *table; + void *raw_a = STBDS_HASH_TO_ARR(a,elemsize); + // adjust a to point to the default element + table = (stbds_hash_index *) stbds_header(raw_a)->hash_table; + if (table == 0) { + *temp = -1; + } else { + ptrdiff_t slot = stbds_hm_find_slot(a, elemsize, key, keysize, keyoffset, mode); + if (slot < 0) { + *temp = STBDS_INDEX_EMPTY; + } else { + stbds_hash_bucket *b = &table->storage[slot >> STBDS_BUCKET_SHIFT]; + *temp = b->index[slot & STBDS_BUCKET_MASK]; + } + } + return a; + } +} + +void * stbds_hmget_key(void *a, size_t elemsize, void *key, size_t keysize, int mode) +{ + ptrdiff_t temp; + void *p = stbds_hmget_key_ts(a, elemsize, key, keysize, &temp, mode); + stbds_temp(STBDS_HASH_TO_ARR(p,elemsize)) = temp; + return p; +} + +void * stbds_hmput_default(void *a, size_t elemsize) +{ + // three cases: + // a is NULL <- allocate + // a has a hash table but no entries, because of shmode <- grow + // a has entries <- do nothing + if (a == NULL || stbds_header(STBDS_HASH_TO_ARR(a,elemsize))->length == 0) { + a = stbds_arrgrowf(a ? STBDS_HASH_TO_ARR(a,elemsize) : NULL, elemsize, 0, 1); + stbds_header(a)->length += 1; + memset(a, 0, elemsize); + a=STBDS_ARR_TO_HASH(a,elemsize); + } + return a; +} + +static char *stbds_strdup(char *str); + +void *stbds_hmput_key(void *a, size_t elemsize, void *key, size_t keysize, int mode) +{ + size_t keyoffset=0; + void *raw_a; + stbds_hash_index *table; + + if (a == NULL) { + a = stbds_arrgrowf(0, elemsize, 0, 1); + memset(a, 0, elemsize); + stbds_header(a)->length += 1; + // adjust a to point AFTER the default element + a = STBDS_ARR_TO_HASH(a,elemsize); + } + + // adjust a to point to the default element + raw_a = a; + a = STBDS_HASH_TO_ARR(a,elemsize); + + table = (stbds_hash_index *) stbds_header(a)->hash_table; + + if (table == NULL || table->used_count >= table->used_count_threshold) { + stbds_hash_index *nt; + size_t slot_count; + + slot_count = (table == NULL) ? STBDS_BUCKET_LENGTH : table->slot_count*2; + nt = stbds_make_hash_index(slot_count, table); + if (table) + STBDS_FREE(NULL, table); + else + nt->string.mode = mode >= STBDS_HM_STRING ? STBDS_SH_DEFAULT : 0; + stbds_header(a)->hash_table = table = nt; + STBDS_STATS(++stbds_hash_grow); + } + + // we iterate hash table explicitly because we want to track if we saw a tombstone + { + size_t hash = mode >= STBDS_HM_STRING ? stbds_hash_string((char*)key,table->seed) : stbds_hash_bytes(key, keysize,table->seed); + size_t step = STBDS_BUCKET_LENGTH; + size_t pos; + ptrdiff_t tombstone = -1; + stbds_hash_bucket *bucket; + + // stored hash values are forbidden from being 0, so we can detect empty slots to early out quickly + if (hash < 2) hash += 2; + + pos = stbds_probe_position(hash, table->slot_count, table->slot_count_log2); + + for (;;) { + size_t limit, i; + STBDS_STATS(++stbds_hash_probes); + bucket = &table->storage[pos >> STBDS_BUCKET_SHIFT]; + + // start searching from pos to end of bucket + for (i=pos & STBDS_BUCKET_MASK; i < STBDS_BUCKET_LENGTH; ++i) { + if (bucket->hash[i] == hash) { + if (stbds_is_key_equal(raw_a, elemsize, key, keysize, keyoffset, mode, bucket->index[i])) { + stbds_temp(a) = bucket->index[i]; + if (mode >= STBDS_HM_STRING) + stbds_temp_key(a) = * (char **) ((char *) raw_a + elemsize*bucket->index[i] + keyoffset); + return STBDS_ARR_TO_HASH(a,elemsize); + } + } else if (bucket->hash[i] == 0) { + pos = (pos & ~STBDS_BUCKET_MASK) + i; + goto found_empty_slot; + } else if (tombstone < 0) { + if (bucket->index[i] == STBDS_INDEX_DELETED) + tombstone = (ptrdiff_t) ((pos & ~STBDS_BUCKET_MASK) + i); + } + } + + // search from beginning of bucket to pos + limit = pos & STBDS_BUCKET_MASK; + for (i = 0; i < limit; ++i) { + if (bucket->hash[i] == hash) { + if (stbds_is_key_equal(raw_a, elemsize, key, keysize, keyoffset, mode, bucket->index[i])) { + stbds_temp(a) = bucket->index[i]; + return STBDS_ARR_TO_HASH(a,elemsize); + } + } else if (bucket->hash[i] == 0) { + pos = (pos & ~STBDS_BUCKET_MASK) + i; + goto found_empty_slot; + } else if (tombstone < 0) { + if (bucket->index[i] == STBDS_INDEX_DELETED) + tombstone = (ptrdiff_t) ((pos & ~STBDS_BUCKET_MASK) + i); + } + } + + // quadratic probing + pos += step; + step += STBDS_BUCKET_LENGTH; + pos &= (table->slot_count-1); + } + found_empty_slot: + if (tombstone >= 0) { + pos = tombstone; + --table->tombstone_count; + } + ++table->used_count; + + { + ptrdiff_t i = (ptrdiff_t) stbds_arrlen(a); + // we want to do stbds_arraddn(1), but we can't use the macros since we don't have something of the right type + if ((size_t) i+1 > stbds_arrcap(a)) + *(void **) &a = stbds_arrgrowf(a, elemsize, 1, 0); + raw_a = STBDS_ARR_TO_HASH(a,elemsize); + + STBDS_ASSERT((size_t) i+1 <= stbds_arrcap(a)); + stbds_header(a)->length = i+1; + bucket = &table->storage[pos >> STBDS_BUCKET_SHIFT]; + bucket->hash[pos & STBDS_BUCKET_MASK] = hash; + bucket->index[pos & STBDS_BUCKET_MASK] = i-1; + stbds_temp(a) = i-1; + + switch (table->string.mode) { + case STBDS_SH_STRDUP: stbds_temp_key(a) = *(char **) ((char *) a + elemsize*i) = stbds_strdup((char*) key); break; + case STBDS_SH_ARENA: stbds_temp_key(a) = *(char **) ((char *) a + elemsize*i) = stbds_stralloc(&table->string, (char*)key); break; + case STBDS_SH_DEFAULT: stbds_temp_key(a) = *(char **) ((char *) a + elemsize*i) = (char *) key; break; + default: memcpy((char *) a + elemsize*i, key, keysize); break; + } + } + return STBDS_ARR_TO_HASH(a,elemsize); + } +} + +void * stbds_shmode_func(size_t elemsize, int mode) +{ + void *a = stbds_arrgrowf(0, elemsize, 0, 1); + stbds_hash_index *h; + memset(a, 0, elemsize); + stbds_header(a)->length = 1; + stbds_header(a)->hash_table = h = (stbds_hash_index *) stbds_make_hash_index(STBDS_BUCKET_LENGTH, NULL); + h->string.mode = (unsigned char) mode; + return STBDS_ARR_TO_HASH(a,elemsize); +} + +void * stbds_hmdel_key(void *a, size_t elemsize, void *key, size_t keysize, size_t keyoffset, int mode) +{ + if (a == NULL) { + return 0; + } else { + stbds_hash_index *table; + void *raw_a = STBDS_HASH_TO_ARR(a,elemsize); + table = (stbds_hash_index *) stbds_header(raw_a)->hash_table; + stbds_temp(raw_a) = 0; + if (table == 0) { + return a; + } else { + ptrdiff_t slot; + slot = stbds_hm_find_slot(a, elemsize, key, keysize, keyoffset, mode); + if (slot < 0) + return a; + else { + stbds_hash_bucket *b = &table->storage[slot >> STBDS_BUCKET_SHIFT]; + int i = slot & STBDS_BUCKET_MASK; + ptrdiff_t old_index = b->index[i]; + ptrdiff_t final_index = (ptrdiff_t) stbds_arrlen(raw_a)-1-1; // minus one for the raw_a vs a, and minus one for 'last' + STBDS_ASSERT(slot < (ptrdiff_t) table->slot_count); + --table->used_count; + ++table->tombstone_count; + stbds_temp(raw_a) = 1; + STBDS_ASSERT(table->used_count >= 0); + //STBDS_ASSERT(table->tombstone_count < table->slot_count/4); + b->hash[i] = STBDS_HASH_DELETED; + b->index[i] = STBDS_INDEX_DELETED; + + if (mode == STBDS_HM_STRING && table->string.mode == STBDS_SH_STRDUP) + STBDS_FREE(NULL, *(char**) ((char *) a+elemsize*old_index)); + + // if indices are the same, memcpy is a no-op, but back-pointer-fixup will fail, so skip + if (old_index != final_index) { + // swap delete + memmove((char*) a + elemsize*old_index, (char*) a + elemsize*final_index, elemsize); + + // now find the slot for the last element + if (mode == STBDS_HM_STRING) + slot = stbds_hm_find_slot(a, elemsize, *(char**) ((char *) a+elemsize*old_index + keyoffset), keysize, keyoffset, mode); + else + slot = stbds_hm_find_slot(a, elemsize, (char* ) a+elemsize*old_index + keyoffset, keysize, keyoffset, mode); + STBDS_ASSERT(slot >= 0); + b = &table->storage[slot >> STBDS_BUCKET_SHIFT]; + i = slot & STBDS_BUCKET_MASK; + STBDS_ASSERT(b->index[i] == final_index); + b->index[i] = old_index; + } + stbds_header(raw_a)->length -= 1; + + if (table->used_count < table->used_count_shrink_threshold && table->slot_count > STBDS_BUCKET_LENGTH) { + stbds_header(raw_a)->hash_table = stbds_make_hash_index(table->slot_count>>1, table); + STBDS_FREE(NULL, table); + STBDS_STATS(++stbds_hash_shrink); + } else if (table->tombstone_count > table->tombstone_count_threshold) { + stbds_header(raw_a)->hash_table = stbds_make_hash_index(table->slot_count , table); + STBDS_FREE(NULL, table); + STBDS_STATS(++stbds_hash_rebuild); + } + + return a; + } + } + } + /* NOTREACHED */ +} + +static char *stbds_strdup(char *str) +{ + // to keep replaceable allocator simple, we don't want to use strdup. + // rolling our own also avoids problem of strdup vs _strdup + size_t len = strlen(str)+1; + char *p = (char*) STBDS_REALLOC(NULL, 0, len); + memmove(p, str, len); + return p; +} + +#ifndef STBDS_STRING_ARENA_BLOCKSIZE_MIN +#define STBDS_STRING_ARENA_BLOCKSIZE_MIN 512u +#endif +#ifndef STBDS_STRING_ARENA_BLOCKSIZE_MAX +#define STBDS_STRING_ARENA_BLOCKSIZE_MAX (1u<<20) +#endif + +char *stbds_stralloc(stbds_string_arena *a, char *str) +{ + char *p; + size_t len = strlen(str)+1; + if (len > a->remaining) { + // compute the next blocksize + size_t blocksize = a->block; + + // size is 512, 512, 1024, 1024, 2048, 2048, 4096, 4096, etc., so that + // there are log(SIZE) allocations to free when we destroy the table + blocksize = (size_t) (STBDS_STRING_ARENA_BLOCKSIZE_MIN) << (blocksize>>1); + + // if size is under 1M, advance to next blocktype + if (blocksize < (size_t)(STBDS_STRING_ARENA_BLOCKSIZE_MAX)) + ++a->block; + + if (len > blocksize) { + // if string is larger than blocksize, then just allocate the full size. + // note that we still advance string_block so block size will continue + // increasing, so e.g. if somebody only calls this with 1000-long strings, + // eventually the arena will start doubling and handling those as well + stbds_string_block *sb = (stbds_string_block *) STBDS_REALLOC(NULL, 0, sizeof(*sb)-8 + len); + memmove(sb->storage, str, len); + if (a->storage) { + // insert it after the first element, so that we don't waste the space there + sb->next = a->storage->next; + a->storage->next = sb; + } else { + sb->next = 0; + a->storage = sb; + a->remaining = 0; // this is redundant, but good for clarity + } + return sb->storage; + } else { + stbds_string_block *sb = (stbds_string_block *) STBDS_REALLOC(NULL, 0, sizeof(*sb)-8 + blocksize); + sb->next = a->storage; + a->storage = sb; + a->remaining = blocksize; + } + } + + STBDS_ASSERT(len <= a->remaining); + p = a->storage->storage + a->remaining - len; + a->remaining -= len; + memmove(p, str, len); + return p; +} + +void stbds_strreset(stbds_string_arena *a) +{ + stbds_string_block *x,*y; + x = a->storage; + while (x) { + y = x->next; + STBDS_FREE(NULL, x); + x = y; + } + memset(a, 0, sizeof(*a)); +} + +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// UNIT TESTS +// + +#ifdef STBDS_UNIT_TESTS +#include +#ifdef STBDS_ASSERT_WAS_UNDEFINED +#undef STBDS_ASSERT +#endif +#ifndef STBDS_ASSERT +#define STBDS_ASSERT assert +#include +#endif + +typedef struct { int key,b,c,d; } stbds_struct; +typedef struct { int key[2],b,c,d; } stbds_struct2; + +static char buffer[256]; +char *strkey(int n) +{ +#if defined(_WIN32) && defined(__STDC_WANT_SECURE_LIB__) + sprintf_s(buffer, sizeof(buffer), "test_%d", n); +#else + sprintf(buffer, "test_%d", n); +#endif + return buffer; +} + +void stbds_unit_tests(void) +{ +#if defined(_MSC_VER) && _MSC_VER <= 1200 && defined(__cplusplus) + // VC6 C++ doesn't like the template<> trick on unnamed structures, so do nothing! + STBDS_ASSERT(0); +#else + const int testsize = 100000; + const int testsize2 = testsize/20; + int *arr=NULL; + struct { int key; int value; } *intmap = NULL; + struct { char *key; int value; } *strmap = NULL, s; + struct { stbds_struct key; int value; } *map = NULL; + stbds_struct *map2 = NULL; + stbds_struct2 *map3 = NULL; + stbds_string_arena sa = { 0 }; + int key3[2] = { 1,2 }; + ptrdiff_t temp; + + int i,j; + + STBDS_ASSERT(arrlen(arr)==0); + for (i=0; i < 20000; i += 50) { + for (j=0; j < i; ++j) + arrpush(arr,j); + arrfree(arr); + } + + for (i=0; i < 4; ++i) { + arrpush(arr,1); arrpush(arr,2); arrpush(arr,3); arrpush(arr,4); + arrdel(arr,i); + arrfree(arr); + arrpush(arr,1); arrpush(arr,2); arrpush(arr,3); arrpush(arr,4); + arrdelswap(arr,i); + arrfree(arr); + } + + for (i=0; i < 5; ++i) { + arrpush(arr,1); arrpush(arr,2); arrpush(arr,3); arrpush(arr,4); + stbds_arrins(arr,i,5); + STBDS_ASSERT(arr[i] == 5); + if (i < 4) + STBDS_ASSERT(arr[4] == 4); + arrfree(arr); + } + + i = 1; + STBDS_ASSERT(hmgeti(intmap,i) == -1); + hmdefault(intmap, -2); + STBDS_ASSERT(hmgeti(intmap, i) == -1); + STBDS_ASSERT(hmget (intmap, i) == -2); + for (i=0; i < testsize; i+=2) + hmput(intmap, i, i*5); + for (i=0; i < testsize; i+=1) { + if (i & 1) STBDS_ASSERT(hmget(intmap, i) == -2 ); + else STBDS_ASSERT(hmget(intmap, i) == i*5); + if (i & 1) STBDS_ASSERT(hmget_ts(intmap, i, temp) == -2 ); + else STBDS_ASSERT(hmget_ts(intmap, i, temp) == i*5); + } + for (i=0; i < testsize; i+=2) + hmput(intmap, i, i*3); + for (i=0; i < testsize; i+=1) + if (i & 1) STBDS_ASSERT(hmget(intmap, i) == -2 ); + else STBDS_ASSERT(hmget(intmap, i) == i*3); + for (i=2; i < testsize; i+=4) + hmdel(intmap, i); // delete half the entries + for (i=0; i < testsize; i+=1) + if (i & 3) STBDS_ASSERT(hmget(intmap, i) == -2 ); + else STBDS_ASSERT(hmget(intmap, i) == i*3); + for (i=0; i < testsize; i+=1) + hmdel(intmap, i); // delete the rest of the entries + for (i=0; i < testsize; i+=1) + STBDS_ASSERT(hmget(intmap, i) == -2 ); + hmfree(intmap); + for (i=0; i < testsize; i+=2) + hmput(intmap, i, i*3); + hmfree(intmap); + + #if defined(__clang__) || defined(__GNUC__) + #ifndef __cplusplus + intmap = NULL; + hmput(intmap, 15, 7); + hmput(intmap, 11, 3); + hmput(intmap, 9, 5); + STBDS_ASSERT(hmget(intmap, 9) == 5); + STBDS_ASSERT(hmget(intmap, 11) == 3); + STBDS_ASSERT(hmget(intmap, 15) == 7); + #endif + #endif + + for (i=0; i < testsize; ++i) + stralloc(&sa, strkey(i)); + strreset(&sa); + + { + s.key = "a", s.value = 1; + shputs(strmap, s); + STBDS_ASSERT(*strmap[0].key == 'a'); + STBDS_ASSERT(strmap[0].key == s.key); + STBDS_ASSERT(strmap[0].value == s.value); + shfree(strmap); + } + + { + s.key = "a", s.value = 1; + sh_new_strdup(strmap); + shputs(strmap, s); + STBDS_ASSERT(*strmap[0].key == 'a'); + STBDS_ASSERT(strmap[0].key != s.key); + STBDS_ASSERT(strmap[0].value == s.value); + shfree(strmap); + } + + { + s.key = "a", s.value = 1; + sh_new_arena(strmap); + shputs(strmap, s); + STBDS_ASSERT(*strmap[0].key == 'a'); + STBDS_ASSERT(strmap[0].key != s.key); + STBDS_ASSERT(strmap[0].value == s.value); + shfree(strmap); + } + + for (j=0; j < 2; ++j) { + STBDS_ASSERT(shgeti(strmap,"foo") == -1); + if (j == 0) + sh_new_strdup(strmap); + else + sh_new_arena(strmap); + STBDS_ASSERT(shgeti(strmap,"foo") == -1); + shdefault(strmap, -2); + STBDS_ASSERT(shgeti(strmap,"foo") == -1); + for (i=0; i < testsize; i+=2) + shput(strmap, strkey(i), i*3); + for (i=0; i < testsize; i+=1) + if (i & 1) STBDS_ASSERT(shget(strmap, strkey(i)) == -2 ); + else STBDS_ASSERT(shget(strmap, strkey(i)) == i*3); + for (i=2; i < testsize; i+=4) + shdel(strmap, strkey(i)); // delete half the entries + for (i=0; i < testsize; i+=1) + if (i & 3) STBDS_ASSERT(shget(strmap, strkey(i)) == -2 ); + else STBDS_ASSERT(shget(strmap, strkey(i)) == i*3); + for (i=0; i < testsize; i+=1) + shdel(strmap, strkey(i)); // delete the rest of the entries + for (i=0; i < testsize; i+=1) + STBDS_ASSERT(shget(strmap, strkey(i)) == -2 ); + shfree(strmap); + } + + { + struct { char *key; char value; } *hash = NULL; + char name[4] = "jen"; + shput(hash, "bob" , 'h'); + shput(hash, "sally" , 'e'); + shput(hash, "fred" , 'l'); + shput(hash, "jen" , 'x'); + shput(hash, "doug" , 'o'); + + shput(hash, name , 'l'); + shfree(hash); + } + + for (i=0; i < testsize; i += 2) { + stbds_struct s = { i,i*2,i*3,i*4 }; + hmput(map, s, i*5); + } + + for (i=0; i < testsize; i += 1) { + stbds_struct s = { i,i*2,i*3 ,i*4 }; + stbds_struct t = { i,i*2,i*3+1,i*4 }; + if (i & 1) STBDS_ASSERT(hmget(map, s) == 0); + else STBDS_ASSERT(hmget(map, s) == i*5); + if (i & 1) STBDS_ASSERT(hmget_ts(map, s, temp) == 0); + else STBDS_ASSERT(hmget_ts(map, s, temp) == i*5); + //STBDS_ASSERT(hmget(map, t.key) == 0); + } + + for (i=0; i < testsize; i += 2) { + stbds_struct s = { i,i*2,i*3,i*4 }; + hmputs(map2, s); + } + hmfree(map); + + for (i=0; i < testsize; i += 1) { + stbds_struct s = { i,i*2,i*3,i*4 }; + stbds_struct t = { i,i*2,i*3+1,i*4 }; + if (i & 1) STBDS_ASSERT(hmgets(map2, s.key).d == 0); + else STBDS_ASSERT(hmgets(map2, s.key).d == i*4); + //STBDS_ASSERT(hmgetp(map2, t.key) == 0); + } + hmfree(map2); + + for (i=0; i < testsize; i += 2) { + stbds_struct2 s = { { i,i*2 }, i*3,i*4, i*5 }; + hmputs(map3, s); + } + for (i=0; i < testsize; i += 1) { + stbds_struct2 s = { { i,i*2}, i*3, i*4, i*5 }; + stbds_struct2 t = { { i,i*2}, i*3+1, i*4, i*5 }; + if (i & 1) STBDS_ASSERT(hmgets(map3, s.key).d == 0); + else STBDS_ASSERT(hmgets(map3, s.key).d == i*5); + //STBDS_ASSERT(hmgetp(map3, t.key) == 0); + } +#endif +} +#endif + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2019 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_image.h b/include/stb_image.h new file mode 100644 index 0000000..9eedabe --- /dev/null +++ b/include/stb_image.h @@ -0,0 +1,7988 @@ +/* stb_image - v2.30 - public domain image loader - http://nothings.org/stb + no warranty implied; use at your own risk + + Do this: + #define STB_IMAGE_IMPLEMENTATION + before you include this file in *one* C or C++ file to create the implementation. + + // i.e. it should look like this: + #include ... + #include ... + #include ... + #define STB_IMAGE_IMPLEMENTATION + #include "stb_image.h" + + You can #define STBI_ASSERT(x) before the #include to avoid using assert.h. + And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free + + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8/16-bit-per-channel + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels, 8/16 bit-per-channel) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + PNM (PPM and PGM binary only) + + Animated GIF still needs a proper API, but here's one way to do it: + http://gist.github.com/urraka/685d9a6340b26b830d49 + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON) + + Full documentation under "DOCUMENTATION" below. + + +LICENSE + + See end of file for license information. + +RECENT REVISION HISTORY: + + 2.30 (2024-05-31) avoid erroneous gcc warning + 2.29 (2023-05-xx) optimizations + 2.28 (2023-01-29) many error fixes, security errors, just tons of stuff + 2.27 (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes + 2.26 (2020-07-13) many minor fixes + 2.25 (2020-02-02) fix warnings + 2.24 (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically + 2.23 (2019-08-11) fix clang static analysis warning + 2.22 (2019-03-04) gif fixes, fix warnings + 2.21 (2019-02-25) fix typo in comment + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings + 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes + 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64 + RGB-format JPEG; remove white matting in PSD; + allocate large structures on the stack; + correct channel count for PNG & BMP + 2.10 (2016-01-22) avoid warning introduced in 2.09 + 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED + + See end of file for full revision history. + + + ============================ Contributors ========================= + + Image formats Extensions, features + Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info) + Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info) + Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG) + Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks) + Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG) + Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip) + Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD) + github:urraka (animated gif) Junggon Kim (PNM comments) + Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA) + socks-the-fox (16-bit PNG) + Jeremy Sawicki (handle all ImageNet JPGs) + Optimizations & bugfixes Mikhail Morozov (1-bit BMP) + Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query) + Arseny Kapoulkine Simon Breuss (16-bit PNM) + John-Mark Allen + Carmelo J Fdez-Aguera + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Jerry Jansson Joseph Thomson Blazej Dariusz Roszkowski + Phil Jordan Dave Moore Roy Eltham + Hayaki Saito Nathan Reed Won Chun + Luke Graham Johan Duparc Nick Verigakis the Horde3D community + Thomas Ruf Ronny Chevalier github:rlyeh + Janez Zemva John Bartholomew Michal Cichon github:romigrou + Jonathan Blow Ken Hamada Tero Hanninen github:svdijk + Eugene Golushkov Laurent Gomila Cort Stratton github:snagar + Aruelien Pocheville Sergio Gonzalez Thibault Reuille github:Zelex + Cass Everitt Ryamond Barbiero github:grim210 + Paul Du Bois Engin Manap Aldo Culquicondor github:sammyhw + Philipp Wiesemann Dale Weiler Oriol Ferrer Mesia github:phprus + Josh Tobin Neil Bickford Matthew Gregan github:poppolopoppo + Julian Raschke Gregory Mullen Christian Floisand github:darealshinji + Baldur Karlsson Kevin Schmidt JR Smith github:Michaelangel007 + Brad Weinberger Matvey Cherevko github:mosra + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + Jacko Dirks + + To add your name to the credits, pick a random blank space in the middle and fill it. + 80% of merge conflicts on stb PRs are due to people adding their name at the end + of the credits. +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below for HDR usage): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data); +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *channels_in_file -- outputs # of image components in image file +// int desired_channels -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data, or NULL on an allocation failure or if the image is +// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'desired_channels' if desired_channels is non-zero, or +// *channels_in_file otherwise. If desired_channels is non-zero, +// *channels_in_file has the number of components that _would_ have been +// output otherwise. E.g. if you set desired_channels to 4, you will always +// get RGBA output, but you can check *channels_in_file to see if it's trivially +// opaque because e.g. there were only 3 channels in the source image. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *channels_in_file will be unchanged. The function +// stbi_failure_reason() can be queried for an extremely brief, end-user +// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS +// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// To query the width, height and component count of an image without having to +// decode the full file, you can use the stbi_info family of functions: +// +// int x,y,n,ok; +// ok = stbi_info(filename, &x, &y, &n); +// // returns ok=1 and sets x, y, n if image is a supported format, +// // 0 otherwise. +// +// Note that stb_image pervasively uses ints in its public API for sizes, +// including sizes of memory buffers. This is now part of the API and thus +// hard to change without causing breakage. As a result, the various image +// loaders all have certain limits on image size; these differ somewhat +// by format but generally boil down to either just under 2GB or just under +// 1GB. When the decoded image would be larger than this, stb_image decoding +// will fail. +// +// Additionally, stb_image will reject image files that have any of their +// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS, +// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit, +// the only way to have an image with such dimensions load correctly +// is for it to have a rather extreme aspect ratio. Either way, the +// assumption here is that such larger images are likely to be malformed +// or malicious. If you do need to load an image with individual dimensions +// larger than that, and it still fits in the overall size limit, you can +// #define STBI_MAX_DIMENSIONS on your own to be something larger. +// +// =========================================================================== +// +// UNICODE: +// +// If compiling for Windows and you wish to use Unicode filenames, compile +// with +// #define STBI_WINDOWS_UTF8 +// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert +// Windows wchar_t filenames to utf8. +// +// =========================================================================== +// +// Philosophy +// +// stb libraries are designed with the following priorities: +// +// 1. easy to use +// 2. easy to maintain +// 3. good performance +// +// Sometimes I let "good performance" creep up in priority over "easy to maintain", +// and for best performance I may provide less-easy-to-use APIs that give higher +// performance, in addition to the easy-to-use ones. Nevertheless, it's important +// to keep in mind that from the standpoint of you, a client of this library, +// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all. +// +// Some secondary priorities arise directly from the first two, some of which +// provide more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small source code footprint ("easy to maintain") +// - No dependencies ("ease of use") +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). +// +// =========================================================================== +// +// SIMD support +// +// The JPEG decoder will try to automatically use SIMD kernels on x86 when +// supported by the compiler. For ARM Neon support, you must explicitly +// request it. +// +// (The old do-it-yourself SIMD API is no longer supported in the current +// code.) +// +// On x86, SSE2 will automatically be used when available based on a run-time +// test; if not, the generic C versions are used as a fall-back. On ARM targets, +// the typical path is to have separate builds for NEON and non-NEON devices +// (at least this is true for iOS and Android). Therefore, the NEON support is +// toggled by a build flag: define STBI_NEON to get NEON loops. +// +// If for some reason you do not want to use any of SIMD code, or if +// you have issues compiling it, you can disable it entirely by +// defining STBI_NO_SIMD. +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image supports loading HDR images in general, and currently the Radiance +// .HDR file format specifically. You can still load any file through the existing +// interface; if you attempt to load an HDR file, it will be automatically remapped +// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// iPhone PNG support: +// +// We optionally support converting iPhone-formatted PNGs (which store +// premultiplied BGRA) back to RGB, even though they're internally encoded +// differently. To enable this conversion, call +// stbi_convert_iphone_png_to_rgb(1). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// ADDITIONAL CONFIGURATION +// +// - You can suppress implementation of any of the decoders to reduce +// your code footprint by #defining one or more of the following +// symbols before creating the implementation. +// +// STBI_NO_JPEG +// STBI_NO_PNG +// STBI_NO_BMP +// STBI_NO_PSD +// STBI_NO_TGA +// STBI_NO_GIF +// STBI_NO_HDR +// STBI_NO_PIC +// STBI_NO_PNM (.ppm and .pgm) +// +// - You can request *only* certain decoders and suppress all other ones +// (this will be more forward-compatible, as addition of new decoders +// doesn't require you to disable them explicitly): +// +// STBI_ONLY_JPEG +// STBI_ONLY_PNG +// STBI_ONLY_BMP +// STBI_ONLY_PSD +// STBI_ONLY_TGA +// STBI_ONLY_GIF +// STBI_ONLY_HDR +// STBI_ONLY_PIC +// STBI_ONLY_PNM (.ppm and .pgm) +// +// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still +// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB +// +// - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater +// than that size (in either width or height) without further processing. +// This is to let programs in the wild set an upper bound to prevent +// denial-of-service attacks on untrusted data, as one could generate a +// valid image of gigantic dimensions and force stb_image to allocate a +// huge block of memory and spend disproportionate time decoding it. By +// default this is set to (1 << 24), which is 16777216, but that's still +// very big. + +#ifndef STBI_NO_STDIO +#include +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for desired_channels + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +#include +typedef unsigned char stbi_uc; +typedef unsigned short stbi_us; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef STBIDEF +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +//////////////////////////////////// +// +// 8-bits-per-channel interface +// + +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +#endif + +#ifdef STBI_WINDOWS_UTF8 +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif + +//////////////////////////////////// +// +// 16-bits-per-channel interface +// + +STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +#endif + +//////////////////////////////////// +// +// float-per-channel interface +// +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); + #endif +#endif + +#ifndef STBI_NO_HDR + STBIDEF void stbi_hdr_to_ldr_gamma(float gamma); + STBIDEF void stbi_hdr_to_ldr_scale(float scale); +#endif // STBI_NO_HDR + +#ifndef STBI_NO_LINEAR + STBIDEF void stbi_ldr_to_hdr_gamma(float gamma); + STBIDEF void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_LINEAR + +// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename); +STBIDEF int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// on most compilers (and ALL modern mainstream compilers) this is threadsafe +STBIDEF const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +STBIDEF void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len); +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user); + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit (char const *filename); +STBIDEF int stbi_is_16_bit_from_file(FILE *f); +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + +// flip the image vertically, so the first pixel in the output array is the bottom left +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip); + +// as above, but only applies to images loaded on the thread that calls the function +// this function is only available if your compiler supports thread-local variables; +// calling it will fail to link if your compiler doesn't +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply); +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert); +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip); + +// ZLIB client - used by PNG, available for other purposes + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header); +STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifdef STB_IMAGE_IMPLEMENTATION + +#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \ + || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \ + || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \ + || defined(STBI_ONLY_ZLIB) + #ifndef STBI_ONLY_JPEG + #define STBI_NO_JPEG + #endif + #ifndef STBI_ONLY_PNG + #define STBI_NO_PNG + #endif + #ifndef STBI_ONLY_BMP + #define STBI_NO_BMP + #endif + #ifndef STBI_ONLY_PSD + #define STBI_NO_PSD + #endif + #ifndef STBI_ONLY_TGA + #define STBI_NO_TGA + #endif + #ifndef STBI_ONLY_GIF + #define STBI_NO_GIF + #endif + #ifndef STBI_ONLY_HDR + #define STBI_NO_HDR + #endif + #ifndef STBI_ONLY_PIC + #define STBI_NO_PIC + #endif + #ifndef STBI_ONLY_PNM + #define STBI_NO_PNM + #endif +#endif + +#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB) +#define STBI_NO_ZLIB +#endif + + +#include +#include // ptrdiff_t on osx +#include +#include +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include // ldexp, pow +#endif + +#ifndef STBI_NO_STDIO +#include +#endif + +#ifndef STBI_ASSERT +#include +#define STBI_ASSERT(x) assert(x) +#endif + +#ifdef __cplusplus +#define STBI_EXTERN extern "C" +#else +#define STBI_EXTERN extern +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + +#ifndef STBI_NO_THREAD_LOCALS + #if defined(__cplusplus) && __cplusplus >= 201103L + #define STBI_THREAD_LOCAL thread_local + #elif defined(__GNUC__) && __GNUC__ < 5 + #define STBI_THREAD_LOCAL __thread + #elif defined(_MSC_VER) + #define STBI_THREAD_LOCAL __declspec(thread) + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__) + #define STBI_THREAD_LOCAL _Thread_local + #endif + + #ifndef STBI_THREAD_LOCAL + #if defined(__GNUC__) + #define STBI_THREAD_LOCAL __thread + #endif + #endif +#endif + +#if defined(_MSC_VER) || defined(__SYMBIAN32__) +typedef unsigned short stbi__uint16; +typedef signed short stbi__int16; +typedef unsigned int stbi__uint32; +typedef signed int stbi__int32; +#else +#include +typedef uint16_t stbi__uint16; +typedef int16_t stbi__int16; +typedef uint32_t stbi__uint32; +typedef int32_t stbi__int32; +#endif + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; + +#ifdef _MSC_VER +#define STBI_NOTUSED(v) (void)(v) +#else +#define STBI_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (-(y) & 31))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// which in turn means it gets to use SSE2 everywhere. This is unfortunate, +// but previous attempts to provide the SSE2 functions with runtime +// detection caused numerous issues. The way architecture extensions are +// exposed in GCC/Clang is, sadly, not really suited for one-file libs. +// New behavior: if compiled with -msse2, we use SSE2 without any +// detection; if not, we don't use it at all. +#define STBI_NO_SIMD +#endif + +#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD) +// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET +// +// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the +// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant. +// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not +// simultaneously enabling "-mstackrealign". +// +// See https://github.com/nothings/stb/issues/81 for more information. +// +// So default to no SSE2 on 32-bit MinGW. If you've read this far and added +// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2. +#define STBI_NO_SIMD +#endif + +#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) +#define STBI_SSE2 +#include + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include // __cpuid +static int stbi__cpuid3(void) +{ + int info[4]; + __cpuid(info,1); + return info[3]; +} +#else +static int stbi__cpuid3(void) +{ + int res; + __asm { + mov eax,1 + cpuid + mov res,edx + } + return res; +} +#endif + +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#endif + +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + // If we're even attempting to compile this on GCC/Clang, that means + // -msse2 is on, which means the compiler is allowed to use SSE2 + // instructions at will, and so are we. + return 1; +} +#endif + +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +#ifdef _MSC_VER +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name +#else +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#ifndef STBI_MAX_DIMENSIONS +#define STBI_MAX_DIMENSIONS (1 << 24) +#endif + +/////////////////////////////////////////////// +// +// stbi__context struct and start_xxx functions + +// stbi__context structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + stbi__uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + stbi_uc buffer_start[128]; + int callback_already_read; + + stbi_uc *img_buffer, *img_buffer_end; + stbi_uc *img_buffer_original, *img_buffer_original_end; +} stbi__context; + + +static void stbi__refill_buffer(stbi__context *s); + +// initialize a memory-decode context +static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; +} + +// initialize a callback-based context +static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = s->buffer_start; + stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; +} + +#ifndef STBI_NO_STDIO + +static int stbi__stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stbi__stdio_skip(void *user, int n) +{ + int ch; + fseek((FILE*) user, n, SEEK_CUR); + ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */ + if (ch != EOF) { + ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */ + } +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user) || ferror((FILE *) user); +} + +static stbi_io_callbacks stbi__stdio_callbacks = +{ + stbi__stdio_read, + stbi__stdio_skip, + stbi__stdio_eof, +}; + +static void stbi__start_file(stbi__context *s, FILE *f) +{ + stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi__context *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi__rewind(stbi__context *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; + s->img_buffer_end = s->img_buffer_original_end; +} + +enum +{ + STBI_ORDER_RGB, + STBI_ORDER_BGR +}; + +typedef struct +{ + int bits_per_channel; + int num_channels; + int channel_order; +} stbi__result_info; + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNG +static int stbi__png_test(stbi__context *s); +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__png_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_TGA +static int stbi__tga_test(stbi__context *s); +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s); +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__psd_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_HDR +static int stbi__hdr_test(stbi__context *s); +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_test(stbi__context *s); +static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_GIF +static int stbi__gif_test(stbi__context *s); +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNM +static int stbi__pnm_test(stbi__context *s); +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__pnm_is16(stbi__context *s); +#endif + +static +#ifdef STBI_THREAD_LOCAL +STBI_THREAD_LOCAL +#endif +const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +#ifndef STBI_NO_FAILURE_STRINGS +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} +#endif + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stb_image uses ints pervasively, including for offset calculations. +// therefore the largest decoded image size we can support with the +// current code, even on 64-bit targets, is INT_MAX. this is not a +// significant limitation for the intended use case. +// +// we do, however, need to make sure our size calculations don't +// overflow. hence a few helper functions for size calculations that +// multiply integers together, making sure that they're non-negative +// and no overflow occurs. + +// return 1 if the sum is valid, 0 on overflow. +// negative terms are considered invalid. +static int stbi__addsizes_valid(int a, int b) +{ + if (b < 0) return 0; + // now 0 <= b <= INT_MAX, hence also + // 0 <= INT_MAX - b <= INTMAX. + // And "a + b <= INT_MAX" (which might overflow) is the + // same as a <= INT_MAX - b (no overflow) + return a <= INT_MAX - b; +} + +// returns 1 if the product is valid, 0 on overflow. +// negative factors are considered invalid. +static int stbi__mul2sizes_valid(int a, int b) +{ + if (a < 0 || b < 0) return 0; + if (b == 0) return 1; // mul-by-0 is always safe + // portable way to check for no overflows in a*b + return a <= INT_MAX/b; +} + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow +static int stbi__mad2sizes_valid(int a, int b, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add); +} +#endif + +// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow +static int stbi__mad3sizes_valid(int a, int b, int c, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__addsizes_valid(a*b*c, add); +} + +// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add); +} +#endif + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// mallocs with size overflow checking +static void *stbi__malloc_mad2(int a, int b, int add) +{ + if (!stbi__mad2sizes_valid(a, b, add)) return NULL; + return stbi__malloc(a*b + add); +} +#endif + +static void *stbi__malloc_mad3(int a, int b, int c, int add) +{ + if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL; + return stbi__malloc(a*b*c + add); +} + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static void *stbi__malloc_mad4(int a, int b, int c, int d, int add) +{ + if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL; + return stbi__malloc(a*b*c*d + add); +} +#endif + +// returns 1 if the sum of two signed ints is valid (between -2^31 and 2^31-1 inclusive), 0 on overflow. +static int stbi__addints_valid(int a, int b) +{ + if ((a >= 0) != (b >= 0)) return 1; // a and b have different signs, so no overflow + if (a < 0 && b < 0) return a >= INT_MIN - b; // same as a + b >= INT_MIN; INT_MIN - b cannot overflow since b < 0. + return a <= INT_MAX - b; +} + +// returns 1 if the product of two ints fits in a signed short, 0 on overflow. +static int stbi__mul2shorts_valid(int a, int b) +{ + if (b == 0 || b == -1) return 1; // multiplication by 0 is always 0; check for -1 so SHRT_MIN/b doesn't overflow + if ((a >= 0) == (b >= 0)) return a <= SHRT_MAX/b; // product is positive, so similar to mul2sizes_valid + if (b < 0) return a <= SHRT_MIN / b; // same as a * b >= SHRT_MIN + return a >= SHRT_MIN / b; +} + +// stbi__err - error +// stbi__errpf - error returning pointer to float +// stbi__errpuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define stbi__err(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define stbi__err(x,y) stbi__err(y) +#else + #define stbi__err(x,y) stbi__err(x) +#endif + +#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL)) +#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL)) + +STBIDEF void stbi_image_free(void *retval_from_stbi_load) +{ + STBI_FREE(retval_from_stbi_load); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_HDR +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static int stbi__vertically_flip_on_load_global = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_global = flag_true_if_should_flip; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__vertically_flip_on_load stbi__vertically_flip_on_load_global +#else +static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set; + +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_local = flag_true_if_should_flip; + stbi__vertically_flip_on_load_set = 1; +} + +#define stbi__vertically_flip_on_load (stbi__vertically_flip_on_load_set \ + ? stbi__vertically_flip_on_load_local \ + : stbi__vertically_flip_on_load_global) +#endif // STBI_THREAD_LOCAL + +static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields + ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed + ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order + ri->num_channels = 0; + + // test the formats with a very explicit header first (at least a FOURCC + // or distinctive magic number first) + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc); + #else + STBI_NOTUSED(bpc); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri); + #endif + + // then the formats that can end up attempting to load with just 1 or 2 + // bytes matching expectations; these are prone to false positives, so + // try them later + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri); + return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + #ifndef STBI_NO_TGA + // test tga last because it's a crappy test! + if (stbi__tga_test(s)) + return stbi__tga_load(s,x,y,comp,req_comp, ri); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi_uc *reduced; + + reduced = (stbi_uc *) stbi__malloc(img_len); + if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling + + STBI_FREE(orig); + return reduced; +} + +static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi__uint16 *enlarged; + + enlarged = (stbi__uint16 *) stbi__malloc(img_len*2); + if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff + + STBI_FREE(orig); + return enlarged; +} + +static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel) +{ + int row; + size_t bytes_per_row = (size_t)w * bytes_per_pixel; + stbi_uc temp[2048]; + stbi_uc *bytes = (stbi_uc *)image; + + for (row = 0; row < (h>>1); row++) { + stbi_uc *row0 = bytes + row*bytes_per_row; + stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row; + // swap row0 with row1 + size_t bytes_left = bytes_per_row; + while (bytes_left) { + size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp); + memcpy(temp, row0, bytes_copy); + memcpy(row0, row1, bytes_copy); + memcpy(row1, temp, bytes_copy); + row0 += bytes_copy; + row1 += bytes_copy; + bytes_left -= bytes_copy; + } + } +} + +#ifndef STBI_NO_GIF +static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel) +{ + int slice; + int slice_size = w * h * bytes_per_pixel; + + stbi_uc *bytes = (stbi_uc *)image; + for (slice = 0; slice < z; ++slice) { + stbi__vertical_flip(bytes, w, h, bytes_per_pixel); + bytes += slice_size; + } +} +#endif + +static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 8) { + result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 8; + } + + // @TODO: move stbi__convert_format to here + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc)); + } + + return (unsigned char *) result; +} + +static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 16) { + result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 16; + } + + // @TODO: move stbi__convert_format16 to here + // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16)); + } + + return (stbi__uint16 *) result; +} + +#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR) +static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp) +{ + if (stbi__vertically_flip_on_load && result != NULL) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(float)); + } +} +#endif + +#ifndef STBI_NO_STDIO + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); +#endif + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + + +STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + unsigned char *result; + if (!f) return stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__uint16 *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + stbi__uint16 *result; + if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file_16(f,x,y,comp,req_comp); + fclose(f); + return result; +} + + +#endif //!STBI_NO_STDIO + +STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_mem(&s,buffer,len); + + result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp); + if (stbi__vertically_flip_on_load) { + stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); + } + + return result; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + stbi__result_info ri; + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp); + if (data) + return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return stbi__errpf("unknown image type", "Image not of any known type, or corrupt"); +} + +STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + float *result; + FILE *f = stbi__fopen(filename, "rb"); + if (!f) return stbi__errpf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_file(&s,f); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_LINEAR + +// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is +// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always +// reports false! + +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +STBIDEF int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + long pos = ftell(f); + int res; + stbi__context s; + stbi__start_file(&s,f); + res = stbi__hdr_test(&s); + fseek(f, pos, SEEK_SET); + return res; + #else + STBI_NOTUSED(f); + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(clbk); + STBI_NOTUSED(user); + return 0; + #endif +} + +#ifndef STBI_NO_LINEAR +static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; + +STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } +STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +#endif + +static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; + +STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } +STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header +}; + +static void stbi__refill_buffer(stbi__context *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original); + if (n == 0) { + // at end of file, treat same as if from memory, but need to handle case + // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file + s->read_from_callbacks = 0; + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start+1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static stbi_uc stbi__get8(stbi__context *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + stbi__refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +stbi_inline static int stbi__at_eof(stbi__context *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) +// nothing +#else +static void stbi__skip(stbi__context *s, int n) +{ + if (n == 0) return; // already there! + if (n < 0) { + s->img_buffer = s->img_buffer_end; + return; + } + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM) +// nothing +#else +static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} +#endif + +#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) +// nothing +#else +static int stbi__get16le(stbi__context *s) +{ + int z = stbi__get8(s); + return z + (stbi__get8(s) << 8); +} +#endif + +#ifndef STBI_NO_BMP +static stbi__uint32 stbi__get32le(stbi__context *s) +{ + stbi__uint32 z = stbi__get16le(s); + z += (stbi__uint32)stbi__get16le(s) << 16; + return z; +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static stbi_uc stbi__compute_y(int r, int g, int b) +{ + return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0); + if (good == NULL) { + STBI_FREE(data); + return stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 stbi__compute_y_16(int r, int g, int b) +{ + return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + stbi__uint16 *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2); + if (good == NULL) { + STBI_FREE(data); + return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + stbi__uint16 *src = data + j * x * img_n ; + stbi__uint16 *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output; + if (!data) return NULL; + output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale); + } + } + if (n < comp) { + for (i=0; i < x*y; ++i) { + output[i*comp + n] = data[i*comp + n]/255.0f; + } + } + STBI_FREE(data); + return output; +} +#endif + +#ifndef STBI_NO_HDR +#define stbi__float2int(x) ((int) (x)) +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output; + if (!data) return NULL; + output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + } + STBI_FREE(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder +// +// simple implementation +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - some SIMD kernels for common paths on targets with SSE2/NEON +// - uses a lot of intermediate memory, could cache poorly + +#ifndef STBI_NO_JPEG + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + stbi_uc fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + stbi__uint16 code[256]; + stbi_uc values[256]; + stbi_uc size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} stbi__huffman; + +typedef struct +{ + stbi__context *s; + stbi__huffman huff_dc[4]; + stbi__huffman huff_ac[4]; + stbi__uint16 dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + stbi_uc *data; + void *raw_data, *raw_coeff; + stbi_uc *linebuf; + short *coeff; // progressive only + int coeff_w, coeff_h; // number of 8x8 coefficient blocks + } img_comp[4]; + + stbi__uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int progressive; + int spec_start; + int spec_end; + int succ_high; + int succ_low; + int eob_run; + int jfif; + int app14_color_transform; // Adobe APP14 tag + int rgb; + + int scan_n, order[4]; + int restart_interval, todo; + +// kernels + void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]); + void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step); + stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs); +} stbi__jpeg; + +static int stbi__build_huffman(stbi__huffman *h, int *count) +{ + int i,j,k=0; + unsigned int code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) { + for (j=0; j < count[i]; ++j) { + h->size[k++] = (stbi_uc) (i+1); + if(k >= 257) return stbi__err("bad size list","Corrupt JPEG"); + } + } + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (stbi__uint16) (code++); + if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (stbi_uc) i; + } + } + } + return 1; +} + +// build a table that decodes both magnitude and value of small ACs in +// one go. +static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h) +{ + int i; + for (i=0; i < (1 << FAST_BITS); ++i) { + stbi_uc fast = h->fast[i]; + fast_ac[i] = 0; + if (fast < 255) { + int rs = h->values[fast]; + int run = (rs >> 4) & 15; + int magbits = rs & 15; + int len = h->size[fast]; + + if (magbits && len + magbits <= FAST_BITS) { + // magnitude code followed by receive_extend code + int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits); + int m = 1 << (magbits - 1); + if (k < m) k += (~0U << magbits) + 1; + // if the result is small enough, we can fit it in fast_ac table + if (k >= -128 && k <= 127) + fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + unsigned int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; + if(c < 0 || c >= 256) // symbol id out of bounds! + return -1; + STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + + sgn = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative) + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & (sgn - 1)); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + if (j->code_bits < 1) return 0; // ran out of bits from stream, return 0s intead of continuing + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static const stbi_uc stbi__jpeg_dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant) +{ + int diff,dc,k; + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? stbi__extend_receive(j, t) : 0; + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta","Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, dequant[0])) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * dequant[0]); + + // decode AC components, see JPEG spec + k = 1; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * dequant[zig]); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]); + } + } + } while (k < 64); + return 1; +} + +static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b) +{ + int diff,dc; + int t; + if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + if (j->succ_high == 0) { + // first scan for DC coefficient, must be first + memset(data,0,64*sizeof(data[0])); // 0 all the ac values now + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + diff = t ? stbi__extend_receive(j, t) : 0; + + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta", "Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, 1 << j->succ_low)) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * (1 << j->succ_low)); + } else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short) (1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * (1 << shift)); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * (1 << shift)); + } + } + } while (k <= j->spec_end); + } else { + // refinement scan for these AC coefficients + + short bit = (short) (1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short *p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } else { + k = j->spec_start; + do { + int r,s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short *p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } else { + if (r == 0) { + *p = (short) s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc) x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) * 4096) + +// derived from jidctint -- DCT_ISLOW +#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * stbi__f2f(0.5411961f); \ + t2 = p1 + p3*stbi__f2f(-1.847759065f); \ + t3 = p1 + p2*stbi__f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*stbi__f2f( 1.175875602f); \ + t0 = t0*stbi__f2f( 0.298631336f); \ + t1 = t1*stbi__f2f( 2.053119869f); \ + t2 = t2*stbi__f2f( 3.072711026f); \ + t3 = t3*stbi__f2f( 1.501321110f); \ + p1 = p5 + p1*stbi__f2f(-0.899976223f); \ + p2 = p5 + p2*stbi__f2f(-2.562915447f); \ + p3 = p3*stbi__f2f(-1.961570560f); \ + p4 = p4*stbi__f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) +{ + int i,val[64],*v=val; + stbi_uc *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0]*4; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = stbi__clamp((x0+t3) >> 17); + o[7] = stbi__clamp((x0-t3) >> 17); + o[1] = stbi__clamp((x1+t2) >> 17); + o[6] = stbi__clamp((x1-t2) >> 17); + o[2] = stbi__clamp((x2+t1) >> 17); + o[5] = stbi__clamp((x2-t1) >> 17); + o[3] = stbi__clamp((x3+t0) >> 17); + o[4] = stbi__clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SSE2 +// sse2 integer IDCT. not the fastest possible implementation but it +// produces bit-identical results to the generic C version so it's +// fully "transparent". +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + // This is constructed to match our regular (generic) integer IDCT exactly. + __m128i row0, row1, row2, row3, row4, row5, row6, row7; + __m128i tmp; + + // dot product constant: even elems=x, odd elems=y + #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y)) + + // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit) + // out(1) = c1[even]*x + c1[odd]*y + #define dct_rot(out0,out1, x,y,c0,c1) \ + __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \ + __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \ + __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \ + __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \ + __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \ + __m128i out1##_h = _mm_madd_epi16(c0##hi, c1) + + // out = in << 12 (in 16-bit, out 32-bit) + #define dct_widen(out, in) \ + __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \ + __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4) + + // wide add + #define dct_wadd(out, a, b) \ + __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_add_epi32(a##_h, b##_h) + + // wide sub + #define dct_wsub(out, a, b) \ + __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_sub_epi32(a##_h, b##_h) + + // butterfly a/b, add bias, then shift by "s" and pack + #define dct_bfly32o(out0, out1, a,b,bias,s) \ + { \ + __m128i abiased_l = _mm_add_epi32(a##_l, bias); \ + __m128i abiased_h = _mm_add_epi32(a##_h, bias); \ + dct_wadd(sum, abiased, b); \ + dct_wsub(dif, abiased, b); \ + out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \ + out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \ + } + + // 8-bit interleave step (for transposes) + #define dct_interleave8(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi8(a, b); \ + b = _mm_unpackhi_epi8(tmp, b) + + // 16-bit interleave step (for transposes) + #define dct_interleave16(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi16(a, b); \ + b = _mm_unpackhi_epi16(tmp, b) + + #define dct_pass(bias,shift) \ + { \ + /* even part */ \ + dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \ + __m128i sum04 = _mm_add_epi16(row0, row4); \ + __m128i dif04 = _mm_sub_epi16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \ + dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \ + __m128i sum17 = _mm_add_epi16(row1, row7); \ + __m128i sum35 = _mm_add_epi16(row3, row5); \ + dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \ + dct_wadd(x4, y0o, y4o); \ + dct_wadd(x5, y1o, y5o); \ + dct_wadd(x6, y2o, y5o); \ + dct_wadd(x7, y3o, y4o); \ + dct_bfly32o(row0,row7, x0,x7,bias,shift); \ + dct_bfly32o(row1,row6, x1,x6,bias,shift); \ + dct_bfly32o(row2,row5, x2,x5,bias,shift); \ + dct_bfly32o(row3,row4, x3,x4,bias,shift); \ + } + + __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f)); + __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f)); + __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f)); + __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f)); + __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f)); + __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f)); + __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f)); + __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f)); + + // rounding biases in column/row passes, see stbi__idct_block for explanation. + __m128i bias_0 = _mm_set1_epi32(512); + __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17)); + + // load + row0 = _mm_load_si128((const __m128i *) (data + 0*8)); + row1 = _mm_load_si128((const __m128i *) (data + 1*8)); + row2 = _mm_load_si128((const __m128i *) (data + 2*8)); + row3 = _mm_load_si128((const __m128i *) (data + 3*8)); + row4 = _mm_load_si128((const __m128i *) (data + 4*8)); + row5 = _mm_load_si128((const __m128i *) (data + 5*8)); + row6 = _mm_load_si128((const __m128i *) (data + 6*8)); + row7 = _mm_load_si128((const __m128i *) (data + 7*8)); + + // column pass + dct_pass(bias_0, 10); + + { + // 16bit 8x8 transpose pass 1 + dct_interleave16(row0, row4); + dct_interleave16(row1, row5); + dct_interleave16(row2, row6); + dct_interleave16(row3, row7); + + // transpose pass 2 + dct_interleave16(row0, row2); + dct_interleave16(row1, row3); + dct_interleave16(row4, row6); + dct_interleave16(row5, row7); + + // transpose pass 3 + dct_interleave16(row0, row1); + dct_interleave16(row2, row3); + dct_interleave16(row4, row5); + dct_interleave16(row6, row7); + } + + // row pass + dct_pass(bias_1, 17); + + { + // pack + __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7 + __m128i p1 = _mm_packus_epi16(row2, row3); + __m128i p2 = _mm_packus_epi16(row4, row5); + __m128i p3 = _mm_packus_epi16(row6, row7); + + // 8bit 8x8 transpose pass 1 + dct_interleave8(p0, p2); // a0e0a1e1... + dct_interleave8(p1, p3); // c0g0c1g1... + + // transpose pass 2 + dct_interleave8(p0, p1); // a0c0e0g0... + dct_interleave8(p2, p3); // b0d0f0h0... + + // transpose pass 3 + dct_interleave8(p0, p2); // a0b0c0d0... + dct_interleave8(p1, p3); // a4b4c4d4... + + // store + _mm_storel_epi64((__m128i *) out, p0); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p2); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p1); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p3); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e)); + } + +#undef dct_const +#undef dct_rot +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_interleave8 +#undef dct_interleave16 +#undef dct_pass +} + +#endif // STBI_SSE2 + +#ifdef STBI_NEON + +// NEON integer IDCT. should produce bit-identical +// results to the generic C version. +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + int16x8_t row0, row1, row2, row3, row4, row5, row6, row7; + + int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f)); + int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f)); + int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f)); + int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f)); + int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f)); + int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f)); + int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f)); + int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f)); + int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f)); + int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f)); + int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f)); + int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f)); + +#define dct_long_mul(out, inq, coeff) \ + int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff) + +#define dct_long_mac(out, acc, inq, coeff) \ + int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff) + +#define dct_widen(out, inq) \ + int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \ + int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12) + +// wide add +#define dct_wadd(out, a, b) \ + int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vaddq_s32(a##_h, b##_h) + +// wide sub +#define dct_wsub(out, a, b) \ + int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vsubq_s32(a##_h, b##_h) + +// butterfly a/b, then shift using "shiftop" by "s" and pack +#define dct_bfly32o(out0,out1, a,b,shiftop,s) \ + { \ + dct_wadd(sum, a, b); \ + dct_wsub(dif, a, b); \ + out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \ + out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \ + } + +#define dct_pass(shiftop, shift) \ + { \ + /* even part */ \ + int16x8_t sum26 = vaddq_s16(row2, row6); \ + dct_long_mul(p1e, sum26, rot0_0); \ + dct_long_mac(t2e, p1e, row6, rot0_1); \ + dct_long_mac(t3e, p1e, row2, rot0_2); \ + int16x8_t sum04 = vaddq_s16(row0, row4); \ + int16x8_t dif04 = vsubq_s16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + int16x8_t sum15 = vaddq_s16(row1, row5); \ + int16x8_t sum17 = vaddq_s16(row1, row7); \ + int16x8_t sum35 = vaddq_s16(row3, row5); \ + int16x8_t sum37 = vaddq_s16(row3, row7); \ + int16x8_t sumodd = vaddq_s16(sum17, sum35); \ + dct_long_mul(p5o, sumodd, rot1_0); \ + dct_long_mac(p1o, p5o, sum17, rot1_1); \ + dct_long_mac(p2o, p5o, sum35, rot1_2); \ + dct_long_mul(p3o, sum37, rot2_0); \ + dct_long_mul(p4o, sum15, rot2_1); \ + dct_wadd(sump13o, p1o, p3o); \ + dct_wadd(sump24o, p2o, p4o); \ + dct_wadd(sump23o, p2o, p3o); \ + dct_wadd(sump14o, p1o, p4o); \ + dct_long_mac(x4, sump13o, row7, rot3_0); \ + dct_long_mac(x5, sump24o, row5, rot3_1); \ + dct_long_mac(x6, sump23o, row3, rot3_2); \ + dct_long_mac(x7, sump14o, row1, rot3_3); \ + dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \ + dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \ + dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \ + dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \ + } + + // load + row0 = vld1q_s16(data + 0*8); + row1 = vld1q_s16(data + 1*8); + row2 = vld1q_s16(data + 2*8); + row3 = vld1q_s16(data + 3*8); + row4 = vld1q_s16(data + 4*8); + row5 = vld1q_s16(data + 5*8); + row6 = vld1q_s16(data + 6*8); + row7 = vld1q_s16(data + 7*8); + + // add DC bias + row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0)); + + // column pass + dct_pass(vrshrn_n_s32, 10); + + // 16bit 8x8 transpose + { +// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively. +// whether compilers actually get this is another story, sadly. +#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); } +#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); } + + // pass 1 + dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6 + dct_trn16(row2, row3); + dct_trn16(row4, row5); + dct_trn16(row6, row7); + + // pass 2 + dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4 + dct_trn32(row1, row3); + dct_trn32(row4, row6); + dct_trn32(row5, row7); + + // pass 3 + dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0 + dct_trn64(row1, row5); + dct_trn64(row2, row6); + dct_trn64(row3, row7); + +#undef dct_trn16 +#undef dct_trn32 +#undef dct_trn64 + } + + // row pass + // vrshrn_n_s32 only supports shifts up to 16, we need + // 17. so do a non-rounding shift of 16 first then follow + // up with a rounding shift by 1. + dct_pass(vshrn_n_s32, 16); + + { + // pack and round + uint8x8_t p0 = vqrshrun_n_s16(row0, 1); + uint8x8_t p1 = vqrshrun_n_s16(row1, 1); + uint8x8_t p2 = vqrshrun_n_s16(row2, 1); + uint8x8_t p3 = vqrshrun_n_s16(row3, 1); + uint8x8_t p4 = vqrshrun_n_s16(row4, 1); + uint8x8_t p5 = vqrshrun_n_s16(row5, 1); + uint8x8_t p6 = vqrshrun_n_s16(row6, 1); + uint8x8_t p7 = vqrshrun_n_s16(row7, 1); + + // again, these can translate into one instruction, but often don't. +#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); } +#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); } + + // sadly can't use interleaved stores here since we only write + // 8 bytes to each scan line! + + // 8x8 8-bit transpose pass 1 + dct_trn8_8(p0, p1); + dct_trn8_8(p2, p3); + dct_trn8_8(p4, p5); + dct_trn8_8(p6, p7); + + // pass 2 + dct_trn8_16(p0, p2); + dct_trn8_16(p1, p3); + dct_trn8_16(p4, p6); + dct_trn8_16(p5, p7); + + // pass 3 + dct_trn8_32(p0, p4); + dct_trn8_32(p1, p5); + dct_trn8_32(p2, p6); + dct_trn8_32(p3, p7); + + // store + vst1_u8(out, p0); out += out_stride; + vst1_u8(out, p1); out += out_stride; + vst1_u8(out, p2); out += out_stride; + vst1_u8(out, p3); out += out_stride; + vst1_u8(out, p4); out += out_stride; + vst1_u8(out, p5); out += out_stride; + vst1_u8(out, p6); out += out_stride; + vst1_u8(out, p7); + +#undef dct_trn8_8 +#undef dct_trn8_16 +#undef dct_trn8_32 + } + +#undef dct_long_mul +#undef dct_long_mac +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_pass +} + +#endif // STBI_NEON + +#define STBI__MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static stbi_uc stbi__get_marker(stbi__jpeg *j) +{ + stbi_uc x; + if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; } + x = stbi__get8(j->s); + if (x != 0xff) return STBI__MARKER_none; + while (x == 0xff) + x = stbi__get8(j->s); // consume repeated 0xff fill bytes + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, stbi__jpeg_reset the entropy decoder and +// the dc prediction +static void stbi__jpeg_reset(stbi__jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0; + j->marker = STBI__MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + j->eob_run = 0; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int stbi__parse_entropy_coded_data(stbi__jpeg *z) +{ + stbi__jpeg_reset(z); + if (!z->progressive) { + if (z->scan_n == 1) { + int i,j; + STBI_SIMD_ALIGN(short, data[64]); + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + STBI_SIMD_ALIGN(short, data[64]); + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } else { + if (z->scan_n == 1) { + int i,j; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + if (z->spec_start == 0) { + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } else { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha])) + return 0; + } + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x); + int y2 = (j*z->img_comp[n].v + y); + short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w); + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant) +{ + int i; + for (i=0; i < 64; ++i) + data[i] *= dequant[i]; +} + +static void stbi__jpeg_finish(stbi__jpeg *z) +{ + if (z->progressive) { + // dequantize and idct the data + int i,j,n; + for (n=0; n < z->s->img_n; ++n) { + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]); + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + } + } + } + } +} + +static int stbi__process_marker(stbi__jpeg *z, int m) +{ + int L; + switch (m) { + case STBI__MARKER_none: // no marker found + return stbi__err("expected marker","Corrupt JPEG"); + + case 0xDD: // DRI - specify restart interval + if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG"); + z->restart_interval = stbi__get16be(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = stbi__get16be(z->s)-2; + while (L > 0) { + int q = stbi__get8(z->s); + int p = q >> 4, sixteen = (p != 0); + int t = q & 15,i; + if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG"); + if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG"); + + for (i=0; i < 64; ++i) + z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s)); + L -= (sixteen ? 