X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.db-test.git;a=blobdiff_plain;f=pctb%2Focr.c;h=6a91935769ecfc7da0e70681ea80db627c49327e;hp=04aecc0d48a0431642d73c85f8b65144fa74b491;hb=663030807a054534cd545be2043ea30eedc4b3e6;hpb=5a2c03e2e4f52b8329f45cf67afc3edec1f2c65b diff --git a/pctb/ocr.c b/pctb/ocr.c index 04aecc0..6a91935 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -1,5 +1,29 @@ /* - */ + * Core OCR algorithm (first exact bitmap match) + */ +/* + * This is part of ypp-sc-tools, a set of third-party tools for assisting + * players of Yohoho Puzzle Pirates. + * + * Copyright (C) 2009 Ian Jackson + * + * This program is free software: you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation, either version 3 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program. If not, see . + * + * Yohoho and Puzzle Pirates are probably trademarks of Three Rings and + * are used without permission. This program is not endorsed or + * sponsored by Three Rings. + */ #include "ocr.h" @@ -8,11 +32,12 @@ typedef struct { struct DatabaseNode *then; } DatabaseLink; -#define MAXGLYPHCHRS 3 +#define MAXGLYPHCHRS 7 typedef struct DatabaseNode { char s[MAXGLYPHCHRS+1]; /* null-terminated; "" means no match here */ int nlinks, alinks; + unsigned endsword:1; DatabaseLink *links; } DatabaseNode; @@ -21,6 +46,25 @@ static const char *context_names[]= { "Upper", "Digit" }; +struct OcrCellTypeInfo { + /* bitmaps of indices into context_names: */ + unsigned initial, nextword, midword; + int space_spaces; + const char *name; +}; +const struct OcrCellTypeInfo ocr_celltype_number= { + 4,4,4, + .space_spaces= 5, + .name= "number" +}; +const struct OcrCellTypeInfo ocr_celltype_text= { + .initial=2, /* Uppercase */ + .nextword=3, /* Either */ + .midword=1, /* Lower only */ + .space_spaces= 4, + .name= "text" +}; + #define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0])) @@ -35,11 +79,21 @@ static FILE *resolver; static pid_t resolver_pid; static int resolver_done; +DEBUG_DEFINE_DEBUGF(ocr) + +#define dbassert(x) \ + ((x) ? (void)0 : \ + fatal("Error in character set