129 : 65); + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = stbi__get16be(z->s)-2; + while (L > 0) { + stbi_uc *v; + int sizes[16],i,n=0; + int q = stbi__get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = stbi__get8(z->s); + n += sizes[i]; + } + if(n > 256) return stbi__err("bad DHT header","Corrupt JPEG"); // Loop over i < n would write past end of values! + L -= 17; + if (tc == 0) { + if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < n; ++i) + v[i] = stbi__get8(z->s); + if (tc != 0) + stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th); + L -= n; + } + return L==0; + } + + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + L = stbi__get16be(z->s); + if (L < 2) { + if (m == 0xFE) + return stbi__err("bad COM len","Corrupt JPEG"); + else + return stbi__err("bad APP len","Corrupt JPEG"); + } + L -= 2; + + if (m == 0xE0 && L >= 5) { // JFIF APP0 segment + static const unsigned char tag[5] = {'J','F','I','F','\0'}; + int ok = 1; + int i; + for (i=0; i < 5; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 5; + if (ok) + z->jfif = 1; + } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment + static const unsigned char tag[6] = {'A','d','o','b','e','\0'}; + int ok = 1; + int i; + for (i=0; i < 6; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 6; + if (ok) { + stbi__get8(z->s); // version + stbi__get16be(z->s); // flags0 + stbi__get16be(z->s); // flags1 + z->app14_color_transform = stbi__get8(z->s); // color transform + L -= 6; + } + } + + stbi__skip(z->s, L); + return 1; + } + + return stbi__err("unknown marker","Corrupt JPEG"); +} + +// after we see SOS +static int stbi__process_scan_header(stbi__jpeg *z) +{ + int i; + int Ls = stbi__get16be(z->s); + z->scan_n = stbi__get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = stbi__get8(z->s), which; + int q = stbi__get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; // no match + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + + { + int aa; + z->spec_start = stbi__get8(z->s); + z->spec_end = stbi__get8(z->s); // should be 63, but might be 0 + aa = stbi__get8(z->s); + z->succ_high = (aa >> 4); + z->succ_low = (aa & 15); + if (z->progressive) { + if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13) + return stbi__err("bad SOS", "Corrupt JPEG"); + } else { + if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG"); + if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG"); + z->spec_end = 63; + } + } + + return 1; +} + +static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why) +{ + int i; + for (i=0; i < ncomp; ++i) { + if (z->img_comp[i].raw_data) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + z->img_comp[i].data = NULL; + } + if (z->img_comp[i].raw_coeff) { + STBI_FREE(z->img_comp[i].raw_coeff); + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].coeff = 0; + } + if (z->img_comp[i].linebuf) { + STBI_FREE(z->img_comp[i].linebuf); + z->img_comp[i].linebuf = NULL; + } + } + return why; +} + +static int stbi__process_frame_header(stbi__jpeg *z, int scan) +{ + stbi__context *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG + p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + c = stbi__get8(s); + if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG"); + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); + + z->rgb = 0; + for (i=0; i < s->img_n; ++i) { + static const unsigned char rgb[3] = { 'R', 'G', 'B' }; + z->img_comp[i].id = stbi__get8(s); + if (s->img_n == 3 && z->img_comp[i].id == rgb[i]) + ++z->rgb; + q = stbi__get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG"); + z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); + } + + if (scan != STBI__SCAN_load) return 1; + + if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios + // and I've never seen a non-corrupted JPEG file actually use them + for (i=0; i < s->img_n; ++i) { + if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG"); + if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG"); + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + // these sizes can't be more than 17 bits + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + // + // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier) + // so these muls can't overflow with 32-bit ints (which we require) + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].linebuf = NULL; + z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15); + if (z->img_comp[i].raw_data == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + // align blocks for idct using mmx/sse + z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + if (z->progressive) { + // w2, h2 are multiples of 8 (see above) + z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8; + z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8; + z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15); + if (z->img_comp[i].raw_coeff == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define stbi__DNL(x) ((x) == 0xdc) +#define stbi__SOI(x) ((x) == 0xd8) +#define stbi__EOI(x) ((x) == 0xd9) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) +#define stbi__SOS(x) ((x) == 0xda) + +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) +{ + int m; + z->jfif = 0; + z->app14_color_transform = -1; // valid values are 0,1,2 + z->marker = STBI__MARKER_none; // initialize cached marker to empty + m = stbi__get_marker(z); + if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG"); + if (scan == STBI__SCAN_type) return 1; + m = stbi__get_marker(z); + while (!stbi__SOF(m)) { + if (!stbi__process_marker(z,m)) return 0; + m = stbi__get_marker(z); + while (m == STBI__MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG"); + m = stbi__get_marker(z); + } + } + z->progressive = stbi__SOF_progressive(m); + if (!stbi__process_frame_header(z, scan)) return 0; + return 1; +} + +static stbi_uc stbi__skip_jpeg_junk_at_end(stbi__jpeg *j) +{ + // some JPEGs have junk at end, skip over it but if we find what looks + // like a valid marker, resume there + while (!stbi__at_eof(j->s)) { + stbi_uc x = stbi__get8(j->s); + while (x == 0xff) { // might be a marker + if (stbi__at_eof(j->s)) return STBI__MARKER_none; + x = stbi__get8(j->s); + if (x != 0x00 && x != 0xff) { + // not a stuffed zero or lead-in to another marker, looks + // like an actual marker, return it + return x; + } + // stuffed zero has x=0 now which ends the loop, meaning we go + // back to regular scan loop. + // repeated 0xff keeps trying to read the next byte of the marker. + } + } + return STBI__MARKER_none; +} + +// decode image to YCbCr format +static int stbi__decode_jpeg_image(stbi__jpeg *j) +{ + int m; + for (m = 0; m < 4; m++) { + j->img_comp[m].raw_data = NULL; + j->img_comp[m].raw_coeff = NULL; + } + j->restart_interval = 0; + if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0; + m = stbi__get_marker(j); + while (!stbi__EOI(m)) { + if (stbi__SOS(m)) { + if (!stbi__process_scan_header(j)) return 0; + if (!stbi__parse_entropy_coded_data(j)) return 0; + if (j->marker == STBI__MARKER_none ) { + j->marker = stbi__skip_jpeg_junk_at_end(j); + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + m = stbi__get_marker(j); + if (STBI__RESTART(m)) + m = stbi__get_marker(j); + } else if (stbi__DNL(m)) { + int Ld = stbi__get16be(j->s); + stbi__uint32 NL = stbi__get16be(j->s); + if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG"); + if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG"); + m = stbi__get_marker(j); + } else { + if (!stbi__process_marker(j, m)) return 1; + m = stbi__get_marker(j); + } + } + if (j->progressive) + stbi__jpeg_finish(j); + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1, + int w, int hs); + +#define stbi__div4(x) ((stbi_uc) ((x) >> 2)) + +static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + stbi_uc *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = stbi__div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = stbi__div4(n+input[i-1]); + out[i*2+1] = stbi__div4(n+input[i+1]); + } + out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define stbi__div16(x) ((stbi_uc) ((x) >> 4)) + +static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = stbi__div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i=0,t0,t1; + + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + // process groups of 8 pixels for as long as we can. + // note we can't handle the last pixel in a row in this loop + // because we need to handle the filter boundary conditions. + for (; i < ((w-1) & ~7); i += 8) { +#if defined(STBI_SSE2) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + __m128i zero = _mm_setzero_si128(); + __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i)); + __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i)); + __m128i farw = _mm_unpacklo_epi8(farb, zero); + __m128i nearw = _mm_unpacklo_epi8(nearb, zero); + __m128i diff = _mm_sub_epi16(farw, nearw); + __m128i nears = _mm_slli_epi16(nearw, 2); + __m128i curr = _mm_add_epi16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + __m128i prv0 = _mm_slli_si128(curr, 2); + __m128i nxt0 = _mm_srli_si128(curr, 2); + __m128i prev = _mm_insert_epi16(prv0, t1, 0); + __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + __m128i bias = _mm_set1_epi16(8); + __m128i curs = _mm_slli_epi16(curr, 2); + __m128i prvd = _mm_sub_epi16(prev, curr); + __m128i nxtd = _mm_sub_epi16(next, curr); + __m128i curb = _mm_add_epi16(curs, bias); + __m128i even = _mm_add_epi16(prvd, curb); + __m128i odd = _mm_add_epi16(nxtd, curb); + + // interleave even and odd pixels, then undo scaling. + __m128i int0 = _mm_unpacklo_epi16(even, odd); + __m128i int1 = _mm_unpackhi_epi16(even, odd); + __m128i de0 = _mm_srli_epi16(int0, 4); + __m128i de1 = _mm_srli_epi16(int1, 4); + + // pack and write output + __m128i outv = _mm_packus_epi16(de0, de1); + _mm_storeu_si128((__m128i *) (out + i*2), outv); +#elif defined(STBI_NEON) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + uint8x8_t farb = vld1_u8(in_far + i); + uint8x8_t nearb = vld1_u8(in_near + i); + int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb)); + int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2)); + int16x8_t curr = vaddq_s16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + int16x8_t prv0 = vextq_s16(curr, curr, 7); + int16x8_t nxt0 = vextq_s16(curr, curr, 1); + int16x8_t prev = vsetq_lane_s16(t1, prv0, 0); + int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + int16x8_t curs = vshlq_n_s16(curr, 2); + int16x8_t prvd = vsubq_s16(prev, curr); + int16x8_t nxtd = vsubq_s16(next, curr); + int16x8_t even = vaddq_s16(curs, prvd); + int16x8_t odd = vaddq_s16(curs, nxtd); + + // undo scaling and round, then store with even/odd phases interleaved + uint8x8x2_t o; + o.val[0] = vqrshrun_n_s16(even, 4); + o.val[1] = vqrshrun_n_s16(odd, 4); + vst2_u8(out + i*2, o); +#endif + + // "previous" value for next iter + t1 = 3*in_near[i+7] + in_far[i+7]; + } + + t0 = t1; + t1 = 3*in_near[i] + in_far[i]; + out[i*2] = stbi__div16(3*t1 + t0 + 8); + + for (++i; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} +#endif + +static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +// this is a reduced-precision calculation of YCbCr-to-RGB introduced +// to make sure the code produces the same results in both SIMD and scalar +#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step) +{ + int i = 0; + +#ifdef STBI_SSE2 + // step == 3 is pretty ugly on the final interleave, and i'm not convinced + // it's useful in practice (you wouldn't use it for textures, for example). + // so just accelerate step == 4 case. + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + __m128i signflip = _mm_set1_epi8(-0x80); + __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f)); + __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f)); + __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f)); + __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f)); + __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128); + __m128i xw = _mm_set1_epi16(255); // alpha channel + + for (; i+7 < count; i += 8) { + // load + __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i)); + __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i)); + __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i)); + __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128 + __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128 + + // unpack to short (and left-shift cr, cb by 8) + __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes); + __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased); + __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased); + + // color transform + __m128i yws = _mm_srli_epi16(yw, 4); + __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw); + __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw); + __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1); + __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1); + __m128i rws = _mm_add_epi16(cr0, yws); + __m128i gwt = _mm_add_epi16(cb0, yws); + __m128i bws = _mm_add_epi16(yws, cb1); + __m128i gws = _mm_add_epi16(gwt, cr1); + + // descale + __m128i rw = _mm_srai_epi16(rws, 4); + __m128i bw = _mm_srai_epi16(bws, 4); + __m128i gw = _mm_srai_epi16(gws, 4); + + // back to byte, set up for transpose + __m128i brb = _mm_packus_epi16(rw, bw); + __m128i gxb = _mm_packus_epi16(gw, xw); + + // transpose to interleave channels + __m128i t0 = _mm_unpacklo_epi8(brb, gxb); + __m128i t1 = _mm_unpackhi_epi8(brb, gxb); + __m128i o0 = _mm_unpacklo_epi16(t0, t1); + __m128i o1 = _mm_unpackhi_epi16(t0, t1); + + // store + _mm_storeu_si128((__m128i *) (out + 0), o0); + _mm_storeu_si128((__m128i *) (out + 16), o1); + out += 32; + } + } +#endif + +#ifdef STBI_NEON + // in this version, step=3 support would be easy to add. but is there demand? + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + uint8x8_t signflip = vdup_n_u8(0x80); + int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f)); + int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f)); + int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f)); + int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f)); + + for (; i+7 < count; i += 8) { + // load + uint8x8_t y_bytes = vld1_u8(y + i); + uint8x8_t cr_bytes = vld1_u8(pcr + i); + uint8x8_t cb_bytes = vld1_u8(pcb + i); + int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip)); + int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip)); + + // expand to s16 + int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4)); + int16x8_t crw = vshll_n_s8(cr_biased, 7); + int16x8_t cbw = vshll_n_s8(cb_biased, 7); + + // color transform + int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0); + int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0); + int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1); + int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1); + int16x8_t rws = vaddq_s16(yws, cr0); + int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1); + int16x8_t bws = vaddq_s16(yws, cb1); + + // undo scaling, round, convert to byte + uint8x8x4_t o; + o.val[0] = vqrshrun_n_s16(rws, 4); + o.val[1] = vqrshrun_n_s16(gws, 4); + o.val[2] = vqrshrun_n_s16(bws, 4); + o.val[3] = vdup_n_u8(255); + + // store, interleaving r/g/b/a + vst4_u8(out, o); + out += 8*4; + } + } +#endif + + for (; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +// set up the kernels +static void stbi__setup_jpeg(stbi__jpeg *j) +{ + j->idct_block_kernel = stbi__idct_block; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2; + +#ifdef STBI_SSE2 + if (stbi__sse2_available()) { + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; +#endif +} + +// clean up the temporary component buffers +static void stbi__cleanup_jpeg(stbi__jpeg *j) +{ + stbi__free_jpeg_components(j, j->s->img_n, 0); +} + +typedef struct +{ + resample_row_func resample; + stbi_uc *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi__resample; + +// fast 0..255 * 0..255 => 0..255 rounded multiplication +static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y) +{ + unsigned int t = x*y + 128; + return (stbi_uc) ((t + (t >>8)) >> 8); +} + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n, is_rgb; + z->s->img_n = 0; // make stbi__cleanup_jpeg safe + + // validate req_comp + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source, but leave in YCbCr format + if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1; + + is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif)); + + if (z->s->img_n == 3 && n < 3 && !is_rgb) + decode_n = 1; + else + decode_n = z->s->img_n; + + // nothing to do if no components requested; check this now to avoid + // accessing uninitialized coutput[0] later + if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; } + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL }; + + stbi__resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel; + else r->resample = stbi__resample_row_generic; + } + + // can't error after this so, this is safe + output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1); + if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + stbi_uc *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + stbi_uc *y = coutput[0]; + if (z->s->img_n == 3) { + if (is_rgb) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = y[i]; + out[1] = coutput[1][i]; + out[2] = coutput[2][i]; + out[3] = 255; + out += n; + } + } else { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else if (z->s->img_n == 4) { + if (z->app14_color_transform == 0) { // CMYK + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(coutput[0][i], m); + out[1] = stbi__blinn_8x8(coutput[1][i], m); + out[2] = stbi__blinn_8x8(coutput[2][i], m); + out[3] = 255; + out += n; + } + } else if (z->app14_color_transform == 2) { // YCCK + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(255 - out[0], m); + out[1] = stbi__blinn_8x8(255 - out[1], m); + out[2] = stbi__blinn_8x8(255 - out[2], m); + out += n; + } + } else { // YCbCr + alpha? Ignore the fourth channel for now + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + if (is_rgb) { + if (n == 1) + for (i=0; i < z->s->img_x; ++i) + *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + else { + for (i=0; i < z->s->img_x; ++i, out += 2) { + out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + out[1] = 255; + } + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 0) { + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + stbi_uc r = stbi__blinn_8x8(coutput[0][i], m); + stbi_uc g = stbi__blinn_8x8(coutput[1][i], m); + stbi_uc b = stbi__blinn_8x8(coutput[2][i], m); + out[0] = stbi__compute_y(r, g, b); + out[1] = 255; + out += n; + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 2) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]); + out[1] = 255; + out += n; + } + } else { + stbi_uc *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; } + } + } + } + stbi__cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output + return output; + } +} + +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + unsigned char* result; + stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__errpuc("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + STBI_NOTUSED(ri); + j->s = s; + stbi__setup_jpeg(j); + result = load_jpeg_image(j, x,y,comp,req_comp); + STBI_FREE(j); + return result; +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + stbi__setup_jpeg(j); + r = stbi__decode_jpeg_header(j, STBI__SCAN_type); + stbi__rewind(s); + STBI_FREE(j); + return r; +} + +static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) +{ + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { + stbi__rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n >= 3 ? 3 : 1; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + int result; + stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg))); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + result = stbi__jpeg_info_raw(j, x, y, comp); + STBI_FREE(j); + return result; +} +#endif + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +#ifndef STBI_NO_ZLIB + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables +#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1) +#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[STBI__ZNSYMS]; + stbi__uint16 value[STBI__ZNSYMS]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16-bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 0, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + if (sizes[i] > (1 << i)) + return stbi__err("bad sizes", "Corrupt PNG"); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (stbi__uint16) code; + z->firstsymbol[i] = (stbi__uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i); + z->size [c] = (stbi_uc ) s; + z->value[c] = (stbi__uint16) i; + if (s <= STBI__ZFAST_BITS) { + int j = stbi__bit_reverse(next_code[s],s); + while (j < (1 << STBI__ZFAST_BITS)) { + z->fast[j] = fastv; + j += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + stbi_uc *zbuffer, *zbuffer_end; + int num_bits; + int hit_zeof_once; + stbi__uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + stbi__zhuffman z_length, z_distance; +} stbi__zbuf; + +stbi_inline static int stbi__zeof(stbi__zbuf *z) +{ + return (z->zbuffer >= z->zbuffer_end); +} + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + return stbi__zeof(z) ? 0 : *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + if (z->code_buffer >= (1U << z->num_bits)) { + z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */ + return; + } + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) stbi__fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s,k; + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = stbi__bit_reverse(a->code_buffer, 16); + for (s=STBI__ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s >= 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere! + if (z->size[b] != s) return -1; // was originally an assert, but report failure instead. + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) { + if (stbi__zeof(a)) { + if (!a->hit_zeof_once) { + // This is the first time we hit eof, insert 16 extra padding btis + // to allow us to keep going; if we actually consume any of them + // though, that is invalid data. This is caught later. + a->hit_zeof_once = 1; + a->num_bits += 16; // add 16 implicit zero bits + } else { + // We already inserted our extra 16 padding bits and are again + // out, this stream is actually prematurely terminated. + return -1; + } + } else { + stbi__fill_bits(a); + } + } + b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; + if (b) { + s = b >> 9; + a->code_buffer >>= s; + a->num_bits -= s; + return b & 511; + } + return stbi__zhuffman_decode_slowpath(a, z); +} + +static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes +{ + char *q; + unsigned int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (unsigned int) (z->zout - z->zout_start); + limit = old_limit = (unsigned) (z->zout_end - z->zout_start); + if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory"); + while (cur + n > limit) { + if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory"); + limit *= 2; + } + q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit); + STBI_NOTUSED(old_limit); + if (q == NULL) return stbi__err("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static const int stbi__zlength_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static const int stbi__zlength_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static const int stbi__zdist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int stbi__parse_huffman_block(stbi__zbuf *a) +{ + char *zout = a->zout; + for(;;) { + int z = stbi__zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= a->zout_end) { + if (!stbi__zexpand(a, zout, 1)) return 0; + zout = a->zout; + } + *zout++ = (char) z; + } else { + stbi_uc *p; + int len,dist; + if (z == 256) { + a->zout = zout; + if (a->hit_zeof_once && a->num_bits < 16) { + // The first time we hit zeof, we inserted 16 extra zero bits into our bit + // buffer so the decoder can just do its speculative decoding. But if we + // actually consumed any of those bits (which is the case when num_bits < 16), + // the stream actually read past the end so it is malformed. + return stbi__err("unexpected end","Corrupt PNG"); + } + return 1; + } + if (z >= 286) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, length codes 286 and 287 must not appear in compressed data + z -= 257; + len = stbi__zlength_base[z]; + if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); + z = stbi__zhuffman_decode(a, &a->z_distance); + if (z < 0 || z >= 30) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, distance codes 30 and 31 must not appear in compressed data + dist = stbi__zdist_base[z]; + if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (len > a->zout_end - zout) { + if (!stbi__zexpand(a, zout, len)) return 0; + zout = a->zout; + } + p = (stbi_uc *) (zout - dist); + if (dist == 1) { // run of one byte; common in images. + stbi_uc v = *p; + if (len) { do *zout++ = v; while (--len); } + } else { + if (len) { do *zout++ = *p++; while (--len); } + } + } + } +} + +static int stbi__compute_huffman_codes(stbi__zbuf *a) +{ + static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + stbi__zhuffman z_codelength; + stbi_uc lencodes[286+32+137];//padding for maximum single op + stbi_uc codelength_sizes[19]; + int i,n; + + int hlit = stbi__zreceive(a,5) + 257; + int hdist = stbi__zreceive(a,5) + 1; + int hclen = stbi__zreceive(a,4) + 4; + int ntot = hlit + hdist; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = stbi__zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (stbi_uc) s; + } + if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < ntot) { + int c = stbi__zhuffman_decode(a, &z_codelength); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); + if (c < 16) + lencodes[n++] = (stbi_uc) c; + else { + stbi_uc fill = 0; + if (c == 16) { + c = stbi__zreceive(a,2)+3; + if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG"); + fill = lencodes[n-1]; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + } else if (c == 18) { + c = stbi__zreceive(a,7)+11; + } else { + return stbi__err("bad codelengths", "Corrupt PNG"); + } + if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG"); + memset(lencodes+n, fill, c); + n += c; + } + } + if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG"); + if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int stbi__parse_uncompressed_block(stbi__zbuf *a) +{ + stbi_uc header[4]; + int len,nlen,k; + if (a->num_bits & 7) + stbi__zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check + a->code_buffer >>= 8; + a->num_bits -= 8; + } + if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG"); + // now fill header the normal way + while (k < 4) + header[k++] = stbi__zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!stbi__zexpand(a, a->zout, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int stbi__parse_zlib_header(stbi__zbuf *a) +{ + int cmf = stbi__zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = stbi__zget8(a); + if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] = +{ + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8 +}; +static const stbi_uc stbi__zdefault_distance[32] = +{ + 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5 +}; +/* +Init algorithm: +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8; + for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9; + for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7; + for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8; + + for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5; +} +*/ + +static int stbi__parse_zlib(stbi__zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!stbi__parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + a->hit_zeof_once = 0; + do { + final = stbi__zreceive(a,1); + type = stbi__zreceive(a,2); + if (type == 0) { + if (!stbi__parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer+len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context *s) +{ + static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi__context *s; + stbi_uc *idata, *expanded, *out; + int depth; +} stbi__png; + + +enum { + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filter used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first +}; + +static stbi_uc first_row_filter[5] = +{ + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_sub // Paeth with b=c=0 turns out to be equivalent to sub +}; + +static int stbi__paeth(int a, int b, int c) +{ + // This formulation looks very different from the reference in the PNG spec, but is + // actually equivalent and has favorable data dependencies and admits straightforward + // generation of branch-free code, which helps performance significantly. + int thresh = c*3 - (a + b); + int lo = a < b ? a : b; + int hi = a < b ? b : a; + int t0 = (hi <= thresh) ? lo : c; + int t1 = (thresh <= lo) ? hi : t0; + return t1; +} + +static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; + +// adds an extra all-255 alpha channel +// dest == src is legal +// img_n must be 1 or 3 +static void stbi__create_png_alpha_expand8(stbi_uc *dest, stbi_uc *src, stbi__uint32 x, int img_n) +{ + int i; + // must process data backwards since we allow dest==src + if (img_n == 1) { + for (i=x-1; i >= 0; --i) { + dest[i*2+1] = 255; + dest[i*2+0] = src[i]; + } + } else { + STBI_ASSERT(img_n == 3); + for (i=x-1; i >= 0; --i) { + dest[i*4+3] = 255; + dest[i*4+2] = src[i*3+2]; + dest[i*4+1] = src[i*3+1]; + dest[i*4+0] = src[i*3+0]; + } + } +} + +// create the png data from post-deflated data +static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color) +{ + int bytes = (depth == 16 ? 2 : 1); + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n*bytes; + stbi__uint32 img_len, img_width_bytes; + stbi_uc *filter_buf; + int all_ok = 1; + int k; + int img_n = s->img_n; // copy it into a local for later + + int output_bytes = out_n*bytes; + int filter_bytes = img_n*bytes; + int width = x; + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + // note: error exits here don't need to clean up a->out individually, + // stbi__do_png always does on error. + if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG"); + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + if (!stbi__mad2sizes_valid(img_width_bytes, y, img_width_bytes)) return stbi__err("too large", "Corrupt PNG"); + img_len = (img_width_bytes + 1) * y; + + // we used to check for exact match between raw_len and img_len on non-interlaced PNGs, + // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros), + // so just check for raw_len < img_len always. + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); + + // Allocate two scan lines worth of filter workspace buffer. + filter_buf = (stbi_uc *) stbi__malloc_mad2(img_width_bytes, 2, 0); + if (!filter_buf) return stbi__err("outofmem", "Out of memory"); + + // Filtering for low-bit-depth images + if (depth < 8) { + filter_bytes = 1; + width = img_width_bytes; + } + + for (j=0; j < y; ++j) { + // cur/prior filter buffers alternate + stbi_uc *cur = filter_buf + (j & 1)*img_width_bytes; + stbi_uc *prior = filter_buf + (~j & 1)*img_width_bytes; + stbi_uc *dest = a->out + stride*j; + int nk = width * filter_bytes; + int filter = *raw++; + + // check filter type + if (filter > 4) { + all_ok = stbi__err("invalid filter","Corrupt PNG"); + break; + } + + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + + // perform actual filtering + switch (filter) { + case STBI__F_none: + memcpy(cur, raw, nk); + break; + case STBI__F_sub: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); + break; + case STBI__F_up: + for (k = 0; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); + break; + case STBI__F_avg: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); + break; + case STBI__F_paeth: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); // prior[k] == stbi__paeth(0,prior[k],0) + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes], prior[k], prior[k-filter_bytes])); + break; + case STBI__F_avg_first: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); + break; + } + + raw += nk; + + // expand decoded bits in cur to dest, also adding an extra alpha channel if desired + if (depth < 8) { + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + stbi_uc *in = cur; + stbi_uc *out = dest; + stbi_uc inb = 0; + stbi__uint32 nsmp = x*img_n; + + // expand bits to bytes first + if (depth == 4) { + for (i=0; i < nsmp; ++i) { + if ((i & 1) == 0) inb = *in++; + *out++ = scale * (inb >> 4); + inb <<= 4; + } + } else if (depth == 2) { + for (i=0; i < nsmp; ++i) { + if ((i & 3) == 0) inb = *in++; + *out++ = scale * (inb >> 6); + inb <<= 2; + } + } else { + STBI_ASSERT(depth == 1); + for (i=0; i < nsmp; ++i) { + if ((i & 7) == 0) inb = *in++; + *out++ = scale * (inb >> 7); + inb <<= 1; + } + } + + // insert alpha=255 values if desired + if (img_n != out_n) + stbi__create_png_alpha_expand8(dest, dest, x, img_n); + } else if (depth == 8) { + if (img_n == out_n) + memcpy(dest, cur, x*img_n); + else + stbi__create_png_alpha_expand8(dest, cur, x, img_n); + } else if (depth == 16) { + // convert the image data from big-endian to platform-native + stbi__uint16 *dest16 = (stbi__uint16*)dest; + stbi__uint32 nsmp = x*img_n; + + if (img_n == out_n) { + for (i = 0; i < nsmp; ++i, ++dest16, cur += 2) + *dest16 = (cur[0] << 8) | cur[1]; + } else { + STBI_ASSERT(img_n+1 == out_n); + if (img_n == 1) { + for (i = 0; i < x; ++i, dest16 += 2, cur += 2) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = 0xffff; + } + } else { + STBI_ASSERT(img_n == 3); + for (i = 0; i < x; ++i, dest16 += 4, cur += 6) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = (cur[2] << 8) | cur[3]; + dest16[2] = (cur[4] << 8) | cur[5]; + dest16[3] = 0xffff; + } + } + } + } + } + + STBI_FREE(filter_buf); + if (!all_ok) return 0; + + return 1; +} + +static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced) +{ + int bytes = (depth == 16 ? 2 : 1); + int out_bytes = out_n * bytes; + stbi_uc *final; + int p; + if (!interlaced) + return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color); + + // de-interlacing + final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0); + if (!final) return stbi__err("outofmem", "Out of memory"); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y; + if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) { + STBI_FREE(final); + return 0; + } + for (j=0; j < y; ++j) { + for (i=0; i < x; ++i) { + int out_y = j*yspc[p]+yorig[p]; + int out_x = i*xspc[p]+xorig[p]; + memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes, + a->out + (j*x+i)*out_bytes, out_bytes); + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; + } + } + a->out = final; + + return 1; +} + +static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi__uint16 *p = (stbi__uint16*) z->out; + + // compute color-based transparency, assuming we've + // already got 65535 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i = 0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 65535); + p += 2; + } + } else { + for (i = 0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n) +{ + stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; + stbi_uc *p, *temp_out, *orig = a->out; + + p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0); + if (p == NULL) return stbi__err("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + STBI_FREE(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi__unpremultiply_on_load_global = 0; +static int stbi__de_iphone_flag_global = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_global = flag_true_if_should_convert; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__unpremultiply_on_load stbi__unpremultiply_on_load_global +#define stbi__de_iphone_flag stbi__de_iphone_flag_global +#else +static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set; +static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set; + +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply; + stbi__unpremultiply_on_load_set = 1; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_local = flag_true_if_should_convert; + stbi__de_iphone_flag_set = 1; +} + +#define stbi__unpremultiply_on_load (stbi__unpremultiply_on_load_set \ + ? stbi__unpremultiply_on_load_local \ + : stbi__unpremultiply_on_load_global) +#define stbi__de_iphone_flag (stbi__de_iphone_flag_set \ + ? stbi__de_iphone_flag_local \ + : stbi__de_iphone_flag_global) +#endif // STBI_THREAD_LOCAL + +static void stbi__de_iphone(stbi__png *z) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + stbi_uc half = a / 2; + p[0] = (p[2] * 255 + half) / a; + p[1] = (p[1] * 255 + half) / a; + p[2] = ( t * 255 + half) / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d)) + +static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) +{ + stbi_uc palette[1024], pal_img_n=0; + stbi_uc has_trans=0, tc[3]={0}; + stbi__uint16 tc16[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=0, is_iphone=0; + stbi__context *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!stbi__check_png_header(s)) return 0; + + if (scan == STBI__SCAN_type) return 1; + + for (;;) { + stbi__pngchunk c = stbi__get_chunk_header(s); + switch (c.type) { + case STBI__PNG_TYPE('C','g','B','I'): + is_iphone = 1; + stbi__skip(s, c.length); + break; + case STBI__PNG_TYPE('I','H','D','R'): { + int comp,filter; + if (!first) return stbi__err("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG"); + s->img_x = stbi__get32be(s); + s->img_y = stbi__get32be(s); + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG"); + comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG"); + filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG"); + interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG"); + } + // even with SCAN_header, have to scan to see if we have a tRNS + break; + } + + case STBI__PNG_TYPE('P','L','T','E'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = stbi__get8(s); + palette[i*4+1] = stbi__get8(s); + palette[i*4+2] = stbi__get8(s); + palette[i*4+3] = 255; + } + break; + } + + case STBI__PNG_TYPE('t','R','N','S'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = stbi__get8(s); + } else { + if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG"); + if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); + has_trans = 1; + // non-paletted with tRNS = constant alpha. if header-scanning, we can stop now. + if (scan == STBI__SCAN_header) { ++s->img_n; return 1; } + if (z->depth == 16) { + for (k = 0; k < s->img_n && k < 3; ++k) // extra loop test to suppress false GCC warning + tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is + } else { + for (k = 0; k < s->img_n && k < 3; ++k) + tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger + } + } + break; + } + + case STBI__PNG_TYPE('I','D','A','T'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG"); + if (scan == STBI__SCAN_header) { + // header scan definitely stops at first IDAT + if (pal_img_n) + s->img_n = pal_img_n; + return 1; + } + if (c.length > (1u << 30)) return stbi__err("IDAT size limit", "IDAT section larger than 2^30 bytes"); + if ((int)(ioff + c.length) < (int)ioff) return 0; + if (ioff + c.length > idata_limit) { + stbi__uint32 idata_limit_old = idata_limit; + stbi_uc *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + STBI_NOTUSED(idata_limit_old); + p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + z->idata = p; + } + if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case STBI__PNG_TYPE('I','E','N','D'): { + stbi__uint32 raw_len, bpl; + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (scan != STBI__SCAN_load) return 1; + if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component + raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */; + z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone); + if (z->expanded == NULL) return 0; // zlib should set error + STBI_FREE(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0; + if (has_trans) { + if (z->depth == 16) { + if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0; + } else { + if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; + } + } + if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) + stbi__de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } else if (has_trans) { + // non-paletted image with tRNS -> source image has (constant) alpha + ++s->img_n; + } + STBI_FREE(z->expanded); z->expanded = NULL; + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + return 1; + } + + default: + // if critical, fail + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX PNG chunk not known"; + invalid_chunk[0] = STBI__BYTECAST(c.type >> 24); + invalid_chunk[1] = STBI__BYTECAST(c.type >> 16); + invalid_chunk[2] = STBI__BYTECAST(c.type >> 8); + invalid_chunk[3] = STBI__BYTECAST(c.type >> 0); + #endif + return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type"); + } + stbi__skip(s, c.length); + break; + } + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + } +} + +static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri) +{ + void *result=NULL; + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { + if (p->depth <= 8) + ri->bits_per_channel = 8; + else if (p->depth == 16) + ri->bits_per_channel = 16; + else + return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth"); + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + if (ri->bits_per_channel == 8) + result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + else + result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_n; + } + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; + + return result; +} + +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp, ri); +} + +static int stbi__png_test(stbi__context *s) +{ + int r; + r = stbi__check_png_header(s); + stbi__rewind(s); + return r; +} + +static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) +{ + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { + stbi__rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__png p; + p.s = s; + return stbi__png_info_raw(&p, x, y, comp); +} + +static int stbi__png_is16(stbi__context *s) +{ + stbi__png p; + p.s = s; + if (!stbi__png_info_raw(&p, NULL, NULL, NULL)) + return 0; + if (p.depth != 16) { + stbi__rewind(p.s); + return 0; + } + return 1; +} +#endif + +// Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP +static int stbi__bmp_test_raw(stbi__context *s) +{ + int r; + int sz; + if (stbi__get8(s) != 'B') return 0; + if (stbi__get8(s) != 'M') return 0; + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + stbi__get32le(s); // discard data offset + sz = stbi__get32le(s); + r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124); + return r; +} + +static int stbi__bmp_test(stbi__context *s) +{ + int r = stbi__bmp_test_raw(s); + stbi__rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int stbi__high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) { n += 16; z >>= 16; } + if (z >= 0x00100) { n += 8; z >>= 8; } + if (z >= 0x00010) { n += 4; z >>= 4; } + if (z >= 0x00004) { n += 2; z >>= 2; } + if (z >= 0x00002) { n += 1;/* >>= 1;*/ } + return n; +} + +static int stbi__bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +// extract an arbitrarily-aligned N-bit value (N=bits) +// from v, and then make it 8-bits long and fractionally +// extend it to full full range. +static int stbi__shiftsigned(unsigned int v, int shift, int bits) +{ + static unsigned int mul_table[9] = { + 0, + 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/, + 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/, + }; + static unsigned int shift_table[9] = { + 0, 0,0,1,0,2,4,6,0, + }; + if (shift < 0) + v <<= -shift; + else + v >>= shift; + STBI_ASSERT(v < 256); + v >>= (8-bits); + STBI_ASSERT(bits >= 0 && bits <= 8); + return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits]; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; + int extra_read; +} stbi__bmp_data; + +static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress) +{ + // BI_BITFIELDS specifies masks explicitly, don't override + if (compress == 3) + return 1; + + if (compress == 0) { + if (info->bpp == 16) { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } else if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } else { + // otherwise, use defaults, which is all-0 + info->mr = info->mg = info->mb = info->ma = 0; + } + return 1; + } + return 0; // error +} + +static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info) +{ + int hsz; + if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP"); + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + info->mr = info->mg = info->mb = info->ma = 0; + info->extra_read = 14; + + if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP"); + + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = stbi__get16le(s); + s->img_y = stbi__get16le(s); + } else { + s->img_x = stbi__get32le(s); + s->img_y = stbi__get32le(s); + } + if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); + info->bpp = stbi__get16le(s); + if (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes + if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel + stbi__get32le(s); // discard sizeof + stbi__get32le(s); // discard hres + stbi__get32le(s); // discard vres + stbi__get32le(s); // discard colorsused + stbi__get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + } + if (info->bpp == 16 || info->bpp == 32) { + if (compress == 0) { + stbi__bmp_set_mask_defaults(info, compress); + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->extra_read += 12; + // not documented, but generated by photoshop and handled by mspaint + if (info->mr == info->mg && info->mg == info->mb) { + // ?!?!? + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else { + // V4/V5 header + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs + stbi__bmp_set_mask_defaults(info, compress); + stbi__get32le(s); // discard color space + for (i=0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void *) 1; +} + + +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + unsigned int mr=0,mg=0,mb=0,ma=0, all_a; + stbi_uc pal[256][4]; + int psize=0,i,j,width; + int flip_vertically, pad, target; + stbi__bmp_data info; + STBI_NOTUSED(ri); + + info.all_a = 255; + if (stbi__bmp_parse_header(s, &info) == NULL) + return NULL; // error code already set + + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + mr = info.mr; + mg = info.mg; + mb = info.mb; + ma = info.ma; + all_a = info.all_a; + + if (info.hsz == 12) { + if (info.bpp < 24) + psize = (info.offset - info.extra_read - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - info.extra_read - info.hsz) >> 2; + } + if (psize == 0) { + // accept some number of extra bytes after the header, but if the offset points either to before + // the header ends or implies a large amount of extra data, reject the file as malformed + int bytes_read_so_far = s->callback_already_read + (int)(s->img_buffer - s->img_buffer_original); + int header_limit = 1024; // max we actually read is below 256 bytes currently. + int extra_data_limit = 256*4; // what ordinarily goes here is a palette; 256 entries*4 bytes is its max size. + if (bytes_read_so_far <= 0 || bytes_read_so_far > header_limit) { + return stbi__errpuc("bad header", "Corrupt BMP"); + } + // we established that bytes_read_so_far is positive and sensible. + // the first half of this test rejects offsets that are either too small positives, or + // negative, and guarantees that info.offset >= bytes_read_so_far > 0. this in turn + // ensures the number computed in the second half of the test can't overflow. + if (info.offset < bytes_read_so_far || info.offset - bytes_read_so_far > extra_data_limit) { + return stbi__errpuc("bad offset", "Corrupt BMP"); + } else { + stbi__skip(s, info.offset - bytes_read_so_far); + } + } + + if (info.bpp == 24 && ma == 0xff000000) + s->img_n = 3; + else + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + + // sanity-check size + if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "Corrupt BMP"); + + out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (info.bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); + pal[i][3] = 255; + } + stbi__skip(s, info.offset - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 1) width = (s->img_x + 7) >> 3; + else if (info.bpp == 4) width = (s->img_x + 1) >> 1; + else if (info.bpp == 8) width = s->img_x; + else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + if (info.bpp == 1) { + for (j=0; j < (int) s->img_y; ++j) { + int bit_offset = 7, v = stbi__get8(s); + for (i=0; i < (int) s->img_x; ++i) { + int color = (v>>bit_offset)&0x1; + out[z++] = pal[color][0]; + out[z++] = pal[color][1]; + out[z++] = pal[color][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + if((--bit_offset) < 0) { + bit_offset = 7; + v = stbi__get8(s); + } + } + stbi__skip(s, pad); + } + } else { + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=stbi__get8(s),v2=0; + if (info.bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (info.bpp == 8) ? stbi__get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + stbi__skip(s, pad); + } + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + stbi__skip(s, info.offset - info.extra_read - info.hsz); + if (info.bpp == 24) width = 3 * s->img_x; + else if (info.bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (info.bpp == 24) { + easy = 1; + } else if (info.bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr); + gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg); + bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb); + ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma); + if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + unsigned char a; + out[z+2] = stbi__get8(s); + out[z+1] = stbi__get8(s); + out[z+0] = stbi__get8(s); + z += 3; + a = (easy == 2 ? stbi__get8(s) : 255); + all_a |= a; + if (target == 4) out[z++] = a; + } + } else { + int bpp = info.bpp; + for (i=0; i < (int) s->img_x; ++i) { + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); + unsigned int a; + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount)); + a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255); + all_a |= a; + if (target == 4) out[z++] = STBI__BYTECAST(a); + } + } + stbi__skip(s, pad); + } + } + + // if alpha channel is all 0s, replace with all 255s + if (target == 4 && all_a == 0) + for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4) + out[i] = 255; + + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i]; p1[i] = p2[i]; p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} +#endif + +// Targa Truevision - TGA +// by Jonathan Dummer +#ifndef STBI_NO_TGA +// returns STBI_rgb or whatever, 0 on error +static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16) +{ + // only RGB or RGBA (incl. 16bit) or grey allowed + if (is_rgb16) *is_rgb16 = 0; + switch(bits_per_pixel) { + case 8: return STBI_grey; + case 16: if(is_grey) return STBI_grey_alpha; + // fallthrough + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fallthrough + case 32: return bits_per_pixel/8; + default: return 0; + } +} + +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp; + int sz, tga_colormap_type; + stbi__get8(s); // discard Offset + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { + stbi__rewind(s); + return 0; // only RGB or indexed allowed + } + tga_image_type = stbi__get8(s); // image type + if ( tga_colormap_type == 1 ) { // colormapped (paletted) image + if (tga_image_type != 1 && tga_image_type != 9) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip image x and y origin + tga_colormap_bpp = sz; + } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE + if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) { + stbi__rewind(s); + return 0; // only RGB or grey allowed, +/- RLE + } + stbi__skip(s,9); // skip colormap specification and image x/y origin + tga_colormap_bpp = 0; + } + tga_w = stbi__get16le(s); + if( tga_w < 1 ) { + stbi__rewind(s); + return 0; // test width + } + tga_h = stbi__get16le(s); + if( tga_h < 1 ) { + stbi__rewind(s); + return 0; // test height + } + tga_bits_per_pixel = stbi__get8(s); // bits per pixel + stbi__get8(s); // ignore alpha bits + if (tga_colormap_bpp != 0) { + if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) { + // when using a colormap, tga_bits_per_pixel is the size of the indexes + // I don't think anything but 8 or 16bit indexes makes sense + stbi__rewind(s); + return 0; + } + tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL); + } else { + tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL); + } + if(!tga_comp) { + stbi__rewind(s); + return 0; + } + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp; + return 1; // seems to have passed everything +} + +static int stbi__tga_test(stbi__context *s) +{ + int res = 0; + int sz, tga_color_type; + stbi__get8(s); // discard Offset + tga_color_type = stbi__get8(s); // color type + if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed + sz = stbi__get8(s); // image type + if ( tga_color_type == 1 ) { // colormapped (paletted) image + if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9 + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + stbi__skip(s,4); // skip image x and y origin + } else { // "normal" image w/o colormap + if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE + stbi__skip(s,9); // skip colormap specification and image x/y origin + } + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height + sz = stbi__get8(s); // bits per pixel + if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + + res = 1; // if we got this far, everything's good and we can return 1 instead of 0 + +errorEnd: + stbi__rewind(s); + return res; +} + +// read 16bit value and convert to 24bit RGB +static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = (stbi__uint16)stbi__get16le(s); + stbi__uint16 fiveBitMask = 31; + // we have 3 channels with 5bits each + int r = (px >> 10) & fiveBitMask; + int g = (px >> 5) & fiveBitMask; + int b = px & fiveBitMask; + // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later + out[0] = (stbi_uc)((r * 255)/31); + out[1] = (stbi_uc)((g * 255)/31); + out[2] = (stbi_uc)((b * 255)/31); + + // some people claim that the most significant bit might be used for alpha + // (possibly if an alpha-bit is set in the "image descriptor byte") + // but that only made 16bit test images completely translucent.. + // so let's treat all 15 and 16bit TGAs as RGB with no alpha. +} + +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + // read in the TGA header stuff + int tga_offset = stbi__get8(s); + int tga_indexed = stbi__get8(s); + int tga_image_type = stbi__get8(s); + int tga_is_RLE = 0; + int tga_palette_start = stbi__get16le(s); + int tga_palette_len = stbi__get16le(s); + int tga_palette_bits = stbi__get8(s); + int tga_x_origin = stbi__get16le(s); + int tga_y_origin = stbi__get16le(s); + int tga_width = stbi__get16le(s); + int tga_height = stbi__get16le(s); + int tga_bits_per_pixel = stbi__get8(s); + int tga_comp, tga_rgb16=0; + int tga_inverted = stbi__get8(s); + // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?) + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4] = {0}; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + STBI_NOTUSED(ri); + STBI_NOTUSED(tga_x_origin); // @TODO + STBI_NOTUSED(tga_y_origin); // @TODO + + if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16); + else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16); + + if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency + return stbi__errpuc("bad format", "Can't find out TGA pixelformat"); + + // tga info + *x = tga_width; + *y = tga_height; + if (comp) *comp = tga_comp; + + if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0)) + return stbi__errpuc("too large", "Corrupt TGA"); + + tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0); + if (!tga_data) return stbi__errpuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + stbi__skip(s, tga_offset ); + + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { + for (i=0; i < tga_height; ++i) { + int row = tga_inverted ? tga_height -i - 1 : i; + stbi_uc *tga_row = tga_data + row*tga_width*tga_comp; + stbi__getn(s, tga_row, tga_width * tga_comp); + } + } else { + // do I need to load a palette? + if ( tga_indexed) + { + if (tga_palette_len == 0) { /* you have to have at least one entry! */ + STBI_FREE(tga_data); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0); + if (!tga_palette) { + STBI_FREE(tga_data); + return stbi__errpuc("outofmem", "Out of memory"); + } + if (tga_rgb16) { + stbi_uc *pal_entry = tga_palette; + STBI_ASSERT(tga_comp == STBI_rgb); + for (i=0; i < tga_palette_len; ++i) { + stbi__tga_read_rgb16(s, pal_entry); + pal_entry += tga_comp; + } + } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) { + STBI_FREE(tga_data); + STBI_FREE(tga_palette); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + } + // load the data + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = stbi__get8(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in index, then perform the lookup + int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s); + if ( pal_idx >= tga_palette_len ) { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else if(tga_rgb16) { + STBI_ASSERT(tga_comp == STBI_rgb); + stbi__tga_read_rgb16(s, raw_data); + } else { + // read in the data raw + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = stbi__get8(s); + } + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + + // copy data + for (j = 0; j < tga_comp; ++j) + tga_data[i*tga_comp+j] = raw_data[j]; + + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * tga_comp; + int index2 = (tga_height - 1 - j) * tga_width * tga_comp; + for (i = tga_width * tga_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + STBI_FREE( tga_palette ); + } + } + + // swap RGB - if the source data was RGB16, it already is in the right order + if (tga_comp >= 3 && !tga_rgb16) + { + unsigned char* tga_pixel = tga_data; + for (i=0; i < tga_width * tga_height; ++i) + { + unsigned char temp = tga_pixel[0]; + tga_pixel[0] = tga_pixel[2]; + tga_pixel[2] = temp; + tga_pixel += tga_comp; + } + } + + // convert to target component count + if (req_comp && req_comp != tga_comp) + tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height); + + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + STBI_NOTUSED(tga_palette_start); + // OK, done + return tga_data; +} +#endif + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s) +{ + int r = (stbi__get32be(s) == 0x38425053); + stbi__rewind(s); + return r; +} + +static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount) +{ + int count, nleft, len; + + count = 0; + while ((nleft = pixelCount - count) > 0) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + if (len > nleft) return 0; // corrupt data + count += len; + while (len) { + *p = stbi__get8(s); + p += 4; + len--; + } + } else if (len > 128) { + stbi_uc val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len = 257 - len; + if (len > nleft) return 0; // corrupt data + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + + return 1; +} + +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + int pixelCount; + int channelCount, compression; + int channel, i; + int bitdepth; + int w,h; + stbi_uc *out; + STBI_NOTUSED(ri); + + // Check identifier + if (stbi__get32be(s) != 0x38425053) // "8BPS" + return stbi__errpuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (stbi__get16be(s) != 1) + return stbi__errpuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + stbi__skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) + return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = stbi__get32be(s); + w = stbi__get32be(s); + + if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // Make sure the depth is 8 bits. + bitdepth = stbi__get16be(s); + if (bitdepth != 8 && bitdepth != 16) + return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (stbi__get16be(s) != 3) + return stbi__errpuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + stbi__skip(s,stbi__get32be(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + stbi__skip(s, stbi__get32be(s) ); + + // Skip the reserved data. + stbi__skip(s, stbi__get32be(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = stbi__get16be(s); + if (compression > 1) + return stbi__errpuc("bad compression", "PSD has an unknown compression format"); + + // Check size + if (!stbi__mad3sizes_valid(4, w, h, 0)) + return stbi__errpuc("too large", "Corrupt PSD"); + + // Create the destination image. + + if (!compression && bitdepth == 16 && bpc == 16) { + out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0); + ri->bits_per_channel = 16; + } else + out = (stbi_uc *) stbi__malloc(4 * w*h); + + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceded by a 2-byte data count for each row in the data, + // which we're going to just skip. + stbi__skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); + } else { + // Read the RLE data. + if (!stbi__psd_decode_rle(s, p, pixelCount)) { + STBI_FREE(out); + return stbi__errpuc("corrupt", "bad RLE data"); + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + if (channel >= channelCount) { + // Fill this channel with default data. + if (bitdepth == 16 && bpc == 16) { + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + stbi__uint16 val = channel == 3 ? 65535 : 0; + for (i = 0; i < pixelCount; i++, q += 4) + *q = val; + } else { + stbi_uc *p = out+channel; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } + } else { + if (ri->bits_per_channel == 16) { // output bpc + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + for (i = 0; i < pixelCount; i++, q += 4) + *q = (stbi__uint16) stbi__get16be(s); + } else { + stbi_uc *p = out+channel; + if (bitdepth == 16) { // input bpc + for (i = 0; i < pixelCount; i++, p += 4) + *p = (stbi_uc) (stbi__get16be(s) >> 8); + } else { + for (i = 0; i < pixelCount; i++, p += 4) + *p = stbi__get8(s); + } + } + } + } + } + + // remove weird white matte from PSD + if (channelCount >= 4) { + if (ri->bits_per_channel == 16) { + for (i=0; i < w*h; ++i) { + stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i; + if (pixel[3] != 0 && pixel[3] != 65535) { + float a = pixel[3] / 65535.0f; + float ra = 1.0f / a; + float inv_a = 65535.0f * (1 - ra); + pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a); + pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a); + pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a); + } + } + } else { + for (i=0; i < w*h; ++i) { + unsigned char *pixel = out + 4*i; + if (pixel[3] != 0 && pixel[3] != 255) { + float a = pixel[3] / 255.0f; + float ra = 1.0f / a; + float inv_a = 255.0f * (1 - ra); + pixel[0] = (unsigned char) (pixel[0]*ra + inv_a); + pixel[1] = (unsigned char) (pixel[1]*ra + inv_a); + pixel[2] = (unsigned char) (pixel[2]*ra + inv_a); + } + } + } + } + + // convert to desired output format + if (req_comp && req_comp != 4) { + if (ri->bits_per_channel == 16) + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h); + else + out = stbi__convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + if (comp) *comp = 4; + *y = h; + *x = w; + + return out; +} +#endif + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +#ifndef STBI_NO_PIC +static int stbi__pic_is4(stbi__context *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (stbi__get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int stbi__pic_test_core(stbi__context *s) +{ + int i; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + stbi__get8(s); + + if (!stbi__pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} stbi__pic_packet; + +static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short"); + dest[i]=stbi__get8(s); + } + } + + return dest; +} + +static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + stbi__pic_packet packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return stbi__errpuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + + act_comp |= packet->channel; + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; ytype) { + default: + return stbi__errpuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;xchannel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=stbi__get8(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (stbi_uc) left; + + if (!stbi__readval(s,packet->channel,value)) return 0; + + for(i=0; ichannel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = stbi__get8(s), i; + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + + if (count==128) + count = stbi__get16be(s); + else + count -= 127; + if (count > left) + return stbi__errpuc("bad file","scanline overrun"); + + if (!stbi__readval(s,packet->channel,value)) + return 0; + + for(i=0;ichannel,dest,value); + } else { // Raw + ++count; + if (count>left) return stbi__errpuc("bad file","scanline overrun"); + + for(i=0;ichannel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri) +{ + stbi_uc *result; + int i, x,y, internal_comp; + STBI_NOTUSED(ri); + + if (!comp) comp = &internal_comp; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + + if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode"); + + stbi__get32be(s); //skip `ratio' + stbi__get16be(s); //skip `fields' + stbi__get16be(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0); + if (!result) return stbi__errpuc("outofmem", "Out of memory"); + memset(result, 0xff, x*y*4); + + if (!stbi__pic_load_core(s,x,y,comp, result)) { + STBI_FREE(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=stbi__convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi__pic_test(stbi__context *s) +{ + int r = stbi__pic_test_core(s); + stbi__rewind(s); + return r; +} +#endif + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb + +#ifndef STBI_NO_GIF +typedef struct +{ + stbi__int16 prefix; + stbi_uc first; + stbi_uc suffix; +} stbi__gif_lzw; + +typedef struct +{ + int w,h; + stbi_uc *out; // output buffer (always 4 components) + stbi_uc *background; // The current "background" as far as a gif is concerned + stbi_uc *history; + int flags, bgindex, ratio, transparent, eflags; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[8192]; + stbi_uc *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; + int delay; +} stbi__gif; + +static int stbi__gif_test_raw(stbi__context *s) +{ + int sz; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0; + sz = stbi__get8(s); + if (sz != '9' && sz != '7') return 0; + if (stbi__get8(s) != 'a') return 0; + return 1; +} + +static int stbi__gif_test(stbi__context *s) +{ + int r = stbi__gif_test_raw(s); + stbi__rewind(s); + return r; +} + +static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + pal[i][3] = transp == i ? 0 : 255; + } +} + +static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) +{ + stbi_uc version; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') + return stbi__err("not GIF", "Corrupt GIF"); + + version = stbi__get8(s); + if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF"); + if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF"); + + stbi__g_failure_reason = ""; + g->w = stbi__get16le(s); + g->h = stbi__get16le(s); + g->flags = stbi__get8(s); + g->bgindex = stbi__get8(s); + g->ratio = stbi__get8(s); + g->transparent = -1; + + if (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif)); + if (!g) return stbi__err("outofmem", "Out of memory"); + if (!stbi__gif_header(s, g, comp, 1)) { + STBI_FREE(g); + stbi__rewind( s ); + return 0; + } + if (x) *x = g->w; + if (y) *y = g->h; + STBI_FREE(g); + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + int idx; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi__out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + idx = g->cur_x + g->cur_y; + p = &g->out[idx]; + g->history[idx / 4] = 1; + + c = &g->color_table[g->codes[code].suffix * 4]; + if (c[3] > 128) { // don't render transparent pixels; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) +{ + stbi_uc lzw_cs; + stbi__int32 len, init_code; + stbi__uint32 first; + stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi__gif_lzw *p; + + lzw_cs = stbi__get8(s); + if (lzw_cs > 12) return NULL; + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (init_code = 0; init_code < clear; init_code++) { + g->codes[init_code].prefix = -1; + g->codes[init_code].first = (stbi_uc) init_code; + g->codes[init_code].suffix = (stbi_uc) init_code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = stbi__get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (stbi__int32) stbi__get8(s) << valid_bits; + valid_bits += 8; + } else { + stbi__int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + stbi__skip(s, len); + while ((len = stbi__get8(s)) > 0) + stbi__skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) { + return stbi__errpuc("no clear code", "Corrupt GIF"); + } + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 8192) { + return stbi__errpuc("too many codes", "Corrupt GIF"); + } + + p->prefix = (stbi__int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + + stbi__out_gif_code(g, (stbi__uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +// two back is the image from two frames ago, used for a very specific disposal format +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back) +{ + int dispose; + int first_frame; + int pi; + int pcount; + STBI_NOTUSED(req_comp); + + // on first frame, any non-written pixels get the background colour (non-transparent) + first_frame = 0; + if (g->out == 0) { + if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header + if (!stbi__mad3sizes_valid(4, g->w, g->h, 0)) + return stbi__errpuc("too large", "GIF image is too large"); + pcount = g->w * g->h; + g->out = (stbi_uc *) stbi__malloc(4 * pcount); + g->background = (stbi_uc *) stbi__malloc(4 * pcount); + g->history = (stbi_uc *) stbi__malloc(pcount); + if (!g->out || !g->background || !g->history) + return stbi__errpuc("outofmem", "Out of memory"); + + // image is treated as "transparent" at the start - ie, nothing overwrites the current background; + // background colour is only used for pixels that are not rendered first frame, after that "background" + // color refers to the color that was there the previous frame. + memset(g->out, 0x00, 4 * pcount); + memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent) + memset(g->history, 0x00, pcount); // pixels that were affected previous frame + first_frame = 1; + } else { + // second frame - how do we dispose of the previous one? + dispose = (g->eflags & 0x1C) >> 2; + pcount = g->w * g->h; + + if ((dispose == 3) && (two_back == 0)) { + dispose = 2; // if I don't have an image to revert back to, default to the old background + } + + if (dispose == 3) { // use previous graphic + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); + } + } + } else if (dispose == 2) { + // restore what was changed last frame to background before that frame; + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); + } + } + } else { + // This is a non-disposal case eithe way, so just + // leave the pixels as is, and they will become the new background + // 1: do not dispose + // 0: not specified. + } + + // background is what out is after the undoing of the previou frame; + memcpy( g->background, g->out, 4 * g->w * g->h ); + } + + // clear my history; + memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame + + for (;;) { + int tag = stbi__get8(s); + switch (tag) { + case 0x2C: /* Image Descriptor */ + { + stbi__int32 x, y, w, h; + stbi_uc *o; + + x = stbi__get16le(s); + y = stbi__get16le(s); + w = stbi__get16le(s); + h = stbi__get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return stbi__errpuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + // if the width of the specified rectangle is 0, that means + // we may not see *any* pixels or the image is malformed; + // to make sure this is caught, move the current y down to + // max_y (which is what out_gif_code checks). + if (w == 0) + g->cur_y = g->max_y; + + g->lflags = stbi__get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (stbi_uc *) g->lpal; + } else if (g->flags & 0x80) { + g->color_table = (stbi_uc *) g->pal; + } else + return stbi__errpuc("missing color table", "Corrupt GIF"); + + o = stbi__process_gif_raster(s, g); + if (!o) return NULL; + + // if this was the first frame, + pcount = g->w * g->h; + if (first_frame && (g->bgindex > 0)) { + // if first frame, any pixel not drawn to gets the background color + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi] == 0) { + g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; + memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); + } + } + } + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + int ext = stbi__get8(s); + if (ext == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths. + + // unset old transparent + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 255; + } + if (g->eflags & 0x01) { + g->transparent = stbi__get8(s); + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 0; + } + } else { + // don't need transparent + stbi__skip(s, 1); + g->transparent = -1; + } + } else { + stbi__skip(s, len); + break; + } + } + while ((len = stbi__get8(s)) != 0) { + stbi__skip(s, len); + } + break; + } + + case 0x3B: // gif stream termination code + return (stbi_uc *) s; // using '1' causes warning on some compilers + + default: + return stbi__errpuc("unknown code", "Corrupt GIF"); + } + } +} + +static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays) +{ + STBI_FREE(g->out); + STBI_FREE(g->history); + STBI_FREE(g->background); + + if (out) STBI_FREE(out); + if (delays && *delays) STBI_FREE(*delays); + return stbi__errpuc("outofmem", "Out of memory"); +} + +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + if (stbi__gif_test(s)) { + int layers = 0; + stbi_uc *u = 0; + stbi_uc *out = 0; + stbi_uc *two_back = 0; + stbi__gif g; + int stride; + int out_size = 0; + int delays_size = 0; + + STBI_NOTUSED(out_size); + STBI_NOTUSED(delays_size); + + memset(&g, 0, sizeof(g)); + if (delays) { + *delays = 0; + } + + do { + u = stbi__gif_load_next(s, &g, comp, req_comp, two_back); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + + if (u) { + *x = g.w; + *y = g.h; + ++layers; + stride = g.w * g.h * 4; + + if (out) { + void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride ); + if (!tmp) + return stbi__load_gif_main_outofmem(&g, out, delays); + else { + out = (stbi_uc*) tmp; + out_size = layers * stride; + } + + if (delays) { + int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers ); + if (!new_delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + *delays = new_delays; + delays_size = layers * sizeof(int); + } + } else { + out = (stbi_uc*)stbi__malloc( layers * stride ); + if (!out) + return stbi__load_gif_main_outofmem(&g, out, delays); + out_size = layers * stride; + if (delays) { + *delays = (int*) stbi__malloc( layers * sizeof(int) ); + if (!*delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + delays_size = layers * sizeof(int); + } + } + memcpy( out + ((layers - 1) * stride), u, stride ); + if (layers >= 2) { + two_back = out - 2 * stride; + } + + if (delays) { + (*delays)[layers - 1U] = g.delay; + } + } + } while (u != 0); + + // free temp buffer; + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + + // do the final conversion after loading everything; + if (req_comp && req_comp != 4) + out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h); + + *z = layers; + return out; + } else { + return stbi__errpuc("not GIF", "Image was not as a gif type."); + } +} + +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + STBI_NOTUSED(ri); + + u = stbi__gif_load_next(s, &g, comp, req_comp, 0); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + + // moved conversion to after successful load so that the same + // can be done for multiple frames. + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } else if (g.out) { + // if there was an error and we allocated an image buffer, free it! + STBI_FREE(g.out); + } + + // free buffers needed for multiple frame loading; + STBI_FREE(g.history); + STBI_FREE(g.background); + + return u; +} + +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp) +{ + return stbi__gif_info_raw(s,x,y,comp); +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int stbi__hdr_test_core(stbi__context *s, const char *signature) +{ + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + stbi__rewind(s); + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s, "#?RADIANCE\n"); + stbi__rewind(s); + if(!r) { + r = stbi__hdr_test_core(s, "#?RGBE\n"); + stbi__rewind(s); + } + return r; +} + +#define STBI__HDR_BUFLEN 1024 +static char *stbi__hdr_gettoken(stbi__context *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) stbi__get8(z); + + while (!stbi__at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == STBI__HDR_BUFLEN-1) { + // flush to end of line + while (!stbi__at_eof(z) && stbi__get8(z) != '\n') + ; + break; + } + c = (char) stbi__get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + const char *headerToken; + STBI_NOTUSED(ri); + + // Check identifier + headerToken = stbi__hdr_gettoken(s,buffer); + if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0) + return stbi__errpf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = (int) strtol(token, NULL, 10); + + if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0)) + return stbi__errpf("too large", "HDR image is too large"); + + // Read data + hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0); + if (!hdr_data) + return stbi__errpf("outofmem", "Out of memory"); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + stbi__getn(s, rgbe, 4); + stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = stbi__get8(s); + c2 = stbi__get8(s); + len = stbi__get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4]; + rgbe[0] = (stbi_uc) c1; + rgbe[1] = (stbi_uc) c2; + rgbe[2] = (stbi_uc) len; + rgbe[3] = (stbi_uc) stbi__get8(s); + stbi__hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + STBI_FREE(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= stbi__get8(s); + if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) { + scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0); + if (!scanline) { + STBI_FREE(hdr_data); + return stbi__errpf("outofmem", "Out of memory"); + } + } + + for (k = 0; k < 4; ++k) { + int nleft; + i = 0; + while ((nleft = width - i) > 0) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = stbi__get8(s); + } + } + } + for (i=0; i < width; ++i) + stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + if (scanline) + STBI_FREE(scanline); + } + + return hdr_data; +} + +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int dummy; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (stbi__hdr_test(s) == 0) { + stbi__rewind( s ); + return 0; + } + + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi__rewind( s ); + return 0; + } + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *y = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *x = (int) strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +#ifndef STBI_NO_BMP +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) +{ + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + if (p == NULL) { + stbi__rewind( s ); + return 0; + } + if (x) *x = s->img_x; + if (y) *y = s->img_y; + if (comp) { + if (info.bpp == 24 && info.ma == 0xff000000) + *comp = 3; + else + *comp = info.ma ? 4 : 3; + } + return 1; +} +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) +{ + int channelCount, dummy, depth; + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + *y = stbi__get32be(s); + *x = stbi__get32be(s); + depth = stbi__get16be(s); + if (depth != 8 && depth != 16) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi__psd_is16(stbi__context *s) +{ + int channelCount, depth; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + STBI_NOTUSED(stbi__get32be(s)); + STBI_NOTUSED(stbi__get32be(s)); + depth = stbi__get16be(s); + if (depth != 16) { + stbi__rewind( s ); + return 0; + } + return 1; +} +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained,dummy; + stbi__pic_packet packets[10]; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) { + stbi__rewind(s); + return 0; + } + + stbi__skip(s, 88); + + *x = stbi__get16be(s); + *y = stbi__get16be(s); + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi__rewind( s ); + return 0; + } + + stbi__skip(s, 8); + + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + act_comp |= packet->channel; + + if (stbi__at_eof(s)) { + stbi__rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi__rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} +#endif + +// ************************************************************************************************* +// Portable Gray Map and Portable Pixel Map loader +// by Ken Miller +// +// PGM: http://netpbm.sourceforge.net/doc/pgm.html +// PPM: http://netpbm.sourceforge.net/doc/ppm.html +// +// Known limitations: +// Does not support comments in the header section +// Does not support ASCII image data (formats P2 and P3) + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context *s) +{ + char p, t; + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + return 1; +} + +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + STBI_NOTUSED(ri); + + ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n); + if (ri->bits_per_channel == 0) + return 0; + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + + if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (!stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8))) { + STBI_FREE(out); + return stbi__errpuc("bad PNM", "PNM file truncated"); + } + + if (req_comp && req_comp != s->img_n) { + if (ri->bits_per_channel == 16) { + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, s->img_n, req_comp, s->img_x, s->img_y); + } else { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + } + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context *s, char *c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char) stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' ) + *c = (char) stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context *s, char *c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value*10 + (*c - '0'); + *c = (char) stbi__get8(s); + if((value > 214748364) || (value == 214748364 && *c > '7')) + return stbi__err("integer parse overflow", "Parsing an integer in the PPM header overflowed a 32-bit int"); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv, dummy; + char c, p, t; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + stbi__rewind(s); + + // Get identifier + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind(s); + return 0; + } + + *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm + + c = (char) stbi__get8(s); + stbi__pnm_skip_whitespace(s, &c); + + *x = stbi__pnm_getinteger(s, &c); // read width + if(*x == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + if (*y == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + if (maxv > 65535) + return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images"); + else if (maxv > 255) + return 16; + else + return 8; +} + +static int stbi__pnm_is16(stbi__context *s) +{ + if (stbi__pnm_info(s, NULL, NULL, NULL) == 16) + return 1; + return 0; +} +#endif + +static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; + #endif + + // test tga last because it's a crappy test! + #ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; + #endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +static int stbi__is_16_main(stbi__context *s) +{ + #ifndef STBI_NO_PNG + if (stbi__png_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_is16(s)) return 1; + #endif + return 0; +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} + +STBIDEF int stbi_is_16_bit(char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_is_16_bit_from_file(f); + fclose(f); + return result; +} + +STBIDEF int stbi_is_16_bit_from_file(FILE *f) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__is_16_main(&s); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__is_16_main(&s); +} + +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__is_16_main(&s); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug + 1-bit BMP + *_is_16_bit api + avoid warnings + 2.16 (2017-07-23) all functions have 16-bit variants; + STBI_NO_STDIO works again; + compilation fixes; + fix rounding in unpremultiply; + optimize vertical flip; + disable raw_len validation; + documentation fixes + 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode; + warning fixes; disable run-time SSE detection on gcc; + uniform handling of optional "return" values; + thread-safe initialization of zlib tables + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) allocate large structures on the stack + remove white matting for transparent PSD + fix reported channel count for PNG & BMP + re-enable SSE2 in non-gcc 64-bit + support RGB-formatted JPEG + read 16-bit PNGs (only as 8-bit) + 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED + 2.09 (2016-01-16) allow comments in PNM files + 16-bit-per-pixel TGA (not bit-per-component) + info() for TGA could break due to .hdr handling + info() for BMP to shares code instead of sloppy parse + can use STBI_REALLOC_SIZED if allocator doesn't support realloc + code cleanup + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) fix compiler warnings + partial animated GIF support + limited 16-bpc PSD support + #ifdef unused functions + bug with < 92 byte PIC,PNM,HDR,TGA + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) extra corruption checking (mmozeiko) + stbi_set_flip_vertically_on_load (nguillemot) + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2 + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg) + progressive JPEG (stb) + PGM/PPM support (Ken Miller) + STBI_MALLOC,STBI_REALLOC,STBI_FREE + GIF bugfix -- seemingly never worked + STBI_NO_*, STBI_ONLY_* + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb) + optimize PNG (ryg) + fix bug in interlaced PNG with user-specified channel count (stb) + 1.46 (2014-08-26) + fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG + 1.45 (2014-08-16) + fix MSVC-ARM internal compiler error by wrapping malloc + 1.44 (2014-08-07) + various warning fixes from Ronny Chevalier + 1.43 (2014-07-15) + fix MSVC-only compiler problem in code changed in 1.42 + 1.42 (2014-07-09) + don't define _CRT_SECURE_NO_WARNINGS (affects user code) + fixes to stbi__cleanup_jpeg path + added STBI_ASSERT to avoid requiring assert.h + 1.41 (2014-06-25) + fix search&replace from 1.36 that messed up comments/error messages + 1.40 (2014-06-22) + fix gcc struct-initialization warning + 1.39 (2014-06-15) + fix to TGA optimization when req_comp != number of components in TGA; + fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite) + add support for BMP version 5 (more ignored fields) + 1.38 (2014-06-06) + suppress MSVC warnings on integer casts truncating values + fix accidental rename of 'skip' field of I/O + 1.37 (2014-06-04) + remove duplicate typedef + 1.36 (2014-06-03) + convert to header file single-file library + if de-iphone isn't set, load iphone images color-swapped instead of returning NULL + 1.35 (2014-05-27) + various warnings + fix broken STBI_SIMD path + fix bug where stbi_load_from_file no longer left file pointer in correct place + fix broken non-easy path for 32-bit BMP (possibly never used) + TGA optimization by Arseny Kapoulkine + 1.34 (unknown) + use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-stbi_uc to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva) + 1.21 fix use of 'stbi_uc' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 (2008-08-02) + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - stbi__convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 (2006-11-19) + first released version +*/ + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_image_write.h b/include/stb_image_write.h new file mode 100644 index 0000000..e4b32ed --- /dev/null +++ b/include/stb_image_write.h @@ -0,0 +1,1724 @@ +/* stb_image_write - v1.16 - public domain - http://nothings.org/stb + writes out PNG/BMP/TGA/JPEG/HDR images to C stdio - Sean Barrett 2010-2015 + no warranty implied; use at your own risk + + Before #including, + + #define STB_IMAGE_WRITE_IMPLEMENTATION + + in the file that you want to have the implementation. + + Will probably not work correctly with strict-aliasing optimizations. + +ABOUT: + + This header file is a library for writing images to C stdio or a callback. + + The PNG output is not optimal; it is 20-50% larger than the file + written by a decent optimizing implementation; though providing a custom + zlib compress function (see STBIW_ZLIB_COMPRESS) can mitigate that. + This library is designed for source code compactness and simplicity, + not optimal image file size or run-time performance. + +BUILDING: + + You can #define STBIW_ASSERT(x) before the #include to avoid using assert.h. + You can #define STBIW_MALLOC(), STBIW_REALLOC(), and STBIW_FREE() to replace + malloc,realloc,free. + You can #define STBIW_MEMMOVE() to replace memmove() + You can #define STBIW_ZLIB_COMPRESS to use a custom zlib-style compress function + for PNG compression (instead of the builtin one), it must have the following signature: + unsigned char * my_compress(unsigned char *data, int data_len, int *out_len, int quality); + The returned data will be freed with STBIW_FREE() (free() by default), + so it must be heap allocated with STBIW_MALLOC() (malloc() by default), + +UNICODE: + + If compiling for Windows and you wish to use Unicode filenames, compile + with + #define STBIW_WINDOWS_UTF8 + and pass utf8-encoded filenames. Call stbiw_convert_wchar_to_utf8 to convert + Windows wchar_t filenames to utf8. + +USAGE: + + There are five functions, one for each image file format: + + int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes); + int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data); + int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data); + int stbi_write_jpg(char const *filename, int w, int h, int comp, const void *data, int quality); + int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data); + + void stbi_flip_vertically_on_write(int flag); // flag is non-zero to flip data vertically + + There are also five equivalent functions that use an arbitrary write function. You are + expected to open/close your file-equivalent before and after calling these: + + int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes); + int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); + int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); + int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data); + int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality); + + where the callback is: + void stbi_write_func(void *context, void *data, int size); + + You can configure it with these global variables: + int stbi_write_tga_with_rle; // defaults to true; set to 0 to disable RLE + int stbi_write_png_compression_level; // defaults to 8; set to higher for more compression + int stbi_write_force_png_filter; // defaults to -1; set to 0..5 to force a filter mode + + + You can define STBI_WRITE_NO_STDIO to disable the file variant of these + functions, so the library will not use stdio.h at all. However, this will + also disable HDR writing, because it requires stdio for formatted output. + + Each function returns 0 on failure and non-0 on success. + + The functions create an image file defined by the parameters. The image + is a rectangle of pixels stored from left-to-right, top-to-bottom. + Each pixel contains 'comp' channels of data stored interleaved with 8-bits + per channel, in the following order: 1=Y, 2=YA, 3=RGB, 4=RGBA. (Y is + monochrome color.) The rectangle is 'w' pixels wide and 'h' pixels tall. + The *data pointer points to the first byte of the top-left-most pixel. + For PNG, "stride_in_bytes" is the distance in bytes from the first byte of + a row of pixels to the first byte of the next row of pixels. + + PNG creates output files with the same number of components as the input. + The BMP format expands Y to RGB in the file format and does not + output alpha. + + PNG supports writing rectangles of data even when the bytes storing rows of + data are not consecutive in memory (e.g. sub-rectangles of a larger image), + by supplying the stride between the beginning of adjacent rows. The other + formats do not. (Thus you cannot write a native-format BMP through the BMP + writer, both because it is in BGR order and because it may have padding + at the end of the line.) + + PNG allows you to set the deflate compression level by setting the global + variable 'stbi_write_png_compression_level' (it defaults to 8). + + HDR expects linear float data. Since the format is always 32-bit rgb(e) + data, alpha (if provided) is discarded, and for monochrome data it is + replicated across all three channels. + + TGA supports RLE or non-RLE compressed data. To use non-RLE-compressed + data, set the global variable 'stbi_write_tga_with_rle' to 0. + + JPEG does ignore alpha channels in input data; quality is between 1 and 100. + Higher quality looks better but results in a bigger image. + JPEG baseline (no JPEG progressive). + +CREDITS: + + + Sean Barrett - PNG/BMP/TGA + Baldur Karlsson - HDR + Jean-Sebastien Guay - TGA monochrome + Tim Kelsey - misc enhancements + Alan Hickman - TGA RLE + Emmanuel Julien - initial file IO callback implementation + Jon Olick - original jo_jpeg.cpp code + Daniel Gibson - integrate JPEG, allow external zlib + Aarni Koskela - allow choosing PNG filter + + bugfixes: + github:Chribba + Guillaume Chereau + github:jry2 + github:romigrou + Sergio Gonzalez + Jonas Karlsson + Filip Wasil + Thatcher Ulrich + github:poppolopoppo + Patrick Boettcher + github:xeekworx + Cap Petschulat + Simon Rodriguez + Ivan Tikhonov + github:ignotion + Adam Schackart + Andrew Kensler + +LICENSE + + See end of file for license information. + +*/ + +#ifndef INCLUDE_STB_IMAGE_WRITE_H +#define INCLUDE_STB_IMAGE_WRITE_H + +#include + +// if STB_IMAGE_WRITE_STATIC causes problems, try defining STBIWDEF to 'inline' or 'static inline' +#ifndef STBIWDEF +#ifdef STB_IMAGE_WRITE_STATIC +#define STBIWDEF static +#else +#ifdef __cplusplus +#define STBIWDEF extern "C" +#else +#define STBIWDEF extern +#endif +#endif +#endif + +#ifndef STB_IMAGE_WRITE_STATIC // C++ forbids static forward declarations +STBIWDEF int stbi_write_tga_with_rle; +STBIWDEF int stbi_write_png_compression_level; +STBIWDEF int stbi_write_force_png_filter; +#endif + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_png(char const *filename, int w, int h, int comp, const void *data, int stride_in_bytes); +STBIWDEF int stbi_write_bmp(char const *filename, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_tga(char const *filename, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_hdr(char const *filename, int w, int h, int comp, const float *data); +STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality); + +#ifdef STBIW_WINDOWS_UTF8 +STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif +#endif + +typedef void stbi_write_func(void *context, void *data, int size); + +STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data, int stride_in_bytes); +STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const void *data); +STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int w, int h, int comp, const float *data); +STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality); + +STBIWDEF void stbi_flip_vertically_on_write(int flip_boolean); + +#endif//INCLUDE_STB_IMAGE_WRITE_H + +#ifdef STB_IMAGE_WRITE_IMPLEMENTATION + +#ifdef _WIN32 + #ifndef _CRT_SECURE_NO_WARNINGS + #define _CRT_SECURE_NO_WARNINGS + #endif + #ifndef _CRT_NONSTDC_NO_DEPRECATE + #define _CRT_NONSTDC_NO_DEPRECATE + #endif +#endif + +#ifndef STBI_WRITE_NO_STDIO +#include +#endif // STBI_WRITE_NO_STDIO + +#include +#include +#include +#include + +#if defined(STBIW_MALLOC) && defined(STBIW_FREE) && (defined(STBIW_REALLOC) || defined(STBIW_REALLOC_SIZED)) +// ok +#elif !defined(STBIW_MALLOC) && !defined(STBIW_FREE) && !defined(STBIW_REALLOC) && !defined(STBIW_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBIW_MALLOC, STBIW_FREE, and STBIW_REALLOC (or STBIW_REALLOC_SIZED)." +#endif + +#ifndef STBIW_MALLOC +#define STBIW_MALLOC(sz) malloc(sz) +#define STBIW_REALLOC(p,newsz) realloc(p,newsz) +#define STBIW_FREE(p) free(p) +#endif + +#ifndef STBIW_REALLOC_SIZED +#define STBIW_REALLOC_SIZED(p,oldsz,newsz) STBIW_REALLOC(p,newsz) +#endif + + +#ifndef STBIW_MEMMOVE +#define STBIW_MEMMOVE(a,b,sz) memmove(a,b,sz) +#endif + + +#ifndef STBIW_ASSERT +#include +#define STBIW_ASSERT(x) assert(x) +#endif + +#define STBIW_UCHAR(x) (unsigned char) ((x) & 0xff) + +#ifdef STB_IMAGE_WRITE_STATIC +static int stbi_write_png_compression_level = 8; +static int stbi_write_tga_with_rle = 1; +static int stbi_write_force_png_filter = -1; +#else +int stbi_write_png_compression_level = 8; +int stbi_write_tga_with_rle = 1; +int stbi_write_force_png_filter = -1; +#endif + +static int stbi__flip_vertically_on_write = 0; + +STBIWDEF void stbi_flip_vertically_on_write(int flag) +{ + stbi__flip_vertically_on_write = flag; +} + +typedef struct +{ + stbi_write_func *func; + void *context; + unsigned char buffer[64]; + int buf_used; +} stbi__write_context; + +// initialize a callback-based context +static void stbi__start_write_callbacks(stbi__write_context *s, stbi_write_func *c, void *context) +{ + s->func = c; + s->context = context; +} + +#ifndef STBI_WRITE_NO_STDIO + +static void stbi__stdio_write(void *context, void *data, int size) +{ + fwrite(data,1,size,(FILE*) context); +} + +#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8) +#ifdef __cplusplus +#define STBIW_EXTERN extern "C" +#else +#define STBIW_EXTERN extern +#endif +STBIW_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBIW_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); + +STBIWDEF int stbiw_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbiw__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBIW_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + +static int stbi__start_write_file(stbi__write_context *s, const char *filename) +{ + FILE *f = stbiw__fopen(filename, "wb"); + stbi__start_write_callbacks(s, stbi__stdio_write, (void *) f); + return f != NULL; +} + +static void stbi__end_write_file(stbi__write_context *s) +{ + fclose((FILE *)s->context); +} + +#endif // !STBI_WRITE_NO_STDIO + +typedef unsigned int stbiw_uint32; +typedef int stb_image_write_test[sizeof(stbiw_uint32)==4 ? 1 : -1]; + +static void stbiw__writefv(stbi__write_context *s, const char *fmt, va_list v) +{ + while (*fmt) { + switch (*fmt++) { + case ' ': break; + case '1': { unsigned char x = STBIW_UCHAR(va_arg(v, int)); + s->func(s->context,&x,1); + break; } + case '2': { int x = va_arg(v,int); + unsigned char b[2]; + b[0] = STBIW_UCHAR(x); + b[1] = STBIW_UCHAR(x>>8); + s->func(s->context,b,2); + break; } + case '4': { stbiw_uint32 x = va_arg(v,int); + unsigned char b[4]; + b[0]=STBIW_UCHAR(x); + b[1]=STBIW_UCHAR(x>>8); + b[2]=STBIW_UCHAR(x>>16); + b[3]=STBIW_UCHAR(x>>24); + s->func(s->context,b,4); + break; } + default: + STBIW_ASSERT(0); + return; + } + } +} + +static void stbiw__writef(stbi__write_context *s, const char *fmt, ...) +{ + va_list v; + va_start(v, fmt); + stbiw__writefv(s, fmt, v); + va_end(v); +} + +static void stbiw__write_flush(stbi__write_context *s) +{ + if (s->buf_used) { + s->func(s->context, &s->buffer, s->buf_used); + s->buf_used = 0; + } +} + +static void stbiw__putc(stbi__write_context *s, unsigned char c) +{ + s->func(s->context, &c, 1); +} + +static void stbiw__write1(stbi__write_context *s, unsigned char a) +{ + if ((size_t)s->buf_used + 1 > sizeof(s->buffer)) + stbiw__write_flush(s); + s->buffer[s->buf_used++] = a; +} + +static void stbiw__write3(stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c) +{ + int n; + if ((size_t)s->buf_used + 3 > sizeof(s->buffer)) + stbiw__write_flush(s); + n = s->buf_used; + s->buf_used = n+3; + s->buffer[n+0] = a; + s->buffer[n+1] = b; + s->buffer[n+2] = c; +} + +static void stbiw__write_pixel(stbi__write_context *s, int rgb_dir, int comp, int write_alpha, int expand_mono, unsigned char *d) +{ + unsigned char bg[3] = { 255, 0, 255}, px[3]; + int k; + + if (write_alpha < 0) + stbiw__write1(s, d[comp - 1]); + + switch (comp) { + case 2: // 2 pixels = mono + alpha, alpha is written separately, so same as 1-channel case + case 1: + if (expand_mono) + stbiw__write3(s, d[0], d[0], d[0]); // monochrome bmp + else + stbiw__write1(s, d[0]); // monochrome TGA + break; + case 4: + if (!write_alpha) { + // composite against pink background + for (k = 0; k < 3; ++k) + px[k] = bg[k] + ((d[k] - bg[k]) * d[3]) / 255; + stbiw__write3(s, px[1 - rgb_dir], px[1], px[1 + rgb_dir]); + break; + } + /* FALLTHROUGH */ + case 3: + stbiw__write3(s, d[1 - rgb_dir], d[1], d[1 + rgb_dir]); + break; + } + if (write_alpha > 0) + stbiw__write1(s, d[comp - 1]); +} + +static void stbiw__write_pixels(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad, int expand_mono) +{ + stbiw_uint32 zero = 0; + int i,j, j_end; + + if (y <= 0) + return; + + if (stbi__flip_vertically_on_write) + vdir *= -1; + + if (vdir < 0) { + j_end = -1; j = y-1; + } else { + j_end = y; j = 0; + } + + for (; j != j_end; j += vdir) { + for (i=0; i < x; ++i) { + unsigned char *d = (unsigned char *) data + (j*x+i)*comp; + stbiw__write_pixel(s, rgb_dir, comp, write_alpha, expand_mono, d); + } + stbiw__write_flush(s); + s->func(s->context, &zero, scanline_pad); + } +} + +static int stbiw__outfile(stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, int expand_mono, void *data, int alpha, int pad, const char *fmt, ...) +{ + if (y < 0 || x < 0) { + return 0; + } else { + va_list v; + va_start(v, fmt); + stbiw__writefv(s, fmt, v); + va_end(v); + stbiw__write_pixels(s,rgb_dir,vdir,x,y,comp,data,alpha,pad, expand_mono); + return 1; + } +} + +static int stbi_write_bmp_core(stbi__write_context *s, int x, int y, int comp, const void *data) +{ + if (comp != 4) { + // write RGB bitmap + int pad = (-x*3) & 3; + return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *) data,0,pad, + "11 4 22 4" "4 44 22 444444", + 'B', 'M', 14+40+(x*3+pad)*y, 0,0, 14+40, // file header + 40, x,y, 1,24, 0,0,0,0,0,0); // bitmap header + } else { + // RGBA bitmaps need a v4 header + // use BI_BITFIELDS mode with 32bpp and alpha mask + // (straight BI_RGB with alpha mask doesn't work in most readers) + return stbiw__outfile(s,-1,-1,x,y,comp,1,(void *)data,1,0, + "11 4 22 4" "4 44 22 444444 4444 4 444 444 444 444", + 'B', 'M', 14+108+x*y*4, 0, 0, 14+108, // file header + 108, x,y, 1,32, 3,0,0,0,0,0, 0xff0000,0xff00,0xff,0xff000000u, 0, 0,0,0, 0,0,0, 0,0,0, 0,0,0); // bitmap V4 header + } +} + +STBIWDEF int stbi_write_bmp_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_bmp_core(&s, x, y, comp, data); +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_bmp(char const *filename, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_bmp_core(&s, x, y, comp, data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif //!STBI_WRITE_NO_STDIO + +static int stbi_write_tga_core(stbi__write_context *s, int x, int y, int comp, void *data) +{ + int has_alpha = (comp == 2 || comp == 4); + int colorbytes = has_alpha ? comp-1 : comp; + int format = colorbytes < 2 ? 3 : 2; // 3 color channels (RGB/RGBA) = 2, 1 color channel (Y/YA) = 3 + + if (y < 0 || x < 0) + return 0; + + if (!stbi_write_tga_with_rle) { + return stbiw__outfile(s, -1, -1, x, y, comp, 0, (void *) data, has_alpha, 0, + "111 221 2222 11", 0, 0, format, 0, 0, 0, 0, 0, x, y, (colorbytes + has_alpha) * 8, has_alpha * 8); + } else { + int i,j,k; + int jend, jdir; + + stbiw__writef(s, "111 221 2222 11", 0,0,format+8, 0,0,0, 0,0,x,y, (colorbytes + has_alpha) * 8, has_alpha * 8); + + if (stbi__flip_vertically_on_write) { + j = 0; + jend = y; + jdir = 1; + } else { + j = y-1; + jend = -1; + jdir = -1; + } + for (; j != jend; j += jdir) { + unsigned char *row = (unsigned char *) data + j * x * comp; + int len; + + for (i = 0; i < x; i += len) { + unsigned char *begin = row + i * comp; + int diff = 1; + len = 1; + + if (i < x - 1) { + ++len; + diff = memcmp(begin, row + (i + 1) * comp, comp); + if (diff) { + const unsigned char *prev = begin; + for (k = i + 2; k < x && len < 128; ++k) { + if (memcmp(prev, row + k * comp, comp)) { + prev += comp; + ++len; + } else { + --len; + break; + } + } + } else { + for (k = i + 2; k < x && len < 128; ++k) { + if (!memcmp(begin, row + k * comp, comp)) { + ++len; + } else { + break; + } + } + } + } + + if (diff) { + unsigned char header = STBIW_UCHAR(len - 1); + stbiw__write1(s, header); + for (k = 0; k < len; ++k) { + stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin + k * comp); + } + } else { + unsigned char header = STBIW_UCHAR(len - 129); + stbiw__write1(s, header); + stbiw__write_pixel(s, -1, comp, has_alpha, 0, begin); + } + } + } + stbiw__write_flush(s); + } + return 1; +} + +STBIWDEF int stbi_write_tga_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_tga_core(&s, x, y, comp, (void *) data); +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_tga(char const *filename, int x, int y, int comp, const void *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_tga_core(&s, x, y, comp, (void *) data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR writer +// by Baldur Karlsson + +#define stbiw__max(a, b) ((a) > (b) ? (a) : (b)) + +#ifndef STBI_WRITE_NO_STDIO + +static void stbiw__linear_to_rgbe(unsigned char *rgbe, float *linear) +{ + int exponent; + float maxcomp = stbiw__max(linear[0], stbiw__max(linear[1], linear[2])); + + if (maxcomp < 1e-32f) { + rgbe[0] = rgbe[1] = rgbe[2] = rgbe[3] = 0; + } else { + float normalize = (float) frexp(maxcomp, &exponent) * 256.0f/maxcomp; + + rgbe[0] = (unsigned char)(linear[0] * normalize); + rgbe[1] = (unsigned char)(linear[1] * normalize); + rgbe[2] = (unsigned char)(linear[2] * normalize); + rgbe[3] = (unsigned char)(exponent + 128); + } +} + +static void stbiw__write_run_data(stbi__write_context *s, int length, unsigned char databyte) +{ + unsigned char lengthbyte = STBIW_UCHAR(length+128); + STBIW_ASSERT(length+128 <= 255); + s->func(s->context, &lengthbyte, 1); + s->func(s->context, &databyte, 1); +} + +static void stbiw__write_dump_data(stbi__write_context *s, int length, unsigned char *data) +{ + unsigned char lengthbyte = STBIW_UCHAR(length); + STBIW_ASSERT(length <= 128); // inconsistent with spec but consistent with official code + s->func(s->context, &lengthbyte, 1); + s->func(s->context, data, length); +} + +static void stbiw__write_hdr_scanline(stbi__write_context *s, int width, int ncomp, unsigned char *scratch, float *scanline) +{ + unsigned char scanlineheader[4] = { 2, 2, 0, 0 }; + unsigned char rgbe[4]; + float linear[3]; + int x; + + scanlineheader[2] = (width&0xff00)>>8; + scanlineheader[3] = (width&0x00ff); + + /* skip RLE for images too small or large */ + if (width < 8 || width >= 32768) { + for (x=0; x < width; x++) { + switch (ncomp) { + case 4: /* fallthrough */ + case 3: linear[2] = scanline[x*ncomp + 2]; + linear[1] = scanline[x*ncomp + 1]; + linear[0] = scanline[x*ncomp + 0]; + break; + default: + linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0]; + break; + } + stbiw__linear_to_rgbe(rgbe, linear); + s->func(s->context, rgbe, 4); + } + } else { + int c,r; + /* encode into scratch buffer */ + for (x=0; x < width; x++) { + switch(ncomp) { + case 4: /* fallthrough */ + case 3: linear[2] = scanline[x*ncomp + 2]; + linear[1] = scanline[x*ncomp + 1]; + linear[0] = scanline[x*ncomp + 0]; + break; + default: + linear[0] = linear[1] = linear[2] = scanline[x*ncomp + 0]; + break; + } + stbiw__linear_to_rgbe(rgbe, linear); + scratch[x + width*0] = rgbe[0]; + scratch[x + width*1] = rgbe[1]; + scratch[x + width*2] = rgbe[2]; + scratch[x + width*3] = rgbe[3]; + } + + s->func(s->context, scanlineheader, 4); + + /* RLE each component separately */ + for (c=0; c < 4; c++) { + unsigned char *comp = &scratch[width*c]; + + x = 0; + while (x < width) { + // find first run + r = x; + while (r+2 < width) { + if (comp[r] == comp[r+1] && comp[r] == comp[r+2]) + break; + ++r; + } + if (r+2 >= width) + r = width; + // dump up to first run + while (x < r) { + int len = r-x; + if (len > 128) len = 128; + stbiw__write_dump_data(s, len, &comp[x]); + x += len; + } + // if there's a run, output it + if (r+2 < width) { // same test as what we break out of in search loop, so only true if we break'd + // find next byte after run + while (r < width && comp[r] == comp[x]) + ++r; + // output run up to r + while (x < r) { + int len = r-x; + if (len > 127) len = 127; + stbiw__write_run_data(s, len, comp[x]); + x += len; + } + } + } + } + } +} + +static int stbi_write_hdr_core(stbi__write_context *s, int x, int y, int comp, float *data) +{ + if (y <= 0 || x <= 0 || data == NULL) + return 0; + else { + // Each component is stored separately. Allocate scratch space for full output scanline. + unsigned char *scratch = (unsigned char *) STBIW_MALLOC(x*4); + int i, len; + char buffer[128]; + char header[] = "#?RADIANCE\n# Written by stb_image_write.h\nFORMAT=32-bit_rle_rgbe\n"; + s->func(s->context, header, sizeof(header)-1); + +#ifdef __STDC_LIB_EXT1__ + len = sprintf_s(buffer, sizeof(buffer), "EXPOSURE= 1.0000000000000\n\n-Y %d +X %d\n", y, x); +#else + len = sprintf(buffer, "EXPOSURE= 1.0000000000000\n\n-Y %d +X %d\n", y, x); +#endif + s->func(s->context, buffer, len); + + for(i=0; i < y; i++) + stbiw__write_hdr_scanline(s, x, comp, scratch, data + comp*x*(stbi__flip_vertically_on_write ? y-1-i : i)); + STBIW_FREE(scratch); + return 1; + } +} + +STBIWDEF int stbi_write_hdr_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const float *data) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_hdr_core(&s, x, y, comp, (float *) data); +} + +STBIWDEF int stbi_write_hdr(char const *filename, int x, int y, int comp, const float *data) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_hdr_core(&s, x, y, comp, (float *) data); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif // STBI_WRITE_NO_STDIO + + +////////////////////////////////////////////////////////////////////////////// +// +// PNG writer +// + +#ifndef STBIW_ZLIB_COMPRESS +// stretchy buffer; stbiw__sbpush() == vector<>::push_back() -- stbiw__sbcount() == vector<>::size() +#define stbiw__sbraw(a) ((int *) (void *) (a) - 2) +#define stbiw__sbm(a) stbiw__sbraw(a)[0] +#define stbiw__sbn(a) stbiw__sbraw(a)[1] + +#define stbiw__sbneedgrow(a,n) ((a)==0 || stbiw__sbn(a)+n >= stbiw__sbm(a)) +#define stbiw__sbmaybegrow(a,n) (stbiw__sbneedgrow(a,(n)) ? stbiw__sbgrow(a,n) : 0) +#define stbiw__sbgrow(a,n) stbiw__sbgrowf((void **) &(a), (n), sizeof(*(a))) + +#define stbiw__sbpush(a, v) (stbiw__sbmaybegrow(a,1), (a)[stbiw__sbn(a)++] = (v)) +#define stbiw__sbcount(a) ((a) ? stbiw__sbn(a) : 0) +#define stbiw__sbfree(a) ((a) ? STBIW_FREE(stbiw__sbraw(a)),0 : 0) + +static void *stbiw__sbgrowf(void **arr, int increment, int itemsize) +{ + int m = *arr ? 2*stbiw__sbm(*arr)+increment : increment+1; + void *p = STBIW_REALLOC_SIZED(*arr ? stbiw__sbraw(*arr) : 0, *arr ? (stbiw__sbm(*arr)*itemsize + sizeof(int)*2) : 0, itemsize * m + sizeof(int)*2); + STBIW_ASSERT(p); + if (p) { + if (!*arr) ((int *) p)[1] = 0; + *arr = (void *) ((int *) p + 2); + stbiw__sbm(*arr) = m; + } + return *arr; +} + +static unsigned char *stbiw__zlib_flushf(unsigned char *data, unsigned int *bitbuffer, int *bitcount) +{ + while (*bitcount >= 8) { + stbiw__sbpush(data, STBIW_UCHAR(*bitbuffer)); + *bitbuffer >>= 8; + *bitcount -= 8; + } + return data; +} + +static int stbiw__zlib_bitrev(int code, int codebits) +{ + int res=0; + while (codebits--) { + res = (res << 1) | (code & 1); + code >>= 1; + } + return res; +} + +static unsigned int stbiw__zlib_countm(unsigned char *a, unsigned char *b, int limit) +{ + int i; + for (i=0; i < limit && i < 258; ++i) + if (a[i] != b[i]) break; + return i; +} + +static unsigned int stbiw__zhash(unsigned char *data) +{ + stbiw_uint32 hash = data[0] + (data[1] << 8) + (data[2] << 16); + hash ^= hash << 3; + hash += hash >> 5; + hash ^= hash << 4; + hash += hash >> 17; + hash ^= hash << 25; + hash += hash >> 6; + return hash; +} + +#define stbiw__zlib_flush() (out = stbiw__zlib_flushf(out, &bitbuf, &bitcount)) +#define stbiw__zlib_add(code,codebits) \ + (bitbuf |= (code) << bitcount, bitcount += (codebits), stbiw__zlib_flush()) +#define stbiw__zlib_huffa(b,c) stbiw__zlib_add(stbiw__zlib_bitrev(b,c),c) +// default huffman tables +#define stbiw__zlib_huff1(n) stbiw__zlib_huffa(0x30 + (n), 8) +#define stbiw__zlib_huff2(n) stbiw__zlib_huffa(0x190 + (n)-144, 9) +#define stbiw__zlib_huff3(n) stbiw__zlib_huffa(0 + (n)-256,7) +#define stbiw__zlib_huff4(n) stbiw__zlib_huffa(0xc0 + (n)-280,8) +#define stbiw__zlib_huff(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : (n) <= 255 ? stbiw__zlib_huff2(n) : (n) <= 279 ? stbiw__zlib_huff3(n) : stbiw__zlib_huff4(n)) +#define stbiw__zlib_huffb(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : stbiw__zlib_huff2(n)) + +#define stbiw__ZHASH 16384 + +#endif // STBIW_ZLIB_COMPRESS + +STBIWDEF unsigned char * stbi_zlib_compress(unsigned char *data, int data_len, int *out_len, int quality) +{ +#ifdef STBIW_ZLIB_COMPRESS + // user provided a zlib compress implementation, use that + return STBIW_ZLIB_COMPRESS(data, data_len, out_len, quality); +#else // use builtin + static unsigned short lengthc[] = { 3,4,5,6,7,8,9,10,11,13,15,17,19,23,27,31,35,43,51,59,67,83,99,115,131,163,195,227,258, 259 }; + static unsigned char lengtheb[]= { 0,0,0,0,0,0,0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0 }; + static unsigned short distc[] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577, 32768 }; + static unsigned char disteb[] = { 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13 }; + unsigned int bitbuf=0; + int i,j, bitcount=0; + unsigned char *out = NULL; + unsigned char ***hash_table = (unsigned char***) STBIW_MALLOC(stbiw__ZHASH * sizeof(unsigned char**)); + if (hash_table == NULL) + return NULL; + if (quality < 5) quality = 5; + + stbiw__sbpush(out, 0x78); // DEFLATE 32K window + stbiw__sbpush(out, 0x5e); // FLEVEL = 1 + stbiw__zlib_add(1,1); // BFINAL = 1 + stbiw__zlib_add(1,2); // BTYPE = 1 -- fixed huffman + + for (i=0; i < stbiw__ZHASH; ++i) + hash_table[i] = NULL; + + i=0; + while (i < data_len-3) { + // hash next 3 bytes of data to be compressed + int h = stbiw__zhash(data+i)&(stbiw__ZHASH-1), best=3; + unsigned char *bestloc = 0; + unsigned char **hlist = hash_table[h]; + int n = stbiw__sbcount(hlist); + for (j=0; j < n; ++j) { + if (hlist[j]-data > i-32768) { // if entry lies within window + int d = stbiw__zlib_countm(hlist[j], data+i, data_len-i); + if (d >= best) { best=d; bestloc=hlist[j]; } + } + } + // when hash table entry is too long, delete half the entries + if (hash_table[h] && stbiw__sbn(hash_table[h]) == 2*quality) { + STBIW_MEMMOVE(hash_table[h], hash_table[h]+quality, sizeof(hash_table[h][0])*quality); + stbiw__sbn(hash_table[h]) = quality; + } + stbiw__sbpush(hash_table[h],data+i); + + if (bestloc) { + // "lazy matching" - check match at *next* byte, and if it's better, do cur byte as literal + h = stbiw__zhash(data+i+1)&(stbiw__ZHASH-1); + hlist = hash_table[h]; + n = stbiw__sbcount(hlist); + for (j=0; j < n; ++j) { + if (hlist[j]-data > i-32767) { + int e = stbiw__zlib_countm(hlist[j], data+i+1, data_len-i-1); + if (e > best) { // if next match is better, bail on current match + bestloc = NULL; + break; + } + } + } + } + + if (bestloc) { + int d = (int) (data+i - bestloc); // distance back + STBIW_ASSERT(d <= 32767 && best <= 258); + for (j=0; best > lengthc[j+1]-1; ++j); + stbiw__zlib_huff(j+257); + if (lengtheb[j]) stbiw__zlib_add(best - lengthc[j], lengtheb[j]); + for (j=0; d > distc[j+1]-1; ++j); + stbiw__zlib_add(stbiw__zlib_bitrev(j,5),5); + if (disteb[j]) stbiw__zlib_add(d - distc[j], disteb[j]); + i += best; + } else { + stbiw__zlib_huffb(data[i]); + ++i; + } + } + // write out final bytes + for (;i < data_len; ++i) + stbiw__zlib_huffb(data[i]); + stbiw__zlib_huff(256); // end of block + // pad with 0 bits to byte boundary + while (bitcount) + stbiw__zlib_add(0,1); + + for (i=0; i < stbiw__ZHASH; ++i) + (void) stbiw__sbfree(hash_table[i]); + STBIW_FREE(hash_table); + + // store uncompressed instead if compression was worse + if (stbiw__sbn(out) > data_len + 2 + ((data_len+32766)/32767)*5) { + stbiw__sbn(out) = 2; // truncate to DEFLATE 32K window and FLEVEL = 1 + for (j = 0; j < data_len;) { + int blocklen = data_len - j; + if (blocklen > 32767) blocklen = 32767; + stbiw__sbpush(out, data_len - j == blocklen); // BFINAL = ?, BTYPE = 0 -- no compression + stbiw__sbpush(out, STBIW_UCHAR(blocklen)); // LEN + stbiw__sbpush(out, STBIW_UCHAR(blocklen >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(~blocklen)); // NLEN + stbiw__sbpush(out, STBIW_UCHAR(~blocklen >> 8)); + memcpy(out+stbiw__sbn(out), data+j, blocklen); + stbiw__sbn(out) += blocklen; + j += blocklen; + } + } + + { + // compute adler32 on input + unsigned int s1=1, s2=0; + int blocklen = (int) (data_len % 5552); + j=0; + while (j < data_len) { + for (i=0; i < blocklen; ++i) { s1 += data[j+i]; s2 += s1; } + s1 %= 65521; s2 %= 65521; + j += blocklen; + blocklen = 5552; + } + stbiw__sbpush(out, STBIW_UCHAR(s2 >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(s2)); + stbiw__sbpush(out, STBIW_UCHAR(s1 >> 8)); + stbiw__sbpush(out, STBIW_UCHAR(s1)); + } + *out_len = stbiw__sbn(out); + // make returned pointer freeable + STBIW_MEMMOVE(stbiw__sbraw(out), out, *out_len); + return (unsigned char *) stbiw__sbraw(out); +#endif // STBIW_ZLIB_COMPRESS +} + +static unsigned int stbiw__crc32(unsigned char *buffer, int len) +{ +#ifdef STBIW_CRC32 + return STBIW_CRC32(buffer, len); +#else + static unsigned int crc_table[256] = + { + 0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, 0xE963A535, 0x9E6495A3, + 0x0eDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91, + 0x1DB71064, 0x6AB020F2, 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7, + 0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, 0xFA0F3D63, 0x8D080DF5, + 0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B, + 0x35B5A8FA, 0x42B2986C, 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59, + 0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, 0xCFBA9599, 0xB8BDA50F, + 0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D, + 0x76DC4190, 0x01DB7106, 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433, + 0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, 0x91646C97, 0xE6635C01, + 0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457, + 0x65B0D9C6, 0x12B7E950, 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65, + 0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, 0xA4D1C46D, 0xD3D6F4FB, + 0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9, + 0x5005713C, 0x270241AA, 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F, + 0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, 0xB7BD5C3B, 0xC0BA6CAD, + 0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683, + 0xE3630B12, 0x94643B84, 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1, + 0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, 0x196C3671, 0x6E6B06E7, + 0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5, + 0xD6D6A3E8, 0xA1D1937E, 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B, + 0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, 0x316E8EEF, 0x4669BE79, + 0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F, + 0xC5BA3BBE, 0xB2BD0B28, 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D, + 0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, 0x72076785, 0x05005713, + 0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21, + 0x86D3D2D4, 0xF1D4E242, 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777, + 0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, 0x616BFFD3, 0x166CCF45, + 0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB, + 0xAED16A4A, 0xD9D65ADC, 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9, + 0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, 0x54DE5729, 0x23D967BF, + 0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D + }; + + unsigned int crc = ~0u; + int i; + for (i=0; i < len; ++i) + crc = (crc >> 8) ^ crc_table[buffer[i] ^ (crc & 0xff)]; + return ~crc; +#endif +} + +#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=STBIW_UCHAR(a),(o)[1]=STBIW_UCHAR(b),(o)[2]=STBIW_UCHAR(c),(o)[3]=STBIW_UCHAR(d),(o)+=4) +#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v)); +#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3]) + +static void stbiw__wpcrc(unsigned char **data, int len) +{ + unsigned int crc = stbiw__crc32(*data - len - 4, len+4); + stbiw__wp32(*data, crc); +} + +static unsigned char stbiw__paeth(int a, int b, int c) +{ + int p = a + b - c, pa = abs(p-a), pb = abs(p-b), pc = abs(p-c); + if (pa <= pb && pa <= pc) return STBIW_UCHAR(a); + if (pb <= pc) return STBIW_UCHAR(b); + return STBIW_UCHAR(c); +} + +// @OPTIMIZE: provide an option that always forces left-predict or paeth predict +static void stbiw__encode_png_line(unsigned char *pixels, int stride_bytes, int width, int height, int y, int n, int filter_type, signed char *line_buffer) +{ + static int mapping[] = { 0,1,2,3,4 }; + static int firstmap[] = { 0,1,0,5,6 }; + int *mymap = (y != 0) ? mapping : firstmap; + int i; + int type = mymap[filter_type]; + unsigned char *z = pixels + stride_bytes * (stbi__flip_vertically_on_write ? height-1-y : y); + int signed_stride = stbi__flip_vertically_on_write ? -stride_bytes : stride_bytes; + + if (type==0) { + memcpy(line_buffer, z, width*n); + return; + } + + // first loop isn't optimized since it's just one pixel + for (i = 0; i < n; ++i) { + switch (type) { + case 1: line_buffer[i] = z[i]; break; + case 2: line_buffer[i] = z[i] - z[i-signed_stride]; break; + case 3: line_buffer[i] = z[i] - (z[i-signed_stride]>>1); break; + case 4: line_buffer[i] = (signed char) (z[i] - stbiw__paeth(0,z[i-signed_stride],0)); break; + case 5: line_buffer[i] = z[i]; break; + case 6: line_buffer[i] = z[i]; break; + } + } + switch (type) { + case 1: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-n]; break; + case 2: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-signed_stride]; break; + case 3: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - ((z[i-n] + z[i-signed_stride])>>1); break; + case 4: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], z[i-signed_stride], z[i-signed_stride-n]); break; + case 5: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - (z[i-n]>>1); break; + case 6: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], 0,0); break; + } +} + +STBIWDEF unsigned char *stbi_write_png_to_mem(const unsigned char *pixels, int stride_bytes, int x, int y, int n, int *out_len) +{ + int force_filter = stbi_write_force_png_filter; + int ctype[5] = { -1, 0, 4, 2, 6 }; + unsigned char sig[8] = { 137,80,78,71,13,10,26,10 }; + unsigned char *out,*o, *filt, *zlib; + signed char *line_buffer; + int j,zlen; + + if (stride_bytes == 0) + stride_bytes = x * n; + + if (force_filter >= 5) { + force_filter = -1; + } + + filt = (unsigned char *) STBIW_MALLOC((x*n+1) * y); if (!filt) return 0; + line_buffer = (signed char *) STBIW_MALLOC(x * n); if (!line_buffer) { STBIW_FREE(filt); return 0; } + for (j=0; j < y; ++j) { + int filter_type; + if (force_filter > -1) { + filter_type = force_filter; + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, force_filter, line_buffer); + } else { // Estimate the best filter by running through all of them: + int best_filter = 0, best_filter_val = 0x7fffffff, est, i; + for (filter_type = 0; filter_type < 5; filter_type++) { + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, filter_type, line_buffer); + + // Estimate the entropy of the line using this filter; the less, the better. + est = 0; + for (i = 0; i < x*n; ++i) { + est += abs((signed char) line_buffer[i]); + } + if (est < best_filter_val) { + best_filter_val = est; + best_filter = filter_type; + } + } + if (filter_type != best_filter) { // If the last iteration already got us the best filter, don't redo it + stbiw__encode_png_line((unsigned char*)(pixels), stride_bytes, x, y, j, n, best_filter, line_buffer); + filter_type = best_filter; + } + } + // when we get here, filter_type contains the filter type, and line_buffer contains the data + filt[j*(x*n+1)] = (unsigned char) filter_type; + STBIW_MEMMOVE(filt+j*(x*n+1)+1, line_buffer, x*n); + } + STBIW_FREE(line_buffer); + zlib = stbi_zlib_compress(filt, y*( x*n+1), &zlen, stbi_write_png_compression_level); + STBIW_FREE(filt); + if (!zlib) return 0; + + // each tag requires 12 bytes of overhead + out = (unsigned char *) STBIW_MALLOC(8 + 12+13 + 12+zlen + 12); + if (!out) return 0; + *out_len = 8 + 12+13 + 12+zlen + 12; + + o=out; + STBIW_MEMMOVE(o,sig,8); o+= 8; + stbiw__wp32(o, 13); // header length + stbiw__wptag(o, "IHDR"); + stbiw__wp32(o, x); + stbiw__wp32(o, y); + *o++ = 8; + *o++ = STBIW_UCHAR(ctype[n]); + *o++ = 0; + *o++ = 0; + *o++ = 0; + stbiw__wpcrc(&o,13); + + stbiw__wp32(o, zlen); + stbiw__wptag(o, "IDAT"); + STBIW_MEMMOVE(o, zlib, zlen); + o += zlen; + STBIW_FREE(zlib); + stbiw__wpcrc(&o, zlen); + + stbiw__wp32(o,0); + stbiw__wptag(o, "IEND"); + stbiw__wpcrc(&o,0); + + STBIW_ASSERT(o == out + *out_len); + + return out; +} + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_png(char const *filename, int x, int y, int comp, const void *data, int stride_bytes) +{ + FILE *f; + int len; + unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len); + if (png == NULL) return 0; + + f = stbiw__fopen(filename, "wb"); + if (!f) { STBIW_FREE(png); return 0; } + fwrite(png, 1, len, f); + fclose(f); + STBIW_FREE(png); + return 1; +} +#endif + +STBIWDEF int stbi_write_png_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int stride_bytes) +{ + int len; + unsigned char *png = stbi_write_png_to_mem((const unsigned char *) data, stride_bytes, x, y, comp, &len); + if (png == NULL) return 0; + func(context, png, len); + STBIW_FREE(png); + return 1; +} + + +/* *************************************************************************** + * + * JPEG writer + * + * This is based on Jon Olick's jo_jpeg.cpp: + * public domain Simple, Minimalistic JPEG writer - http://www.jonolick.com/code.html + */ + +static const unsigned char stbiw__jpg_ZigZag[] = { 0,1,5,6,14,15,27,28,2,4,7,13,16,26,29,42,3,8,12,17,25,30,41,43,9,11,18, + 24,31,40,44,53,10,19,23,32,39,45,52,54,20,22,33,38,46,51,55,60,21,34,37,47,50,56,59,61,35,36,48,49,57,58,62,63 }; + +static void stbiw__jpg_writeBits(stbi__write_context *s, int *bitBufP, int *bitCntP, const unsigned short *bs) { + int bitBuf = *bitBufP, bitCnt = *bitCntP; + bitCnt += bs[1]; + bitBuf |= bs[0] << (24 - bitCnt); + while(bitCnt >= 8) { + unsigned char c = (bitBuf >> 16) & 255; + stbiw__putc(s, c); + if(c == 255) { + stbiw__putc(s, 0); + } + bitBuf <<= 8; + bitCnt -= 8; + } + *bitBufP = bitBuf; + *bitCntP = bitCnt; +} + +static void stbiw__jpg_DCT(float *d0p, float *d1p, float *d2p, float *d3p, float *d4p, float *d5p, float *d6p, float *d7p) { + float d0 = *d0p, d1 = *d1p, d2 = *d2p, d3 = *d3p, d4 = *d4p, d5 = *d5p, d6 = *d6p, d7 = *d7p; + float z1, z2, z3, z4, z5, z11, z13; + + float tmp0 = d0 + d7; + float tmp7 = d0 - d7; + float tmp1 = d1 + d6; + float tmp6 = d1 - d6; + float tmp2 = d2 + d5; + float tmp5 = d2 - d5; + float tmp3 = d3 + d4; + float tmp4 = d3 - d4; + + // Even part + float tmp10 = tmp0 + tmp3; // phase 2 + float tmp13 = tmp0 - tmp3; + float tmp11 = tmp1 + tmp2; + float tmp12 = tmp1 - tmp2; + + d0 = tmp10 + tmp11; // phase 3 + d4 = tmp10 - tmp11; + + z1 = (tmp12 + tmp13) * 0.707106781f; // c4 + d2 = tmp13 + z1; // phase 5 + d6 = tmp13 - z1; + + // Odd part + tmp10 = tmp4 + tmp5; // phase 2 + tmp11 = tmp5 + tmp6; + tmp12 = tmp6 + tmp7; + + // The rotator is modified from fig 4-8 to avoid extra negations. + z5 = (tmp10 - tmp12) * 0.382683433f; // c6 + z2 = tmp10 * 0.541196100f + z5; // c2-c6 + z4 = tmp12 * 1.306562965f + z5; // c2+c6 + z3 = tmp11 * 0.707106781f; // c4 + + z11 = tmp7 + z3; // phase 5 + z13 = tmp7 - z3; + + *d5p = z13 + z2; // phase 6 + *d3p = z13 - z2; + *d1p = z11 + z4; + *d7p = z11 - z4; + + *d0p = d0; *d2p = d2; *d4p = d4; *d6p = d6; +} + +static void stbiw__jpg_calcBits(int val, unsigned short bits[2]) { + int tmp1 = val < 0 ? -val : val; + val = val < 0 ? val-1 : val; + bits[1] = 1; + while(tmp1 >>= 1) { + ++bits[1]; + } + bits[0] = val & ((1<0)&&(DU[end0pos]==0); --end0pos) { + } + // end0pos = first element in reverse order !=0 + if(end0pos == 0) { + stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB); + return DU[0]; + } + for(i = 1; i <= end0pos; ++i) { + int startpos = i; + int nrzeroes; + unsigned short bits[2]; + for (; DU[i]==0 && i<=end0pos; ++i) { + } + nrzeroes = i-startpos; + if ( nrzeroes >= 16 ) { + int lng = nrzeroes>>4; + int nrmarker; + for (nrmarker=1; nrmarker <= lng; ++nrmarker) + stbiw__jpg_writeBits(s, bitBuf, bitCnt, M16zeroes); + nrzeroes &= 15; + } + stbiw__jpg_calcBits(DU[i], bits); + stbiw__jpg_writeBits(s, bitBuf, bitCnt, HTAC[(nrzeroes<<4)+bits[1]]); + stbiw__jpg_writeBits(s, bitBuf, bitCnt, bits); + } + if(end0pos != 63) { + stbiw__jpg_writeBits(s, bitBuf, bitCnt, EOB); + } + return DU[0]; +} + +static int stbi_write_jpg_core(stbi__write_context *s, int width, int height, int comp, const void* data, int quality) { + // Constants that don't pollute global namespace + static const unsigned char std_dc_luminance_nrcodes[] = {0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0}; + static const unsigned char std_dc_luminance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11}; + static const unsigned char std_ac_luminance_nrcodes[] = {0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d}; + static const unsigned char std_ac_luminance_values[] = { + 0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08, + 0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0,0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28, + 0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59, + 0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89, + 0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6, + 0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2, + 0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa + }; + static const unsigned char std_dc_chrominance_nrcodes[] = {0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0}; + static const unsigned char std_dc_chrominance_values[] = {0,1,2,3,4,5,6,7,8,9,10,11}; + static const unsigned char std_ac_chrominance_nrcodes[] = {0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77}; + static const unsigned char std_ac_chrominance_values[] = { + 0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91, + 0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0,0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26, + 0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58, + 0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87, + 0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4, + 0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda, + 0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8,0xf9,0xfa + }; + // Huffman tables + static const unsigned short YDC_HT[256][2] = { {0,2},{2,3},{3,3},{4,3},{5,3},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9}}; + static const unsigned short UVDC_HT[256][2] = { {0,2},{1,2},{2,2},{6,3},{14,4},{30,5},{62,6},{126,7},{254,8},{510,9},{1022,10},{2046,11}}; + static const unsigned short YAC_HT[256][2] = { + {10,4},{0,2},{1,2},{4,3},{11,4},{26,5},{120,7},{248,8},{1014,10},{65410,16},{65411,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {12,4},{27,5},{121,7},{502,9},{2038,11},{65412,16},{65413,16},{65414,16},{65415,16},{65416,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {28,5},{249,8},{1015,10},{4084,12},{65417,16},{65418,16},{65419,16},{65420,16},{65421,16},{65422,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {58,6},{503,9},{4085,12},{65423,16},{65424,16},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {59,6},{1016,10},{65430,16},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {122,7},{2039,11},{65438,16},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {123,7},{4086,12},{65446,16},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {250,8},{4087,12},{65454,16},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {504,9},{32704,15},{65462,16},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {505,9},{65470,16},{65471,16},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {506,9},{65479,16},{65480,16},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1017,10},{65488,16},{65489,16},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1018,10},{65497,16},{65498,16},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2040,11},{65506,16},{65507,16},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {65515,16},{65516,16},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2041,11},{65525,16},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0} + }; + static const unsigned short UVAC_HT[256][2] = { + {0,2},{1,2},{4,3},{10,4},{24,5},{25,5},{56,6},{120,7},{500,9},{1014,10},{4084,12},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {11,4},{57,6},{246,8},{501,9},{2038,11},{4085,12},{65416,16},{65417,16},{65418,16},{65419,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {26,5},{247,8},{1015,10},{4086,12},{32706,15},{65420,16},{65421,16},{65422,16},{65423,16},{65424,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {27,5},{248,8},{1016,10},{4087,12},{65425,16},{65426,16},{65427,16},{65428,16},{65429,16},{65430,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {58,6},{502,9},{65431,16},{65432,16},{65433,16},{65434,16},{65435,16},{65436,16},{65437,16},{65438,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {59,6},{1017,10},{65439,16},{65440,16},{65441,16},{65442,16},{65443,16},{65444,16},{65445,16},{65446,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {121,7},{2039,11},{65447,16},{65448,16},{65449,16},{65450,16},{65451,16},{65452,16},{65453,16},{65454,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {122,7},{2040,11},{65455,16},{65456,16},{65457,16},{65458,16},{65459,16},{65460,16},{65461,16},{65462,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {249,8},{65463,16},{65464,16},{65465,16},{65466,16},{65467,16},{65468,16},{65469,16},{65470,16},{65471,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {503,9},{65472,16},{65473,16},{65474,16},{65475,16},{65476,16},{65477,16},{65478,16},{65479,16},{65480,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {504,9},{65481,16},{65482,16},{65483,16},{65484,16},{65485,16},{65486,16},{65487,16},{65488,16},{65489,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {505,9},{65490,16},{65491,16},{65492,16},{65493,16},{65494,16},{65495,16},{65496,16},{65497,16},{65498,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {506,9},{65499,16},{65500,16},{65501,16},{65502,16},{65503,16},{65504,16},{65505,16},{65506,16},{65507,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {2041,11},{65508,16},{65509,16},{65510,16},{65511,16},{65512,16},{65513,16},{65514,16},{65515,16},{65516,16},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}, + {16352,14},{65517,16},{65518,16},{65519,16},{65520,16},{65521,16},{65522,16},{65523,16},{65524,16},{65525,16},{0,0},{0,0},{0,0},{0,0},{0,0}, + {1018,10},{32707,15},{65526,16},{65527,16},{65528,16},{65529,16},{65530,16},{65531,16},{65532,16},{65533,16},{65534,16},{0,0},{0,0},{0,0},{0,0},{0,0} + }; + static const int YQT[] = {16,11,10,16,24,40,51,61,12,12,14,19,26,58,60,55,14,13,16,24,40,57,69,56,14,17,22,29,51,87,80,62,18,22, + 37,56,68,109,103,77,24,35,55,64,81,104,113,92,49,64,78,87,103,121,120,101,72,92,95,98,112,100,103,99}; + static const int UVQT[] = {17,18,24,47,99,99,99,99,18,21,26,66,99,99,99,99,24,26,56,99,99,99,99,99,47,66,99,99,99,99,99,99, + 99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99}; + static const float aasf[] = { 1.0f * 2.828427125f, 1.387039845f * 2.828427125f, 1.306562965f * 2.828427125f, 1.175875602f * 2.828427125f, + 1.0f * 2.828427125f, 0.785694958f * 2.828427125f, 0.541196100f * 2.828427125f, 0.275899379f * 2.828427125f }; + + int row, col, i, k, subsample; + float fdtbl_Y[64], fdtbl_UV[64]; + unsigned char YTable[64], UVTable[64]; + + if(!data || !width || !height || comp > 4 || comp < 1) { + return 0; + } + + quality = quality ? quality : 90; + subsample = quality <= 90 ? 1 : 0; + quality = quality < 1 ? 1 : quality > 100 ? 100 : quality; + quality = quality < 50 ? 5000 / quality : 200 - quality * 2; + + for(i = 0; i < 64; ++i) { + int uvti, yti = (YQT[i]*quality+50)/100; + YTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (yti < 1 ? 1 : yti > 255 ? 255 : yti); + uvti = (UVQT[i]*quality+50)/100; + UVTable[stbiw__jpg_ZigZag[i]] = (unsigned char) (uvti < 1 ? 1 : uvti > 255 ? 255 : uvti); + } + + for(row = 0, k = 0; row < 8; ++row) { + for(col = 0; col < 8; ++col, ++k) { + fdtbl_Y[k] = 1 / (YTable [stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]); + fdtbl_UV[k] = 1 / (UVTable[stbiw__jpg_ZigZag[k]] * aasf[row] * aasf[col]); + } + } + + // Write Headers + { + static const unsigned char head0[] = { 0xFF,0xD8,0xFF,0xE0,0,0x10,'J','F','I','F',0,1,1,0,0,1,0,1,0,0,0xFF,0xDB,0,0x84,0 }; + static const unsigned char head2[] = { 0xFF,0xDA,0,0xC,3,1,0,2,0x11,3,0x11,0,0x3F,0 }; + const unsigned char head1[] = { 0xFF,0xC0,0,0x11,8,(unsigned char)(height>>8),STBIW_UCHAR(height),(unsigned char)(width>>8),STBIW_UCHAR(width), + 3,1,(unsigned char)(subsample?0x22:0x11),0,2,0x11,1,3,0x11,1,0xFF,0xC4,0x01,0xA2,0 }; + s->func(s->context, (void*)head0, sizeof(head0)); + s->func(s->context, (void*)YTable, sizeof(YTable)); + stbiw__putc(s, 1); + s->func(s->context, UVTable, sizeof(UVTable)); + s->func(s->context, (void*)head1, sizeof(head1)); + s->func(s->context, (void*)(std_dc_luminance_nrcodes+1), sizeof(std_dc_luminance_nrcodes)-1); + s->func(s->context, (void*)std_dc_luminance_values, sizeof(std_dc_luminance_values)); + stbiw__putc(s, 0x10); // HTYACinfo + s->func(s->context, (void*)(std_ac_luminance_nrcodes+1), sizeof(std_ac_luminance_nrcodes)-1); + s->func(s->context, (void*)std_ac_luminance_values, sizeof(std_ac_luminance_values)); + stbiw__putc(s, 1); // HTUDCinfo + s->func(s->context, (void*)(std_dc_chrominance_nrcodes+1), sizeof(std_dc_chrominance_nrcodes)-1); + s->func(s->context, (void*)std_dc_chrominance_values, sizeof(std_dc_chrominance_values)); + stbiw__putc(s, 0x11); // HTUACinfo + s->func(s->context, (void*)(std_ac_chrominance_nrcodes+1), sizeof(std_ac_chrominance_nrcodes)-1); + s->func(s->context, (void*)std_ac_chrominance_values, sizeof(std_ac_chrominance_values)); + s->func(s->context, (void*)head2, sizeof(head2)); + } + + // Encode 8x8 macroblocks + { + static const unsigned short fillBits[] = {0x7F, 7}; + int DCY=0, DCU=0, DCV=0; + int bitBuf=0, bitCnt=0; + // comp == 2 is grey+alpha (alpha is ignored) + int ofsG = comp > 2 ? 1 : 0, ofsB = comp > 2 ? 2 : 0; + const unsigned char *dataR = (const unsigned char *)data; + const unsigned char *dataG = dataR + ofsG; + const unsigned char *dataB = dataR + ofsB; + int x, y, pos; + if(subsample) { + for(y = 0; y < height; y += 16) { + for(x = 0; x < width; x += 16) { + float Y[256], U[256], V[256]; + for(row = y, pos = 0; row < y+16; ++row) { + // row >= height => use last input row + int clamped_row = (row < height) ? row : height - 1; + int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp; + for(col = x; col < x+16; ++col, ++pos) { + // if col >= width => use pixel from last input column + int p = base_p + ((col < width) ? col : (width-1))*comp; + float r = dataR[p], g = dataG[p], b = dataB[p]; + Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128; + U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b; + V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b; + } + } + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+0, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+8, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+128, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y+136, 16, fdtbl_Y, DCY, YDC_HT, YAC_HT); + + // subsample U,V + { + float subU[64], subV[64]; + int yy, xx; + for(yy = 0, pos = 0; yy < 8; ++yy) { + for(xx = 0; xx < 8; ++xx, ++pos) { + int j = yy*32+xx*2; + subU[pos] = (U[j+0] + U[j+1] + U[j+16] + U[j+17]) * 0.25f; + subV[pos] = (V[j+0] + V[j+1] + V[j+16] + V[j+17]) * 0.25f; + } + } + DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subU, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT); + DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, subV, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT); + } + } + } + } else { + for(y = 0; y < height; y += 8) { + for(x = 0; x < width; x += 8) { + float Y[64], U[64], V[64]; + for(row = y, pos = 0; row < y+8; ++row) { + // row >= height => use last input row + int clamped_row = (row < height) ? row : height - 1; + int base_p = (stbi__flip_vertically_on_write ? (height-1-clamped_row) : clamped_row)*width*comp; + for(col = x; col < x+8; ++col, ++pos) { + // if col >= width => use pixel from last input column + int p = base_p + ((col < width) ? col : (width-1))*comp; + float r = dataR[p], g = dataG[p], b = dataB[p]; + Y[pos]= +0.29900f*r + 0.58700f*g + 0.11400f*b - 128; + U[pos]= -0.16874f*r - 0.33126f*g + 0.50000f*b; + V[pos]= +0.50000f*r - 0.41869f*g - 0.08131f*b; + } + } + + DCY = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, Y, 8, fdtbl_Y, DCY, YDC_HT, YAC_HT); + DCU = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, U, 8, fdtbl_UV, DCU, UVDC_HT, UVAC_HT); + DCV = stbiw__jpg_processDU(s, &bitBuf, &bitCnt, V, 8, fdtbl_UV, DCV, UVDC_HT, UVAC_HT); + } + } + } + + // Do the bit alignment of the EOI marker + stbiw__jpg_writeBits(s, &bitBuf, &bitCnt, fillBits); + } + + // EOI + stbiw__putc(s, 0xFF); + stbiw__putc(s, 0xD9); + + return 1; +} + +STBIWDEF int stbi_write_jpg_to_func(stbi_write_func *func, void *context, int x, int y, int comp, const void *data, int quality) +{ + stbi__write_context s = { 0 }; + stbi__start_write_callbacks(&s, func, context); + return stbi_write_jpg_core(&s, x, y, comp, (void *) data, quality); +} + + +#ifndef STBI_WRITE_NO_STDIO +STBIWDEF int stbi_write_jpg(char const *filename, int x, int y, int comp, const void *data, int quality) +{ + stbi__write_context s = { 0 }; + if (stbi__start_write_file(&s,filename)) { + int r = stbi_write_jpg_core(&s, x, y, comp, data, quality); + stbi__end_write_file(&s); + return r; + } else + return 0; +} +#endif + +#endif // STB_IMAGE_WRITE_IMPLEMENTATION + +/* Revision history + 1.16 (2021-07-11) + make Deflate code emit uncompressed blocks when it would otherwise expand + support writing BMPs with alpha channel + 1.15 (2020-07-13) unknown + 1.14 (2020-02-02) updated JPEG writer to downsample chroma channels + 1.13 + 1.12 + 1.11 (2019-08-11) + + 1.10 (2019-02-07) + support utf8 filenames in Windows; fix warnings and platform ifdefs + 1.09 (2018-02-11) + fix typo in zlib quality API, improve STB_I_W_STATIC in C++ + 1.08 (2018-01-29) + add stbi__flip_vertically_on_write, external zlib, zlib quality, choose PNG filter + 1.07 (2017-07-24) + doc fix + 1.06 (2017-07-23) + writing JPEG (using Jon Olick's code) + 1.05 ??? + 1.04 (2017-03-03) + monochrome BMP expansion + 1.03 ??? + 1.02 (2016-04-02) + avoid allocating large structures on the stack + 1.01 (2016-01-16) + STBIW_REALLOC_SIZED: support allocators with no realloc support + avoid race-condition in crc initialization + minor compile issues + 1.00 (2015-09-14) + installable file IO function + 0.99 (2015-09-13) + warning fixes; TGA rle support + 0.98 (2015-04-08) + added STBIW_MALLOC, STBIW_ASSERT etc + 0.97 (2015-01-18) + fixed HDR asserts, rewrote HDR rle logic + 0.96 (2015-01-17) + add HDR output + fix monochrome BMP + 0.95 (2014-08-17) + add monochrome TGA output + 0.94 (2014-05-31) + rename private functions to avoid conflicts with stb_image.h + 0.93 (2014-05-27) + warning fixes + 0.92 (2010-08-01) + casts to unsigned char to fix warnings + 0.91 (2010-07-17) + first public release + 0.90 first internal release +*/ + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_include.h b/include/stb_include.h new file mode 100644 index 0000000..c5db201 --- /dev/null +++ b/include/stb_include.h @@ -0,0 +1,295 @@ +// stb_include.h - v0.02 - parse and process #include directives - public domain +// +// To build this, in one source file that includes this file do +// #define STB_INCLUDE_IMPLEMENTATION +// +// This program parses a string and replaces lines of the form +// #include "foo" +// with the contents of a file named "foo". It also embeds the +// appropriate #line directives. Note that all include files must +// reside in the location specified in the path passed to the API; +// it does not check multiple directories. +// +// If the string contains a line of the form +// #inject +// then it will be replaced with the contents of the string 'inject' passed to the API. +// +// Options: +// +// Define STB_INCLUDE_LINE_GLSL to get GLSL-style #line directives +// which use numbers instead of filenames. +// +// Define STB_INCLUDE_LINE_NONE to disable output of #line directives. +// +// Standard libraries: +// +// stdio.h FILE, fopen, fclose, fseek, ftell +// stdlib.h malloc, realloc, free +// string.h strcpy, strncmp, memcpy +// +// Credits: +// +// Written by Sean Barrett. +// +// Fixes: +// Michal Klos + +#ifndef STB_INCLUDE_STB_INCLUDE_H +#define STB_INCLUDE_STB_INCLUDE_H + +// Do include-processing on the string 'str'. To free the return value, pass it to free() +char *stb_include_string(char *str, char *inject, char *path_to_includes, char *filename_for_line_directive, char error[256]); + +// Concatenate the strings 'strs' and do include-processing on the result. To free the return value, pass it to free() +char *stb_include_strings(char **strs, int count, char *inject, char *path_to_includes, char *filename_for_line_directive, char error[256]); + +// Load the file 'filename' and do include-processing on the string therein. note that +// 'filename' is opened directly; 'path_to_includes' is not used. To free the return value, pass it to free() +char *stb_include_file(char *filename, char *inject, char *path_to_includes, char error[256]); + +#endif + + +#ifdef STB_INCLUDE_IMPLEMENTATION + +#include +#include +#include + +static char *stb_include_load_file(char *filename, size_t *plen) +{ + char *text; + size_t len; + FILE *f = fopen(filename, "rb"); + if (f == 0) return 0; + fseek(f, 0, SEEK_END); + len = (size_t) ftell(f); + if (plen) *plen = len; + text = (char *) malloc(len+1); + if (text == 0) return 0; + fseek(f, 0, SEEK_SET); + fread(text, 1, len, f); + fclose(f); + text[len] = 0; + return text; +} + +typedef struct +{ + int offset; + int end; + char *filename; + int next_line_after; +} include_info; + +static include_info *stb_include_append_include(include_info *array, int len, int offset, int end, char *filename, int next_line) +{ + include_info *z = (include_info *) realloc(array, sizeof(*z) * (len+1)); + z[len].offset = offset; + z[len].end = end; + z[len].filename = filename; + z[len].next_line_after = next_line; + return z; +} + +static void stb_include_free_includes(include_info *array, int len) +{ + int i; + for (i=0; i < len; ++i) + free(array[i].filename); + free(array); +} + +static int stb_include_isspace(int ch) +{ + return (ch == ' ' || ch == '\t' || ch == '\r' || ch == '\n'); +} + +// find location of all #include and #inject +static int stb_include_find_includes(char *text, include_info **plist) +{ + int line_count = 1; + int inc_count = 0; + char *s = text, *start; + include_info *list = NULL; + while (*s) { + // parse is always at start of line when we reach here + start = s; + while (*s == ' ' || *s == '\t') + ++s; + if (*s == '#') { + ++s; + while (*s == ' ' || *s == '\t') + ++s; + if (0==strncmp(s, "include", 7) && stb_include_isspace(s[7])) { + s += 7; + while (*s == ' ' || *s == '\t') + ++s; + if (*s == '"') { + char *t = ++s; + while (*t != '"' && *t != '\n' && *t != '\r' && *t != 0) + ++t; + if (*t == '"') { + char *filename = (char *) malloc(t-s+1); + memcpy(filename, s, t-s); + filename[t-s] = 0; + s=t; + while (*s != '\r' && *s != '\n' && *s != 0) + ++s; + // s points to the newline, so s-start is everything except the newline + list = stb_include_append_include(list, inc_count++, start-text, s-text, filename, line_count+1); + } + } + } else if (0==strncmp(s, "inject", 6) && (stb_include_isspace(s[6]) || s[6]==0)) { + while (*s != '\r' && *s != '\n' && *s != 0) + ++s; + list = stb_include_append_include(list, inc_count++, start-text, s-text, NULL, line_count+1); + } + } + while (*s != '\r' && *s != '\n' && *s != 0) + ++s; + if (*s == '\r' || *s == '\n') { + s = s + (s[0] + s[1] == '\r' + '\n' ? 2 : 1); + } + ++line_count; + } + *plist = list; + return inc_count; +} + +// avoid dependency on sprintf() +static void stb_include_itoa(char str[9], int n) +{ + int i; + for (i=0; i < 8; ++i) + str[i] = ' '; + str[i] = 0; + + for (i=1; i < 8; ++i) { + str[7-i] = '0' + (n % 10); + n /= 10; + if (n == 0) + break; + } +} + +static char *stb_include_append(char *str, size_t *curlen, char *addstr, size_t addlen) +{ + str = (char *) realloc(str, *curlen + addlen); + memcpy(str + *curlen, addstr, addlen); + *curlen += addlen; + return str; +} + +char *stb_include_string(char *str, char *inject, char *path_to_includes, char *filename, char error[256]) +{ + char temp[4096]; + include_info *inc_list; + int i, num = stb_include_find_includes(str, &inc_list); + size_t source_len = strlen(str); + char *text=0; + size_t textlen=0, last=0; + for (i=0; i < num; ++i) { + text = stb_include_append(text, &textlen, str+last, inc_list[i].offset - last); + // write out line directive for the include + #ifndef STB_INCLUDE_LINE_NONE + #ifdef STB_INCLUDE_LINE_GLSL + if (textlen != 0) // GLSL #version must appear first, so don't put a #line at the top + #endif + { + strcpy(temp, "#line "); + stb_include_itoa(temp+6, 1); + strcat(temp, " "); + #ifdef STB_INCLUDE_LINE_GLSL + stb_include_itoa(temp+15, i+1); + #else + strcat(temp, "\""); + if (inc_list[i].filename == 0) + strcmp(temp, "INJECT"); + else + strcat(temp, inc_list[i].filename); + strcat(temp, "\""); + #endif + strcat(temp, "\n"); + text = stb_include_append(text, &textlen, temp, strlen(temp)); + } + #endif + if (inc_list[i].filename == 0) { + if (inject != 0) + text = stb_include_append(text, &textlen, inject, strlen(inject)); + } else { + char *inc; + strcpy(temp, path_to_includes); + strcat(temp, "/"); + strcat(temp, inc_list[i].filename); + inc = stb_include_file(temp, inject, path_to_includes, error); + if (inc == NULL) { + stb_include_free_includes(inc_list, num); + return NULL; + } + text = stb_include_append(text, &textlen, inc, strlen(inc)); + free(inc); + } + // write out line directive + #ifndef STB_INCLUDE_LINE_NONE + strcpy(temp, "\n#line "); + stb_include_itoa(temp+6, inc_list[i].next_line_after); + strcat(temp, " "); + #ifdef STB_INCLUDE_LINE_GLSL + stb_include_itoa(temp+15, 0); + #else + strcat(temp, filename != 0 ? filename : "source-file"); + #endif + text = stb_include_append(text, &textlen, temp, strlen(temp)); + // no newlines, because we kept the #include newlines, which will get appended next + #endif + last = inc_list[i].end; + } + text = stb_include_append(text, &textlen, str+last, source_len - last + 1); // append '\0' + stb_include_free_includes(inc_list, num); + return text; +} + +char *stb_include_strings(char **strs, int count, char *inject, char *path_to_includes, char *filename, char error[256]) +{ + char *text; + char *result; + int i; + size_t length=0; + for (i=0; i < count; ++i) + length += strlen(strs[i]); + text = (char *) malloc(length+1); + length = 0; + for (i=0; i < count; ++i) { + strcpy(text + length, strs[i]); + length += strlen(strs[i]); + } + result = stb_include_string(text, inject, path_to_includes, filename, error); + free(text); + return result; +} + +char *stb_include_file(char *filename, char *inject, char *path_to_includes, char error[256]) +{ + size_t len; + char *result; + char *text = stb_include_load_file(filename, &len); + if (text == NULL) { + strcpy(error, "Error: couldn't load '"); + strcat(error, filename); + strcat(error, "'"); + return 0; + } + result = stb_include_string(text, inject, path_to_includes, filename, error); + free(text); + return result; +} + +#if 0 // @TODO, GL_ARB_shader_language_include-style system that doesn't touch filesystem +char *stb_include_preloaded(char *str, char *inject, char *includes[][2], char error[256]) +{ + +} +#endif + +#endif // STB_INCLUDE_IMPLEMENTATION diff --git a/include/stb_leakcheck.h b/include/stb_leakcheck.h new file mode 100644 index 0000000..19ee6e7 --- /dev/null +++ b/include/stb_leakcheck.h @@ -0,0 +1,194 @@ +// stb_leakcheck.h - v0.6 - quick & dirty malloc leak-checking - public domain +// LICENSE +// +// See end of file. + +#ifdef STB_LEAKCHECK_IMPLEMENTATION +#undef STB_LEAKCHECK_IMPLEMENTATION // don't implement more than once + +// if we've already included leakcheck before, undefine the macros +#ifdef malloc +#undef malloc +#undef free +#undef realloc +#endif + +#ifndef STB_LEAKCHECK_OUTPUT_PIPE +#define STB_LEAKCHECK_OUTPUT_PIPE stdout +#endif + +#include +#include +#include +#include +#include +typedef struct malloc_info stb_leakcheck_malloc_info; + +struct malloc_info +{ + const char *file; + int line; + size_t size; + stb_leakcheck_malloc_info *next,*prev; +}; + +static stb_leakcheck_malloc_info *mi_head; + +void *stb_leakcheck_malloc(size_t sz, const char *file, int line) +{ + stb_leakcheck_malloc_info *mi = (stb_leakcheck_malloc_info *) malloc(sz + sizeof(*mi)); + if (mi == NULL) return mi; + mi->file = file; + mi->line = line; + mi->next = mi_head; + if (mi_head) + mi->next->prev = mi; + mi->prev = NULL; + mi->size = (int) sz; + mi_head = mi; + return mi+1; +} + +void stb_leakcheck_free(void *ptr) +{ + if (ptr != NULL) { + stb_leakcheck_malloc_info *mi = (stb_leakcheck_malloc_info *) ptr - 1; + mi->size = ~mi->size; + #ifndef STB_LEAKCHECK_SHOWALL + if (mi->prev == NULL) { + assert(mi_head == mi); + mi_head = mi->next; + } else + mi->prev->next = mi->next; + if (mi->next) + mi->next->prev = mi->prev; + free(mi); + #endif + } +} + +void *stb_leakcheck_realloc(void *ptr, size_t sz, const char *file, int line) +{ + if (ptr == NULL) { + return stb_leakcheck_malloc(sz, file, line); + } else if (sz == 0) { + stb_leakcheck_free(ptr); + return NULL; + } else { + stb_leakcheck_malloc_info *mi = (stb_leakcheck_malloc_info *) ptr - 1; + if (sz <= mi->size) + return ptr; + else { + #ifdef STB_LEAKCHECK_REALLOC_PRESERVE_MALLOC_FILELINE + void *q = stb_leakcheck_malloc(sz, mi->file, mi->line); + #else + void *q = stb_leakcheck_malloc(sz, file, line); + #endif + if (q) { + memcpy(q, ptr, mi->size); + stb_leakcheck_free(ptr); + } + return q; + } + } +} + +static void stblkck_internal_print(const char *reason, stb_leakcheck_malloc_info *mi) +{ +#if defined(_MSC_VER) && _MSC_VER < 1900 // 1900=VS 2015 + // Compilers that use the old MS C runtime library don't have %zd + // and the older ones don't even have %lld either... however, the old compilers + // without "long long" don't support 64-bit targets either, so here's the + // compromise: + #if _MSC_VER < 1400 // before VS 2005 + fprintf(STB_LEAKCHECK_OUTPUT_PIPE, "%s: %s (%4d): %8d bytes at %p\n", reason, mi->file, mi->line, (int)mi->size, (void*)(mi+1)); + #else + fprintf(STB_LEAKCHECK_OUTPUT_PIPE, "%s: %s (%4d): %16lld bytes at %p\n", reason, mi->file, mi->line, (long long)mi->size, (void*)(mi+1)); + #endif +#else + // Assume we have %zd on other targets. + #ifdef __MINGW32__ + __mingw_fprintf(STB_LEAKCHECK_OUTPUT_PIPE, "%s: %s (%4d): %zd bytes at %p\n", reason, mi->file, mi->line, mi->size, (void*)(mi+1)); + #else + fprintf(STB_LEAKCHECK_OUTPUT_PIPE, "%s: %s (%4d): %zd bytes at %p\n", reason, mi->file, mi->line, mi->size, (void*)(mi+1)); + #endif +#endif +} + +void stb_leakcheck_dumpmem(void) +{ + stb_leakcheck_malloc_info *mi = mi_head; + while (mi) { + if ((ptrdiff_t) mi->size >= 0) + stblkck_internal_print("LEAKED", mi); + mi = mi->next; + } + #ifdef STB_LEAKCHECK_SHOWALL + mi = mi_head; + while (mi) { + if ((ptrdiff_t) mi->size < 0) + stblkck_internal_print("FREED ", mi); + mi = mi->next; + } + #endif +} +#endif // STB_LEAKCHECK_IMPLEMENTATION + +#if !defined(INCLUDE_STB_LEAKCHECK_H) || !defined(malloc) +#define INCLUDE_STB_LEAKCHECK_H + +#include // we want to define the macros *after* stdlib to avoid a slew of errors + +#define malloc(sz) stb_leakcheck_malloc(sz, __FILE__, __LINE__) +#define free(p) stb_leakcheck_free(p) +#define realloc(p,sz) stb_leakcheck_realloc(p,sz, __FILE__, __LINE__) + +extern void * stb_leakcheck_malloc(size_t sz, const char *file, int line); +extern void * stb_leakcheck_realloc(void *ptr, size_t sz, const char *file, int line); +extern void stb_leakcheck_free(void *ptr); +extern void stb_leakcheck_dumpmem(void); + +#endif // INCLUDE_STB_LEAKCHECK_H + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/include/stb_sprintf.h b/include/stb_sprintf.h new file mode 100644 index 0000000..ca432a6 --- /dev/null +++ b/include/stb_sprintf.h @@ -0,0 +1,1906 @@ +// stb_sprintf - v1.10 - public domain snprintf() implementation +// originally by Jeff Roberts / RAD Game Tools, 2015/10/20 +// http://github.com/nothings/stb +// +// allowed types: sc uidBboXx p AaGgEef n +// lengths : hh h ll j z t I64 I32 I +// +// Contributors: +// Fabian "ryg" Giesen (reformatting) +// github:aganm (attribute format) +// +// Contributors (bugfixes): +// github:d26435 +// github:trex78 +// github:account-login +// Jari Komppa (SI suffixes) +// Rohit Nirmal +// Marcin Wojdyr +// Leonard Ritter +// Stefano Zanotti +// Adam Allison +// Arvid Gerstmann +// Markus Kolb +// +// LICENSE: +// +// See end of file for license information. + +#ifndef STB_SPRINTF_H_INCLUDE +#define STB_SPRINTF_H_INCLUDE + +/* +Single file sprintf replacement. + +Originally written by Jeff Roberts at RAD Game Tools - 2015/10/20. +Hereby placed in public domain. + +This is a full sprintf replacement that supports everything that +the C runtime sprintfs support, including float/double, 64-bit integers, +hex floats, field parameters (%*.*d stuff), length reads backs, etc. + +Why would you need this if sprintf already exists? Well, first off, +it's *much* faster (see below). It's also much smaller than the CRT +versions code-space-wise. We've also added some simple improvements +that are super handy (commas in thousands, callbacks at buffer full, +for example). Finally, the format strings for MSVC and GCC differ +for 64-bit integers (among other small things), so this lets you use +the same format strings in cross platform code. + +It uses the standard single file trick of being both the header file +and the source itself. If you just include it normally, you just get +the header file function definitions. To get the code, you include +it from a C or C++ file and define STB_SPRINTF_IMPLEMENTATION first. + +It only uses va_args macros from the C runtime to do it's work. It +does cast doubles to S64s and shifts and divides U64s, which does +drag in CRT code on most platforms. + +It compiles to roughly 8K with float support, and 4K without. +As a comparison, when using MSVC static libs, calling sprintf drags +in 16K. + +API: +==== +int stbsp_sprintf( char * buf, char const * fmt, ... ) +int stbsp_snprintf( char * buf, int count, char const * fmt, ... ) + Convert an arg list into a buffer. stbsp_snprintf always returns + a zero-terminated string (unlike regular snprintf). + +int stbsp_vsprintf( char * buf, char const * fmt, va_list va ) +int stbsp_vsnprintf( char * buf, int count, char const * fmt, va_list va ) + Convert a va_list arg list into a buffer. stbsp_vsnprintf always returns + a zero-terminated string (unlike regular snprintf). + +int stbsp_vsprintfcb( STBSP_SPRINTFCB * callback, void * user, char * buf, char const * fmt, va_list va ) + typedef char * STBSP_SPRINTFCB( char const * buf, void * user, int len ); + Convert into a buffer, calling back every STB_SPRINTF_MIN chars. + Your callback can then copy the chars out, print them or whatever. + This function is actually the workhorse for everything else. + The buffer you pass in must hold at least STB_SPRINTF_MIN characters. + // you return the next buffer to use or 0 to stop converting + +void stbsp_set_separators( char comma, char period ) + Set the comma and period characters to use. + +FLOATS/DOUBLES: +=============== +This code uses a internal float->ascii conversion method that uses +doubles with error correction (double-doubles, for ~105 bits of +precision). This conversion is round-trip perfect - that is, an atof +of the values output here will give you the bit-exact double back. + +One difference is that our insignificant digits will be different than +with MSVC or GCC (but they don't match each other either). We also +don't attempt to find the minimum length matching float (pre-MSVC15 +doesn't either). + +If you don't need float or doubles at all, define STB_SPRINTF_NOFLOAT +and you'll save 4K of code space. + +64-BIT INTS: +============ +This library also supports 64-bit integers and you can use MSVC style or +GCC style indicators (%I64d or %lld). It supports the C99 specifiers +for size_t and ptr_diff_t (%jd %zd) as well. + +EXTRAS: +======= +Like some GCCs, for integers and floats, you can use a ' (single quote) +specifier and commas will be inserted on the thousands: "%'d" on 12345 +would print 12,345. + +For integers and floats, you can use a "$" specifier and the number +will be converted to float and then divided to get kilo, mega, giga or +tera and then printed, so "%$d" 1000 is "1.0 k", "%$.2d" 2536000 is +"2.53 M", etc. For byte values, use two $:s, like "%$$d" to turn +2536000 to "2.42 Mi". If you prefer JEDEC suffixes to SI ones, use three +$:s: "%$$$d" -> "2.42 M". To remove the space between the number and the +suffix, add "_" specifier: "%_$d" -> "2.53M". + +In addition to octal and hexadecimal conversions, you can print +integers in binary: "%b" for 256 would print 100. + +PERFORMANCE vs MSVC 2008 32-/64-bit (GCC is even slower than MSVC): +=================================================================== +"%d" across all 32-bit ints (4.8x/4.0x faster than 32-/64-bit MSVC) +"%24d" across all 32-bit ints (4.5x/4.2x faster) +"%x" across all 32-bit ints (4.5x/3.8x faster) +"%08x" across all 32-bit ints (4.3x/3.8x faster) +"%f" across e-10 to e+10 floats (7.3x/6.0x faster) +"%e" across e-10 to e+10 floats (8.1x/6.0x faster) +"%g" across e-10 to e+10 floats (10.0x/7.1x faster) +"%f" for values near e-300 (7.9x/6.5x faster) +"%f" for values near e+300 (10.0x/9.1x faster) +"%e" for values near e-300 (10.1x/7.0x faster) +"%e" for values near e+300 (9.2x/6.0x faster) +"%.320f" for values near e-300 (12.6x/11.2x faster) +"%a" for random values (8.6x/4.3x faster) +"%I64d" for 64-bits with 32-bit values (4.8x/3.4x faster) +"%I64d" for 64-bits > 32-bit values (4.9x/5.5x faster) +"%s%s%s" for 64 char strings (7.1x/7.3x faster) +"...512 char string..." ( 35.0x/32.5x faster!) +*/ + +#if defined(__clang__) + #if defined(__has_feature) && defined(__has_attribute) + #if __has_feature(address_sanitizer) + #if __has_attribute(__no_sanitize__) + #define STBSP__ASAN __attribute__((__no_sanitize__("address"))) + #elif __has_attribute(__no_sanitize_address__) + #define STBSP__ASAN __attribute__((__no_sanitize_address__)) + #elif __has_attribute(__no_address_safety_analysis__) + #define STBSP__ASAN __attribute__((__no_address_safety_analysis__)) + #endif + #endif + #endif +#elif defined(__GNUC__) && (__GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8)) + #if defined(__SANITIZE_ADDRESS__) && __SANITIZE_ADDRESS__ + #define STBSP__ASAN __attribute__((__no_sanitize_address__)) + #endif +#endif + +#ifndef STBSP__ASAN +#define STBSP__ASAN +#endif + +#ifdef STB_SPRINTF_STATIC +#define STBSP__PUBLICDEC static +#define STBSP__PUBLICDEF static STBSP__ASAN +#else +#ifdef __cplusplus +#define STBSP__PUBLICDEC extern "C" +#define STBSP__PUBLICDEF extern "C" STBSP__ASAN +#else +#define STBSP__PUBLICDEC extern +#define STBSP__PUBLICDEF STBSP__ASAN +#endif +#endif + +#if defined(__has_attribute) + #if __has_attribute(format) + #define STBSP__ATTRIBUTE_FORMAT(fmt,va) __attribute__((format(printf,fmt,va))) + #endif +#endif + +#ifndef STBSP__ATTRIBUTE_FORMAT +#define STBSP__ATTRIBUTE_FORMAT(fmt,va) +#endif + +#ifdef _MSC_VER +#define STBSP__NOTUSED(v) (void)(v) +#else +#define STBSP__NOTUSED(v) (void)sizeof(v) +#endif + +#include // for va_arg(), va_list() +#include // size_t, ptrdiff_t + +#ifndef STB_SPRINTF_MIN +#define STB_SPRINTF_MIN 512 // how many characters per callback +#endif +typedef char *STBSP_SPRINTFCB(const char *buf, void *user, int len); + +#ifndef STB_SPRINTF_DECORATE +#define STB_SPRINTF_DECORATE(name) stbsp_##name // define this before including if you want to change the names +#endif + +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsprintf)(char *buf, char const *fmt, va_list va); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsnprintf)(char *buf, int count, char const *fmt, va_list va); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(sprintf)(char *buf, char const *fmt, ...) STBSP__ATTRIBUTE_FORMAT(2,3); +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(snprintf)(char *buf, int count, char const *fmt, ...) STBSP__ATTRIBUTE_FORMAT(3,4); + +STBSP__PUBLICDEC int STB_SPRINTF_DECORATE(vsprintfcb)(STBSP_SPRINTFCB *callback, void *user, char *buf, char const *fmt, va_list va); +STBSP__PUBLICDEC void STB_SPRINTF_DECORATE(set_separators)(char comma, char period); + +#endif // STB_SPRINTF_H_INCLUDE + +#ifdef STB_SPRINTF_IMPLEMENTATION + +#define stbsp__uint32 unsigned int +#define stbsp__int32 signed int + +#ifdef _MSC_VER +#define stbsp__uint64 unsigned __int64 +#define stbsp__int64 signed __int64 +#else +#define stbsp__uint64 unsigned long long +#define stbsp__int64 signed long long +#endif +#define stbsp__uint16 unsigned short + +#ifndef stbsp__uintptr +#if defined(__ppc64__) || defined(__powerpc64__) || defined(__aarch64__) || defined(_M_X64) || defined(__x86_64__) || defined(__x86_64) || defined(__s390x__) +#define stbsp__uintptr stbsp__uint64 +#else +#define stbsp__uintptr stbsp__uint32 +#endif +#endif + +#ifndef STB_SPRINTF_MSVC_MODE // used for MSVC2013 and earlier (MSVC2015 matches GCC) +#if defined(_MSC_VER) && (_MSC_VER < 1900) +#define STB_SPRINTF_MSVC_MODE +#endif +#endif + +#ifdef STB_SPRINTF_NOUNALIGNED // define this before inclusion to force stbsp_sprintf to always use aligned accesses +#define STBSP__UNALIGNED(code) +#else +#define STBSP__UNALIGNED(code) code +#endif + +#ifndef STB_SPRINTF_NOFLOAT +// internal float utility functions +static stbsp__int32 stbsp__real_to_str(char const **start, stbsp__uint32 *len, char *out, stbsp__int32 *decimal_pos, double value, stbsp__uint32 frac_digits); +static stbsp__int32 stbsp__real_to_parts(stbsp__int64 *bits, stbsp__int32 *expo, double value); +#define STBSP__SPECIAL 0x7000 +#endif + +static char stbsp__period = '.'; +static char stbsp__comma = ','; +static struct +{ + short temp; // force next field to be 2-byte aligned + char pair[201]; +} stbsp__digitpair = +{ + 0, + "00010203040506070809101112131415161718192021222324" + "25262728293031323334353637383940414243444546474849" + "50515253545556575859606162636465666768697071727374" + "75767778798081828384858687888990919293949596979899" +}; + +STBSP__PUBLICDEF void STB_SPRINTF_DECORATE(set_separators)(char pcomma, char pperiod) +{ + stbsp__period = pperiod; + stbsp__comma = pcomma; +} + +#define STBSP__LEFTJUST 1 +#define STBSP__LEADINGPLUS 2 +#define STBSP__LEADINGSPACE 4 +#define STBSP__LEADING_0X 8 +#define STBSP__LEADINGZERO 16 +#define STBSP__INTMAX 32 +#define STBSP__TRIPLET_COMMA 64 +#define STBSP__NEGATIVE 128 +#define STBSP__METRIC_SUFFIX 256 +#define STBSP__HALFWIDTH 512 +#define STBSP__METRIC_NOSPACE 1024 +#define STBSP__METRIC_1024 2048 +#define STBSP__METRIC_JEDEC 4096 + +static void stbsp__lead_sign(stbsp__uint32 fl, char *sign) +{ + sign[0] = 0; + if (fl & STBSP__NEGATIVE) { + sign[0] = 1; + sign[1] = '-'; + } else if (fl & STBSP__LEADINGSPACE) { + sign[0] = 1; + sign[1] = ' '; + } else if (fl & STBSP__LEADINGPLUS) { + sign[0] = 1; + sign[1] = '+'; + } +} + +static STBSP__ASAN stbsp__uint32 stbsp__strlen_limited(char const *s, stbsp__uint32 limit) +{ + char const * sn = s; + + // get up to 4-byte alignment + for (;;) { + if (((stbsp__uintptr)sn & 3) == 0) + break; + + if (!limit || *sn == 0) + return (stbsp__uint32)(sn - s); + + ++sn; + --limit; + } + + // scan over 4 bytes at a time to find terminating 0 + // this will intentionally scan up to 3 bytes past the end of buffers, + // but becase it works 4B aligned, it will never cross page boundaries + // (hence the STBSP__ASAN markup; the over-read here is intentional + // and harmless) + while (limit >= 4) { + stbsp__uint32 v = *(stbsp__uint32 *)sn; + // bit hack to find if there's a 0 byte in there + if ((v - 0x01010101) & (~v) & 0x80808080UL) + break; + + sn += 4; + limit -= 4; + } + + // handle the last few characters to find actual size + while (limit && *sn) { + ++sn; + --limit; + } + + return (stbsp__uint32)(sn - s); +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(vsprintfcb)(STBSP_SPRINTFCB *callback, void *user, char *buf, char const *fmt, va_list va) +{ + static char hex[] = "0123456789abcdefxp"; + static char hexu[] = "0123456789ABCDEFXP"; + char *bf; + char const *f; + int tlen = 0; + + bf = buf; + f = fmt; + for (;;) { + stbsp__int32 fw, pr, tz; + stbsp__uint32 fl; + + // macros for the callback buffer stuff + #define stbsp__chk_cb_bufL(bytes) \ + { \ + int len = (int)(bf - buf); \ + if ((len + (bytes)) >= STB_SPRINTF_MIN) { \ + tlen += len; \ + if (0 == (bf = buf = callback(buf, user, len))) \ + goto done; \ + } \ + } + #define stbsp__chk_cb_buf(bytes) \ + { \ + if (callback) { \ + stbsp__chk_cb_bufL(bytes); \ + } \ + } + #define stbsp__flush_cb() \ + { \ + stbsp__chk_cb_bufL(STB_SPRINTF_MIN - 1); \ + } // flush if there is even one byte in the buffer + #define stbsp__cb_buf_clamp(cl, v) \ + cl = v; \ + if (callback) { \ + int lg = STB_SPRINTF_MIN - (int)(bf - buf); \ + if (cl > lg) \ + cl = lg; \ + } + + // fast copy everything up to the next % (or end of string) + for (;;) { + while (((stbsp__uintptr)f) & 3) { + schk1: + if (f[0] == '%') + goto scandd; + schk2: + if (f[0] == 0) + goto endfmt; + stbsp__chk_cb_buf(1); + *bf++ = f[0]; + ++f; + } + for (;;) { + // Check if the next 4 bytes contain %(0x25) or end of string. + // Using the 'hasless' trick: + // https://graphics.stanford.edu/~seander/bithacks.html#HasLessInWord + stbsp__uint32 v, c; + v = *(stbsp__uint32 *)f; + c = (~v) & 0x80808080; + if (((v ^ 0x25252525) - 0x01010101) & c) + goto schk1; + if ((v - 0x01010101) & c) + goto schk2; + if (callback) + if ((STB_SPRINTF_MIN - (int)(bf - buf)) < 4) + goto schk1; + #ifdef STB_SPRINTF_NOUNALIGNED + if(((stbsp__uintptr)bf) & 3) { + bf[0] = f[0]; + bf[1] = f[1]; + bf[2] = f[2]; + bf[3] = f[3]; + } else + #endif + { + *(stbsp__uint32 *)bf = v; + } + bf += 4; + f += 4; + } + } + scandd: + + ++f; + + // ok, we have a percent, read the modifiers first + fw = 0; + pr = -1; + fl = 0; + tz = 0; + + // flags + for (;;) { + switch (f[0]) { + // if we have left justify + case '-': + fl |= STBSP__LEFTJUST; + ++f; + continue; + // if we have leading plus + case '+': + fl |= STBSP__LEADINGPLUS; + ++f; + continue; + // if we have leading space + case ' ': + fl |= STBSP__LEADINGSPACE; + ++f; + continue; + // if we have leading 0x + case '#': + fl |= STBSP__LEADING_0X; + ++f; + continue; + // if we have thousand commas + case '\'': + fl |= STBSP__TRIPLET_COMMA; + ++f; + continue; + // if we have kilo marker (none->kilo->kibi->jedec) + case '$': + if (fl & STBSP__METRIC_SUFFIX) { + if (fl & STBSP__METRIC_1024) { + fl |= STBSP__METRIC_JEDEC; + } else { + fl |= STBSP__METRIC_1024; + } + } else { + fl |= STBSP__METRIC_SUFFIX; + } + ++f; + continue; + // if we don't want space between metric suffix and number + case '_': + fl |= STBSP__METRIC_NOSPACE; + ++f; + continue; + // if we have leading zero + case '0': + fl |= STBSP__LEADINGZERO; + ++f; + goto flags_done; + default: goto flags_done; + } + } + flags_done: + + // get the field width + if (f[0] == '*') { + fw = va_arg(va, stbsp__uint32); + ++f; + } else { + while ((f[0] >= '0') && (f[0] <= '9')) { + fw = fw * 10 + f[0] - '0'; + f++; + } + } + // get the precision + if (f[0] == '.') { + ++f; + if (f[0] == '*') { + pr = va_arg(va, stbsp__uint32); + ++f; + } else { + pr = 0; + while ((f[0] >= '0') && (f[0] <= '9')) { + pr = pr * 10 + f[0] - '0'; + f++; + } + } + } + + // handle integer size overrides + switch (f[0]) { + // are we halfwidth? + case 'h': + fl |= STBSP__HALFWIDTH; + ++f; + if (f[0] == 'h') + ++f; // QUARTERWIDTH + break; + // are we 64-bit (unix style) + case 'l': + fl |= ((sizeof(long) == 8) ? STBSP__INTMAX : 0); + ++f; + if (f[0] == 'l') { + fl |= STBSP__INTMAX; + ++f; + } + break; + // are we 64-bit on intmax? (c99) + case 'j': + fl |= (sizeof(size_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + // are we 64-bit on size_t or ptrdiff_t? (c99) + case 'z': + fl |= (sizeof(ptrdiff_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + case 't': + fl |= (sizeof(ptrdiff_t) == 8) ? STBSP__INTMAX : 0; + ++f; + break; + // are we 64-bit (msft style) + case 'I': + if ((f[1] == '6') && (f[2] == '4')) { + fl |= STBSP__INTMAX; + f += 3; + } else if ((f[1] == '3') && (f[2] == '2')) { + f += 3; + } else { + fl |= ((sizeof(void *) == 8) ? STBSP__INTMAX : 0); + ++f; + } + break; + default: break; + } + + // handle each replacement + switch (f[0]) { + #define STBSP__NUMSZ 512 // big enough for e308 (with commas) or e-307 + char num[STBSP__NUMSZ]; + char lead[8]; + char tail[8]; + char *s; + char const *h; + stbsp__uint32 l, n, cs; + stbsp__uint64 n64; +#ifndef STB_SPRINTF_NOFLOAT + double fv; +#endif + stbsp__int32 dp; + char const *sn; + + case 's': + // get the string + s = va_arg(va, char *); + if (s == 0) + s = (char *)"null"; + // get the length, limited to desired precision + // always limit to ~0u chars since our counts are 32b + l = stbsp__strlen_limited(s, (pr >= 0) ? pr : ~0u); + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + // copy the string in + goto scopy; + + case 'c': // char + // get the character + s = num + STBSP__NUMSZ - 1; + *s = (char)va_arg(va, int); + l = 1; + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + goto scopy; + + case 'n': // weird write-bytes specifier + { + int *d = va_arg(va, int *); + *d = tlen + (int)(bf - buf); + } break; + +#ifdef STB_SPRINTF_NOFLOAT + case 'A': // float + case 'a': // hex float + case 'G': // float + case 'g': // float + case 'E': // float + case 'e': // float + case 'f': // float + va_arg(va, double); // eat it + s = (char *)"No float"; + l = 8; + lead[0] = 0; + tail[0] = 0; + pr = 0; + cs = 0; + STBSP__NOTUSED(dp); + goto scopy; +#else + case 'A': // hex float + case 'a': // hex float + h = (f[0] == 'A') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_parts((stbsp__int64 *)&n64, &dp, fv)) + fl |= STBSP__NEGATIVE; + + s = num + 64; + + stbsp__lead_sign(fl, lead); + + if (dp == -1023) + dp = (n64) ? -1022 : 0; + else + n64 |= (((stbsp__uint64)1) << 52); + n64 <<= (64 - 56); + if (pr < 15) + n64 += ((((stbsp__uint64)8) << 56) >> (pr * 4)); +// add leading chars + +#ifdef STB_SPRINTF_MSVC_MODE + *s++ = '0'; + *s++ = 'x'; +#else + lead[1 + lead[0]] = '0'; + lead[2 + lead[0]] = 'x'; + lead[0] += 2; +#endif + *s++ = h[(n64 >> 60) & 15]; + n64 <<= 4; + if (pr) + *s++ = stbsp__period; + sn = s; + + // print the bits + n = pr; + if (n > 13) + n = 13; + if (pr > (stbsp__int32)n) + tz = pr - n; + pr = 0; + while (n--) { + *s++ = h[(n64 >> 60) & 15]; + n64 <<= 4; + } + + // print the expo + tail[1] = h[17]; + if (dp < 0) { + tail[2] = '-'; + dp = -dp; + } else + tail[2] = '+'; + n = (dp >= 1000) ? 6 : ((dp >= 100) ? 5 : ((dp >= 10) ? 4 : 3)); + tail[0] = (char)n; + for (;;) { + tail[n] = '0' + dp % 10; + if (n <= 3) + break; + --n; + dp /= 10; + } + + dp = (int)(s - sn); + l = (int)(s - (num + 64)); + s = num + 64; + cs = 1 + (3 << 24); + goto scopy; + + case 'G': // float + case 'g': // float + h = (f[0] == 'G') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; + else if (pr == 0) + pr = 1; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, (pr - 1) | 0x80000000)) + fl |= STBSP__NEGATIVE; + + // clamp the precision and delete extra zeros after clamp + n = pr; + if (l > (stbsp__uint32)pr) + l = pr; + while ((l > 1) && (pr) && (sn[l - 1] == '0')) { + --pr; + --l; + } + + // should we use %e + if ((dp <= -4) || (dp > (stbsp__int32)n)) { + if (pr > (stbsp__int32)l) + pr = l - 1; + else if (pr) + --pr; // when using %e, there is one digit before the decimal + goto doexpfromg; + } + // this is the insane action to get the pr to match %g semantics for %f + if (dp > 0) { + pr = (dp < (stbsp__int32)l) ? l - dp : 0; + } else { + pr = -dp + ((pr > (stbsp__int32)l) ? (stbsp__int32) l : pr); + } + goto dofloatfromg; + + case 'E': // float + case 'e': // float + h = (f[0] == 'E') ? hexu : hex; + fv = va_arg(va, double); + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, pr | 0x80000000)) + fl |= STBSP__NEGATIVE; + doexpfromg: + tail[0] = 0; + stbsp__lead_sign(fl, lead); + if (dp == STBSP__SPECIAL) { + s = (char *)sn; + cs = 0; + pr = 0; + goto scopy; + } + s = num + 64; + // handle leading chars + *s++ = sn[0]; + + if (pr) + *s++ = stbsp__period; + + // handle after decimal + if ((l - 1) > (stbsp__uint32)pr) + l = pr + 1; + for (n = 1; n < l; n++) + *s++ = sn[n]; + // trailing zeros + tz = pr - (l - 1); + pr = 0; + // dump expo + tail[1] = h[0xe]; + dp -= 1; + if (dp < 0) { + tail[2] = '-'; + dp = -dp; + } else + tail[2] = '+'; +#ifdef STB_SPRINTF_MSVC_MODE + n = 5; +#else + n = (dp >= 100) ? 5 : 4; +#endif + tail[0] = (char)n; + for (;;) { + tail[n] = '0' + dp % 10; + if (n <= 3) + break; + --n; + dp /= 10; + } + cs = 1 + (3 << 24); // how many tens + goto flt_lead; + + case 'f': // float + fv = va_arg(va, double); + doafloat: + // do kilos + if (fl & STBSP__METRIC_SUFFIX) { + double divisor; + divisor = 1000.0f; + if (fl & STBSP__METRIC_1024) + divisor = 1024.0; + while (fl < 0x4000000) { + if ((fv < divisor) && (fv > -divisor)) + break; + fv /= divisor; + fl += 0x1000000; + } + } + if (pr == -1) + pr = 6; // default is 6 + // read the double into a string + if (stbsp__real_to_str(&sn, &l, num, &dp, fv, pr)) + fl |= STBSP__NEGATIVE; + dofloatfromg: + tail[0] = 0; + stbsp__lead_sign(fl, lead); + if (dp == STBSP__SPECIAL) { + s = (char *)sn; + cs = 0; + pr = 0; + goto scopy; + } + s = num + 64; + + // handle the three decimal varieties + if (dp <= 0) { + stbsp__int32 i; + // handle 0.000*000xxxx + *s++ = '0'; + if (pr) + *s++ = stbsp__period; + n = -dp; + if ((stbsp__int32)n > pr) + n = pr; + i = n; + while (i) { + if ((((stbsp__uintptr)s) & 3) == 0) + break; + *s++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)s = 0x30303030; + s += 4; + i -= 4; + } + while (i) { + *s++ = '0'; + --i; + } + if ((stbsp__int32)(l + n) > pr) + l = pr - n; + i = l; + while (i) { + *s++ = *sn++; + --i; + } + tz = pr - (n + l); + cs = 1 + (3 << 24); // how many tens did we write (for commas below) + } else { + cs = (fl & STBSP__TRIPLET_COMMA) ? ((600 - (stbsp__uint32)dp) % 3) : 0; + if ((stbsp__uint32)dp >= l) { + // handle xxxx000*000.0 + n = 0; + for (;;) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = sn[n]; + ++n; + if (n >= l) + break; + } + } + if (n < (stbsp__uint32)dp) { + n = dp - n; + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + while (n) { + if ((((stbsp__uintptr)s) & 3) == 0) + break; + *s++ = '0'; + --n; + } + while (n >= 4) { + *(stbsp__uint32 *)s = 0x30303030; + s += 4; + n -= 4; + } + } + while (n) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = '0'; + --n; + } + } + } + cs = (int)(s - (num + 64)) + (3 << 24); // cs is how many tens + if (pr) { + *s++ = stbsp__period; + tz = pr; + } + } else { + // handle xxxxx.xxxx000*000 + n = 0; + for (;;) { + if ((fl & STBSP__TRIPLET_COMMA) && (++cs == 4)) { + cs = 0; + *s++ = stbsp__comma; + } else { + *s++ = sn[n]; + ++n; + if (n >= (stbsp__uint32)dp) + break; + } + } + cs = (int)(s - (num + 64)) + (3 << 24); // cs is how many tens + if (pr) + *s++ = stbsp__period; + if ((l - dp) > (stbsp__uint32)pr) + l = pr + dp; + while (n < l) { + *s++ = sn[n]; + ++n; + } + tz = pr - (l - dp); + } + } + pr = 0; + + // handle k,m,g,t + if (fl & STBSP__METRIC_SUFFIX) { + char idx; + idx = 1; + if (fl & STBSP__METRIC_NOSPACE) + idx = 0; + tail[0] = idx; + tail[1] = ' '; + { + if (fl >> 24) { // SI kilo is 'k', JEDEC and SI kibits are 'K'. + if (fl & STBSP__METRIC_1024) + tail[idx + 1] = "_KMGT"[fl >> 24]; + else + tail[idx + 1] = "_kMGT"[fl >> 24]; + idx++; + // If printing kibits and not in jedec, add the 'i'. + if (fl & STBSP__METRIC_1024 && !(fl & STBSP__METRIC_JEDEC)) { + tail[idx + 1] = 'i'; + idx++; + } + tail[0] = idx; + } + } + }; + + flt_lead: + // get the length that we copied + l = (stbsp__uint32)(s - (num + 64)); + s = num + 64; + goto scopy; +#endif + + case 'B': // upper binary + case 'b': // lower binary + h = (f[0] == 'B') ? hexu : hex; + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 2; + lead[1] = '0'; + lead[2] = h[0xb]; + } + l = (8 << 4) | (1 << 8); + goto radixnum; + + case 'o': // octal + h = hexu; + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 1; + lead[1] = '0'; + } + l = (3 << 4) | (3 << 8); + goto radixnum; + + case 'p': // pointer + fl |= (sizeof(void *) == 8) ? STBSP__INTMAX : 0; + pr = sizeof(void *) * 2; + fl &= ~STBSP__LEADINGZERO; // 'p' only prints the pointer with zeros + // fall through - to X + + case 'X': // upper hex + case 'x': // lower hex + h = (f[0] == 'X') ? hexu : hex; + l = (4 << 4) | (4 << 8); + lead[0] = 0; + if (fl & STBSP__LEADING_0X) { + lead[0] = 2; + lead[1] = '0'; + lead[2] = h[16]; + } + radixnum: + // get the number + if (fl & STBSP__INTMAX) + n64 = va_arg(va, stbsp__uint64); + else + n64 = va_arg(va, stbsp__uint32); + + s = num + STBSP__NUMSZ; + dp = 0; + // clear tail, and clear leading if value is zero + tail[0] = 0; + if (n64 == 0) { + lead[0] = 0; + if (pr == 0) { + l = 0; + cs = 0; + goto scopy; + } + } + // convert to string + for (;;) { + *--s = h[n64 & ((1 << (l >> 8)) - 1)]; + n64 >>= (l >> 8); + if (!((n64) || ((stbsp__int32)((num + STBSP__NUMSZ) - s) < pr))) + break; + if (fl & STBSP__TRIPLET_COMMA) { + ++l; + if ((l & 15) == ((l >> 4) & 15)) { + l &= ~15; + *--s = stbsp__comma; + } + } + }; + // get the tens and the comma pos + cs = (stbsp__uint32)((num + STBSP__NUMSZ) - s) + ((((l >> 4) & 15)) << 24); + // get the length that we copied + l = (stbsp__uint32)((num + STBSP__NUMSZ) - s); + // copy it + goto scopy; + + case 'u': // unsigned + case 'i': + case 'd': // integer + // get the integer and abs it + if (fl & STBSP__INTMAX) { + stbsp__int64 i64 = va_arg(va, stbsp__int64); + n64 = (stbsp__uint64)i64; + if ((f[0] != 'u') && (i64 < 0)) { + n64 = (stbsp__uint64)-i64; + fl |= STBSP__NEGATIVE; + } + } else { + stbsp__int32 i = va_arg(va, stbsp__int32); + n64 = (stbsp__uint32)i; + if ((f[0] != 'u') && (i < 0)) { + n64 = (stbsp__uint32)-i; + fl |= STBSP__NEGATIVE; + } + } + +#ifndef STB_SPRINTF_NOFLOAT + if (fl & STBSP__METRIC_SUFFIX) { + if (n64 < 1024) + pr = 0; + else if (pr == -1) + pr = 1; + fv = (double)(stbsp__int64)n64; + goto doafloat; + } +#endif + + // convert to string + s = num + STBSP__NUMSZ; + l = 0; + + for (;;) { + // do in 32-bit chunks (avoid lots of 64-bit divides even with constant denominators) + char *o = s - 8; + if (n64 >= 100000000) { + n = (stbsp__uint32)(n64 % 100000000); + n64 /= 100000000; + } else { + n = (stbsp__uint32)n64; + n64 = 0; + } + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + do { + s -= 2; + *(stbsp__uint16 *)s = *(stbsp__uint16 *)&stbsp__digitpair.pair[(n % 100) * 2]; + n /= 100; + } while (n); + } + while (n) { + if ((fl & STBSP__TRIPLET_COMMA) && (l++ == 3)) { + l = 0; + *--s = stbsp__comma; + --o; + } else { + *--s = (char)(n % 10) + '0'; + n /= 10; + } + } + if (n64 == 0) { + if ((s[0] == '0') && (s != (num + STBSP__NUMSZ))) + ++s; + break; + } + while (s != o) + if ((fl & STBSP__TRIPLET_COMMA) && (l++ == 3)) { + l = 0; + *--s = stbsp__comma; + --o; + } else { + *--s = '0'; + } + } + + tail[0] = 0; + stbsp__lead_sign(fl, lead); + + // get the length that we copied + l = (stbsp__uint32)((num + STBSP__NUMSZ) - s); + if (l == 0) { + *--s = '0'; + l = 1; + } + cs = l + (3 << 24); + if (pr < 0) + pr = 0; + + scopy: + // get fw=leading/trailing space, pr=leading zeros + if (pr < (stbsp__int32)l) + pr = l; + n = pr + lead[0] + tail[0] + tz; + if (fw < (stbsp__int32)n) + fw = n; + fw -= n; + pr -= l; + + // handle right justify and leading zeros + if ((fl & STBSP__LEFTJUST) == 0) { + if (fl & STBSP__LEADINGZERO) // if leading zeros, everything is in pr + { + pr = (fw > pr) ? fw : pr; + fw = 0; + } else { + fl &= ~STBSP__TRIPLET_COMMA; // if no leading zeros, then no commas + } + } + + // copy the spaces and/or zeros + if (fw + pr) { + stbsp__int32 i; + stbsp__uint32 c; + + // copy leading spaces (or when doing %8.4d stuff) + if ((fl & STBSP__LEFTJUST) == 0) + while (fw > 0) { + stbsp__cb_buf_clamp(i, fw); + fw -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = ' '; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x20202020; + bf += 4; + i -= 4; + } + while (i) { + *bf++ = ' '; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy leader + sn = lead + 1; + while (lead[0]) { + stbsp__cb_buf_clamp(i, lead[0]); + lead[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy leading zeros + c = cs >> 24; + cs &= 0xffffff; + cs = (fl & STBSP__TRIPLET_COMMA) ? ((stbsp__uint32)(c - ((pr + cs) % (c + 1)))) : 0; + while (pr > 0) { + stbsp__cb_buf_clamp(i, pr); + pr -= i; + if ((fl & STBSP__TRIPLET_COMMA) == 0) { + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x30303030; + bf += 4; + i -= 4; + } + } + while (i) { + if ((fl & STBSP__TRIPLET_COMMA) && (cs++ == c)) { + cs = 0; + *bf++ = stbsp__comma; + } else + *bf++ = '0'; + --i; + } + stbsp__chk_cb_buf(1); + } + } + + // copy leader if there is still one + sn = lead + 1; + while (lead[0]) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, lead[0]); + lead[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy the string + n = l; + while (n) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, n); + n -= i; + STBSP__UNALIGNED(while (i >= 4) { + *(stbsp__uint32 volatile *)bf = *(stbsp__uint32 volatile *)s; + bf += 4; + s += 4; + i -= 4; + }) + while (i) { + *bf++ = *s++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy trailing zeros + while (tz) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, tz); + tz -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = '0'; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x30303030; + bf += 4; + i -= 4; + } + while (i) { + *bf++ = '0'; + --i; + } + stbsp__chk_cb_buf(1); + } + + // copy tail if there is one + sn = tail + 1; + while (tail[0]) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, tail[0]); + tail[0] -= (char)i; + while (i) { + *bf++ = *sn++; + --i; + } + stbsp__chk_cb_buf(1); + } + + // handle the left justify + if (fl & STBSP__LEFTJUST) + if (fw > 0) { + while (fw) { + stbsp__int32 i; + stbsp__cb_buf_clamp(i, fw); + fw -= i; + while (i) { + if ((((stbsp__uintptr)bf) & 3) == 0) + break; + *bf++ = ' '; + --i; + } + while (i >= 4) { + *(stbsp__uint32 *)bf = 0x20202020; + bf += 4; + i -= 4; + } + while (i--) + *bf++ = ' '; + stbsp__chk_cb_buf(1); + } + } + break; + + default: // unknown, just copy code + s = num + STBSP__NUMSZ - 1; + *s = f[0]; + l = 1; + fw = fl = 0; + lead[0] = 0; + tail[0] = 0; + pr = 0; + dp = 0; + cs = 0; + goto scopy; + } + ++f; + } +endfmt: + + if (!callback) + *bf = 0; + else + stbsp__flush_cb(); + +done: + return tlen + (int)(bf - buf); +} + +// cleanup +#undef STBSP__LEFTJUST +#undef STBSP__LEADINGPLUS +#undef STBSP__LEADINGSPACE +#undef STBSP__LEADING_0X +#undef STBSP__LEADINGZERO +#undef STBSP__INTMAX +#undef STBSP__TRIPLET_COMMA +#undef STBSP__NEGATIVE +#undef STBSP__METRIC_SUFFIX +#undef STBSP__NUMSZ +#undef stbsp__chk_cb_bufL +#undef stbsp__chk_cb_buf +#undef stbsp__flush_cb +#undef stbsp__cb_buf_clamp + +// ============================================================================ +// wrapper functions + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(sprintf)(char *buf, char const *fmt, ...) +{ + int result; + va_list va; + va_start(va, fmt); + result = STB_SPRINTF_DECORATE(vsprintfcb)(0, 0, buf, fmt, va); + va_end(va); + return result; +} + +typedef struct stbsp__context { + char *buf; + int count; + int length; + char tmp[STB_SPRINTF_MIN]; +} stbsp__context; + +static char *stbsp__clamp_callback(const char *buf, void *user, int len) +{ + stbsp__context *c = (stbsp__context *)user; + c->length += len; + + if (len > c->count) + len = c->count; + + if (len) { + if (buf != c->buf) { + const char *s, *se; + char *d; + d = c->buf; + s = buf; + se = buf + len; + do { + *d++ = *s++; + } while (s < se); + } + c->buf += len; + c->count -= len; + } + + if (c->count <= 0) + return c->tmp; + return (c->count >= STB_SPRINTF_MIN) ? c->buf : c->tmp; // go direct into buffer if you can +} + +static char * stbsp__count_clamp_callback( const char * buf, void * user, int len ) +{ + stbsp__context * c = (stbsp__context*)user; + (void) sizeof(buf); + + c->length += len; + return c->tmp; // go direct into buffer if you can +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE( vsnprintf )( char * buf, int count, char const * fmt, va_list va ) +{ + stbsp__context c; + + if ( (count == 0) && !buf ) + { + c.length = 0; + + STB_SPRINTF_DECORATE( vsprintfcb )( stbsp__count_clamp_callback, &c, c.tmp, fmt, va ); + } + else + { + int l; + + c.buf = buf; + c.count = count; + c.length = 0; + + STB_SPRINTF_DECORATE( vsprintfcb )( stbsp__clamp_callback, &c, stbsp__clamp_callback(0,&c,0), fmt, va ); + + // zero-terminate + l = (int)( c.buf - buf ); + if ( l >= count ) // should never be greater, only equal (or less) than count + l = count - 1; + buf[l] = 0; + } + + return c.length; +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(snprintf)(char *buf, int count, char const *fmt, ...) +{ + int result; + va_list va; + va_start(va, fmt); + + result = STB_SPRINTF_DECORATE(vsnprintf)(buf, count, fmt, va); + va_end(va); + + return result; +} + +STBSP__PUBLICDEF int STB_SPRINTF_DECORATE(vsprintf)(char *buf, char const *fmt, va_list va) +{ + return STB_SPRINTF_DECORATE(vsprintfcb)(0, 0, buf, fmt, va); +} + +// ======================================================================= +// low level float utility functions + +#ifndef STB_SPRINTF_NOFLOAT + +// copies d to bits w/ strict aliasing (this compiles to nothing on /Ox) +#define STBSP__COPYFP(dest, src) \ + { \ + int cn; \ + for (cn = 0; cn < 8; cn++) \ + ((char *)&dest)[cn] = ((char *)&src)[cn]; \ + } + +// get float info +static stbsp__int32 stbsp__real_to_parts(stbsp__int64 *bits, stbsp__int32 *expo, double value) +{ + double d; + stbsp__int64 b = 0; + + // load value and round at the frac_digits + d = value; + + STBSP__COPYFP(b, d); + + *bits = b & ((((stbsp__uint64)1) << 52) - 1); + *expo = (stbsp__int32)(((b >> 52) & 2047) - 1023); + + return (stbsp__int32)((stbsp__uint64) b >> 63); +} + +static double const stbsp__bot[23] = { + 1e+000, 1e+001, 1e+002, 1e+003, 1e+004, 1e+005, 1e+006, 1e+007, 1e+008, 1e+009, 1e+010, 1e+011, + 1e+012, 1e+013, 1e+014, 1e+015, 1e+016, 1e+017, 1e+018, 1e+019, 1e+020, 1e+021, 1e+022 +}; +static double const stbsp__negbot[22] = { + 1e-001, 1e-002, 1e-003, 1e-004, 1e-005, 1e-006, 1e-007, 1e-008, 1e-009, 1e-010, 1e-011, + 1e-012, 1e-013, 1e-014, 1e-015, 1e-016, 1e-017, 1e-018, 1e-019, 1e-020, 1e-021, 1e-022 +}; +static double const stbsp__negboterr[22] = { + -5.551115123125783e-018, -2.0816681711721684e-019, -2.0816681711721686e-020, -4.7921736023859299e-021, -8.1803053914031305e-022, 4.5251888174113741e-023, + 4.5251888174113739e-024, -2.0922560830128471e-025, -6.2281591457779853e-026, -3.6432197315497743e-027, 6.0503030718060191e-028, 2.0113352370744385e-029, + -3.0373745563400371e-030, 1.1806906454401013e-032, -7.7705399876661076e-032, 2.0902213275965398e-033, -7.1542424054621921e-034, -7.1542424054621926e-035, + 2.4754073164739869e-036, 5.4846728545790429e-037, 9.2462547772103625e-038, -4.8596774326570872e-039 +}; +static double const stbsp__top[13] = { + 1e+023, 1e+046, 1e+069, 1e+092, 1e+115, 1e+138, 1e+161, 1e+184, 1e+207, 1e+230, 1e+253, 1e+276, 1e+299 +}; +static double const stbsp__negtop[13] = { + 1e-023, 1e-046, 1e-069, 1e-092, 1e-115, 1e-138, 1e-161, 1e-184, 1e-207, 1e-230, 1e-253, 1e-276, 1e-299 +}; +static double const stbsp__toperr[13] = { + 8388608, + 6.8601809640529717e+028, + -7.253143638152921e+052, + -4.3377296974619174e+075, + -1.5559416129466825e+098, + -3.2841562489204913e+121, + -3.7745893248228135e+144, + -1.7356668416969134e+167, + -3.8893577551088374e+190, + -9.9566444326005119e+213, + 6.3641293062232429e+236, + -5.2069140800249813e+259, + -5.2504760255204387e+282 +}; +static double const stbsp__negtoperr[13] = { + 3.9565301985100693e-040, -2.299904345391321e-063, 3.6506201437945798e-086, 1.1875228833981544e-109, + -5.0644902316928607e-132, -6.7156837247865426e-155, -2.812077463003139e-178, -5.7778912386589953e-201, + 7.4997100559334532e-224, -4.6439668915134491e-247, -6.3691100762962136e-270, -9.436808465446358e-293, + 8.0970921678014997e-317 +}; + +#if defined(_MSC_VER) && (_MSC_VER <= 1200) +static stbsp__uint64 const stbsp__powten[20] = { + 1, + 10, + 100, + 1000, + 10000, + 100000, + 1000000, + 10000000, + 100000000, + 1000000000, + 10000000000, + 100000000000, + 1000000000000, + 10000000000000, + 100000000000000, + 1000000000000000, + 10000000000000000, + 100000000000000000, + 1000000000000000000, + 10000000000000000000U +}; +#define stbsp__tento19th ((stbsp__uint64)1000000000000000000) +#else +static stbsp__uint64 const stbsp__powten[20] = { + 1, + 10, + 100, + 1000, + 10000, + 100000, + 1000000, + 10000000, + 100000000, + 1000000000, + 10000000000ULL, + 100000000000ULL, + 1000000000000ULL, + 10000000000000ULL, + 100000000000000ULL, + 1000000000000000ULL, + 10000000000000000ULL, + 100000000000000000ULL, + 1000000000000000000ULL, + 10000000000000000000ULL +}; +#define stbsp__tento19th (1000000000000000000ULL) +#endif + +#define stbsp__ddmulthi(oh, ol, xh, yh) \ + { \ + double ahi = 0, alo, bhi = 0, blo; \ + stbsp__int64 bt; \ + oh = xh * yh; \ + STBSP__COPYFP(bt, xh); \ + bt &= ((~(stbsp__uint64)0) << 27); \ + STBSP__COPYFP(ahi, bt); \ + alo = xh - ahi; \ + STBSP__COPYFP(bt, yh); \ + bt &= ((~(stbsp__uint64)0) << 27); \ + STBSP__COPYFP(bhi, bt); \ + blo = yh - bhi; \ + ol = ((ahi * bhi - oh) + ahi * blo + alo * bhi) + alo * blo; \ + } + +#define stbsp__ddtoS64(ob, xh, xl) \ + { \ + double ahi = 0, alo, vh, t; \ + ob = (stbsp__int64)xh; \ + vh = (double)ob; \ + ahi = (xh - vh); \ + t = (ahi - xh); \ + alo = (xh - (ahi - t)) - (vh + t); \ + ob += (stbsp__int64)(ahi + alo + xl); \ + } + +#define stbsp__ddrenorm(oh, ol) \ + { \ + double s; \ + s = oh + ol; \ + ol = ol - (s - oh); \ + oh = s; \ + } + +#define stbsp__ddmultlo(oh, ol, xh, xl, yh, yl) ol = ol + (xh * yl + xl * yh); + +#define stbsp__ddmultlos(oh, ol, xh, yl) ol = ol + (xh * yl); + +static void stbsp__raise_to_power10(double *ohi, double *olo, double d, stbsp__int32 power) // power can be -323 to +350 +{ + double ph, pl; + if ((power >= 0) && (power <= 22)) { + stbsp__ddmulthi(ph, pl, d, stbsp__bot[power]); + } else { + stbsp__int32 e, et, eb; + double p2h, p2l; + + e = power; + if (power < 0) + e = -e; + et = (e * 0x2c9) >> 14; /* %23 */ + if (et > 13) + et = 13; + eb = e - (et * 23); + + ph = d; + pl = 0.0; + if (power < 0) { + if (eb) { + --eb; + stbsp__ddmulthi(ph, pl, d, stbsp__negbot[eb]); + stbsp__ddmultlos(ph, pl, d, stbsp__negboterr[eb]); + } + if (et) { + stbsp__ddrenorm(ph, pl); + --et; + stbsp__ddmulthi(p2h, p2l, ph, stbsp__negtop[et]); + stbsp__ddmultlo(p2h, p2l, ph, pl, stbsp__negtop[et], stbsp__negtoperr[et]); + ph = p2h; + pl = p2l; + } + } else { + if (eb) { + e = eb; + if (eb > 22) + eb = 22; + e -= eb; + stbsp__ddmulthi(ph, pl, d, stbsp__bot[eb]); + if (e) { + stbsp__ddrenorm(ph, pl); + stbsp__ddmulthi(p2h, p2l, ph, stbsp__bot[e]); + stbsp__ddmultlos(p2h, p2l, stbsp__bot[e], pl); + ph = p2h; + pl = p2l; + } + } + if (et) { + stbsp__ddrenorm(ph, pl); + --et; + stbsp__ddmulthi(p2h, p2l, ph, stbsp__top[et]); + stbsp__ddmultlo(p2h, p2l, ph, pl, stbsp__top[et], stbsp__toperr[et]); + ph = p2h; + pl = p2l; + } + } + } + stbsp__ddrenorm(ph, pl); + *ohi = ph; + *olo = pl; +} + +// given a float value, returns the significant bits in bits, and the position of the +// decimal point in decimal_pos. +/-INF and NAN are specified by special values +// returned in the decimal_pos parameter. +// frac_digits is absolute normally, but if you want from first significant digits (got %g and %e), or in 0x80000000 +static stbsp__int32 stbsp__real_to_str(char const **start, stbsp__uint32 *len, char *out, stbsp__int32 *decimal_pos, double value, stbsp__uint32 frac_digits) +{ + double d; + stbsp__int64 bits = 0; + stbsp__int32 expo, e, ng, tens; + + d = value; + STBSP__COPYFP(bits, d); + expo = (stbsp__int32)((bits >> 52) & 2047); + ng = (stbsp__int32)((stbsp__uint64) bits >> 63); + if (ng) + d = -d; + + if (expo == 2047) // is nan or inf? + { + *start = (bits & ((((stbsp__uint64)1) << 52) - 1)) ? "NaN" : "Inf"; + *decimal_pos = STBSP__SPECIAL; + *len = 3; + return ng; + } + + if (expo == 0) // is zero or denormal + { + if (((stbsp__uint64) bits << 1) == 0) // do zero + { + *decimal_pos = 1; + *start = out; + out[0] = '0'; + *len = 1; + return ng; + } + // find the right expo for denormals + { + stbsp__int64 v = ((stbsp__uint64)1) << 51; + while ((bits & v) == 0) { + --expo; + v >>= 1; + } + } + } + + // find the decimal exponent as well as the decimal bits of the value + { + double ph, pl; + + // log10 estimate - very specifically tweaked to hit or undershoot by no more than 1 of log10 of all expos 1..2046 + tens = expo - 1023; + tens = (tens < 0) ? ((tens * 617) / 2048) : (((tens * 1233) / 4096) + 1); + + // move the significant bits into position and stick them into an int + stbsp__raise_to_power10(&ph, &pl, d, 18 - tens); + + // get full as much precision from double-double as possible + stbsp__ddtoS64(bits, ph, pl); + + // check if we undershot + if (((stbsp__uint64)bits) >= stbsp__tento19th) + ++tens; + } + + // now do the rounding in integer land + frac_digits = (frac_digits & 0x80000000) ? ((frac_digits & 0x7ffffff) + 1) : (tens + frac_digits); + if ((frac_digits < 24)) { + stbsp__uint32 dg = 1; + if ((stbsp__uint64)bits >= stbsp__powten[9]) + dg = 10; + while ((stbsp__uint64)bits >= stbsp__powten[dg]) { + ++dg; + if (dg == 20) + goto noround; + } + if (frac_digits < dg) { + stbsp__uint64 r; + // add 0.5 at the right position and round + e = dg - frac_digits; + if ((stbsp__uint32)e >= 24) + goto noround; + r = stbsp__powten[e]; + bits = bits + (r / 2); + if ((stbsp__uint64)bits >= stbsp__powten[dg]) + ++tens; + bits /= r; + } + noround:; + } + + // kill long trailing runs of zeros + if (bits) { + stbsp__uint32 n; + for (;;) { + if (bits <= 0xffffffff) + break; + if (bits % 1000) + goto donez; + bits /= 1000; + } + n = (stbsp__uint32)bits; + while ((n % 1000) == 0) + n /= 1000; + bits = n; + donez:; + } + + // convert to string + out += 64; + e = 0; + for (;;) { + stbsp__uint32 n; + char *o = out - 8; + // do the conversion in chunks of U32s (avoid most 64-bit divides, worth it, constant denomiators be damned) + if (bits >= 100000000) { + n = (stbsp__uint32)(bits % 100000000); + bits /= 100000000; + } else { + n = (stbsp__uint32)bits; + bits = 0; + } + while (n) { + out -= 2; + *(stbsp__uint16 *)out = *(stbsp__uint16 *)&stbsp__digitpair.pair[(n % 100) * 2]; + n /= 100; + e += 2; + } + if (bits == 0) { + if ((e) && (out[0] == '0')) { + ++out; + --e; + } + break; + } + while (out != o) { + *--out = '0'; + ++e; + } + } + + *decimal_pos = tens; + *start = out; + *len = e; + return ng; +} + +#undef stbsp__ddmulthi +#undef stbsp__ddrenorm +#undef stbsp__ddmultlo +#undef stbsp__ddmultlos +#undef STBSP__SPECIAL +#undef STBSP__COPYFP + +#endif // STB_SPRINTF_NOFLOAT + +// clean up +#undef stbsp__uint16 +#undef stbsp__uint32 +#undef stbsp__int32 +#undef stbsp__uint64 +#undef stbsp__int64 +#undef STBSP__UNALIGNED + +#endif // STB_SPRINTF_IMPLEMENTATION + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ diff --git a/obj/main.o b/obj/main.o new file mode 100644 index 0000000000000000000000000000000000000000..9289905bd9d556507f1ec59dd8e0ec746441aec6 GIT binary patch literal 4689 zcmbtXYitzP6+U-n?(VF|kF^)54KyByN1(yG28@HDP8Q6`i zK@9>CWfP>;%2iycZGKQG?T<>WqO?`>gFb#hAUsr6A`(hmDUlorR3lLdMUm>NQRq2$ zX4bPcQh)SHbMJS~`R=*rp1JqVoy5gly=Lum1KJ=es0k|6vI?QSyt2Bk%E;wpl-yLrMBX#c6^A|a$a zhlZQun?e2Gk+;oSGC(O0LwNt>5u^I>m@%8#IAPSrb)&Xm)UFvfBTM8G+eg&ij64L) zRXy_1c+03RGNTq*u!Sof#??i|xVoS$iXL098?y&CqMyk8&B$G$nq4iLdIYssEmke^ zi2BFx81?Z-Mm;h=R=Yh`yYokUDd3Q5k-PQC-Rc!)yw#*f*Gp(H;8mrxM zI-&T;g_n%lC&sL)e*fiblSb`6PIaO-?l)?a-6D?ac|W_*W6W~NHSXodSRQ;yntG($ zs9sTxx(MpSuO2_f!5?8(=EaO|My%P~j4TLwc6{ELT}?ZT!f;20AwT$`rJpyRjbK=j z1>`KCcT8p@DT{|mS07)(nUQ^Qymr-cW7Ihf@fZ^}xhQ(FJ2%8Qi^go*T5?@`---si zHC)bOf@V2IgxN*LAIORG3YLuO`6cLhBZc#&fSq2KiDXo-PrQ8X%{PmX%j?N>#?(s% zGoDVS%>?Mamc1L`+@{jG7GRn(smY#SNCRwwsWMl8q{xgGC6)(PBig4h#o`FUg~x%Z zWNJP1tx0`!cv#=QE4U*#p!a9Xm7F<}GIM4z9Y--%j8Biu>>lbL8r1t!sl=4t|81m| z%88Nq@UZ32Ue97E8pJ6Q^TelPMQBN*##>6CHKS#aHf(zXMM?E$^0|~Qs!FbuPUXym zp3Wgb<;xbKki4boe6g%I(Y$7{n9GyHSK2-zhYBZhi;aZ~(OdC0^^=&|kVSNoswDE2 zsf>v(lw>9!LzjL<^tbhG(R6;Qk_1)mYdMDdqdhv2iN($7d?sNQLEfN|5}-rn1iS~w z4o5%^^=wDMpDdat@DG^(Rg{<@s#!A|FPw>%VCOnqK=3Qa!cL>agh&WH@1mts$;L97 zd>r@<)=B+u6P;2BYN=c-=TAk8aLV@&)Y8K~r|L_pJoG9&lzUDV&ntxfNr*(w(5w`S zn4=`f@3Rg{+-B4g( zTfx_fLl$cW3Z^HP068pm(Tsmj*6)!AkqJ>nP=~?f3uX@Fy;3KQ`yuhf9H{~=8%yUP zz-b1~g5O03EbLN32Jpq=5NId6G>5`z2=vhOr8!j=-?(m-QXG0Iy23rXg_|CZi$@we?eUD$N}hI$Vc7FS%6aE?@A-@qC5$f|q72 z&%Jj6R?#w8O^KK1!ua849L56JeIsbv^O}rjfjkJZx~Grr)rPgPfYgnyg6xGJ)MAOG(GS~d$$n$tmyO&vN} zmp9yfP7K%D$u_A0P3~RSOKShu0NMgfRyEBUl0s^4b49QlKd~Jc95(*u8CdamZpJKh z;rt+Az$e*CbGH@;|Y<4u5y@hEaH?Dh${E#)=(m#SHY2c5L?FBzcC^42Q%R-{mX`zpLe zO>q@ZET(?W*Nkvql1pkxt#aR=IIHllpT5WWaW1EL%YDA~ZQk)3SMQ^-%G;}4Rc~|u zajw3`eJ^rZ9pN77=>Ttghsz^s2p!00)Q`9%``+OyIy)|Ta0aD(r5HCuw9$fb7~OMX zvSb!ZCnDK&CLKF*z$~3A=L;u}l+9u$ol7AhBnTmk5V8rO2Ep=2ye#s6FfD-~C z{YUZ3Wjn86zy3nLl%DA?nbX_!sY*Ii)~C+s`_bieSw9lX;fY?_rWbc@wZDj%eJu3( zraK7(Hlw$C)wk^}Jt(Zpm9xF~wfiqh``Wv!@`3h3q}8KKpIqB_qs!~FsT=Lq;X9|5 z=r`Ktkq6if90uS;BcFf&PLut#AcC#`+n7{)P3^ z0(mfS%S$_=EkgI*hAxI^tI%CU^bLW2ji^tc4-i>>|BA@!`!__-3hTXw^$SFnmkwOP zmK7=B!AHip97J1-t`4&BYFYgaE2Z4B5|1Pe#<8Z0mImKxph-m3mx&C~=GEX+9M{qv z#Us%|0y=DsyH_-D9G}=H$11(2N8{UdwJ))=Z(LitL zUTWx$HFQC7=A($^tH%H-X0LopCF;iK1to_5B+U78p&WzQHq_E0Tgf2{H2tfR3z^Xo zN7-I3?*CLYB4^p=Aoo8iCi=-7*Z(O+C^H@F=HmW0S{y>nC1ZZo%Zx3OMP~Fzy0TofRX%KGHh?u{p}ElMO@)3JW%ikkNV8miZbRBbGNZrC zmF0J~ihf~bM!SDy`LnGOyMxgTtoQ9^-gXIAv~(44iUiRpttm&w^#j%ot4Sm7W^Rpr znX^^Y%euD#j)?0E=}X!Ds;k+bxm|}#KqZ`=?s+MK1#uaIaNtT)462c!Lg~S zqItR@6=S)SS#m`D8kr8pQ5{I?sF1~)+<5lvS;v$t(tYgz8RvgGk4B^M=~Jchsc5!@ NZz7#7L>s>Z{{fWK7X1JK literal 0 HcmV?d00001 diff --git a/source/array.h b/source/array.h new file mode 100644 index 0000000..bb021ed --- /dev/null +++ b/source/array.h @@ -0,0 +1,66 @@ +#ifndef ARRAY_H +# define ARRAY_H + +#include +#include +#include +#include + +typedef struct Header_s{ + int size; + int capacity; + int type; +} Header; + +#ifndef memset + +void *memset(void *src, char c, size_t size) { + for (int i = 0; i < size; i++) { + *(char *)(src + i) = c; + } + return (src); +} + +#endif + +typedef void *Array; + +#define ARRAY_BASE_CAPACITY 64 + +#define ArraySize(arr) ((Header*)(arr) - 1)->size + +#define ArrayFree(arr) free(((Header*)(arr) - 1)) + +#define ArrayPush(arr, x)\ + do {\ + Header *head = NULL;\ + if (!arr) { \ + head = malloc(sizeof(x) * ARRAY_BASE_CAPACITY + sizeof(Header));\ + head->size = 0;\ + head->type = sizeof(x);\ + head->capacity = ARRAY_BASE_CAPACITY;\ + arr = (void *)(head + 1);\ + }\ + head = (Header*)(arr)-1;\ + assert(sizeof(x) == head->type);\ + if (head->size >= head->capacity) {\ + head->capacity *= 2;\ + head = realloc(head, head->type *head->capacity + sizeof(Header));\ + arr = (void *)(head + 1);\ + }\ + (arr)[head->size++] = x;\ + } while(0) + +#define ArrayClear(arr)\ + do {\ + assert(arr);\ + Header * head = (Header*)(arr)-1;\ + memset(arr, 0, head->size * head->type);\ + } while(0)\ + + +# ifdef ARRAY_IMPL + +# endif + +#endif diff --git a/source/graphic.h b/source/graphic.h new file mode 100644 index 0000000..dde863e --- /dev/null +++ b/source/graphic.h @@ -0,0 +1,7 @@ +#ifndef GRAPHIC_H +# define GRAPHIC_H + +//software renderer allowing for system agnostic rendering (other then drawing the frame on the screen) + + +#endif diff --git a/source/main.c b/source/main.c new file mode 100644 index 0000000..67ada91 --- /dev/null +++ b/source/main.c @@ -0,0 +1,94 @@ +#include +#include +#include + +char *LoadFile(const char *filename) { + char *data = NULL; + FILE *file = NULL; + const int buff_size = 2048; + char buffer[buff_size]; + int count = 0; + file = fopen(filename, "r"); + count = fread(buffer, 1, buff_size, file); + data = calloc(count + 1, sizeof(char)); + memcpy(data, buffer, count); + fclose(file); + return (data); +} + +typedef enum { + TOK_NONE, + TOK_STRING, + TOK_TOKEN, + TOK_PREPROCESSOR, +} TKN_CTX; + +typedef struct Token_s { + int size; + char *data; + TKN_CTX ctx; +} Token_t; + +typedef struct { + int size; + Token_t *data; +} TokenList; + +TokenList SeparateString(char *data) { + char* span = NULL; + char* tok = NULL; + char* tmp = NULL; + Token_t buffer[256]; + tok = strtok_r(data, "\"\'", &span); + int i = 0; + buffer[i].ctx = TOK_NONE; + while (tok) { + size_t size = (span - tok); + tmp = calloc(size + 1, sizeof(char)); + memcpy_s(tmp, size, tok, size); + tmp[size] = 0x00; + buffer[i].data = tmp; + buffer[i].size = size; + + //printf("|%s|\n", tmp); + //free(tmp); + tok = strtok_r(NULL, "\"\'", &span); + i++; + buffer[i].ctx = TOK_STRING; + } + TokenList token_list; + token_list.data = calloc(i, sizeof(Token_t)); + token_list.size = i; + return (token_list); +} + +int main(int ac, char **av) { + if (ac <= 1) { + printf("no file specified"); + return (-1); + } + char* data = LoadFile(av[1]); + TokenList tkn_lst; + //first pass on string + tkn_lst = SeparateString(data); + //second pass on ; and \n + //third pass on \t and space + + char* span = NULL; + char* tmp = NULL; + char* delim = " \t\n"; + char* special = ";#"; + size_t size = 0; + char* tok = strtok_r(data, delim, &span); + while (tok) { + size = (span - tok); + tmp = malloc(size + 1); + memcpy_s(tmp, size, tok, size); + tmp[size] = 0x00; + printf("|%s|\n", tmp); + free(tmp); + tok = strtok_r(NULL, " \t\n", &span); + } + free(data); + return(0); +} diff --git a/source/sterling_runtime_x86_64elf.nasm b/source/sterling_runtime_x86_64elf.nasm new file mode 100644 index 0000000..6261b98 --- /dev/null +++ b/source/sterling_runtime_x86_64elf.nasm @@ -0,0 +1,34 @@ +BITS 64 +CPU X64 + +section .text +global _start + +extern sterling_runtime_init;runtime setup +extern sterling_main;user entrypoint + +_start: + ;Clear RBP (Recommended by the ABI to mark the outermost frame) + xor rbp, rbp + + ;Extract argc and argv from the stack + ;[rsp] is argc, [rsp + 8] is argv[0] on 64bit + pop rdi ; rdi = argc + mov rsi, rsp ; rsi = argv (pointer to the current top of stack) + + ;The System V ABI requires the stack to be 16-byte aligned before a 'call'. + ;The kernel starts us aligned, but our 'pop' just moved rsp by 8 bytes. + and rsp, -16 + + call sterling_runtime_init ;elf loader hook for the exokernel ? + + ;Call to the 'main' function + ;Assuming signature like C: int sterling_main(int argc, char **argv) + ;May change it to reflect the object file implementation + call sterling_main + + ; 5. Exit System Call + ; sterling_main return its status in RAX. Move it to RDI for exit. + mov rdi, rax ; exit status + mov rax, 60 ; sys_exit (60 for x86_64) + syscall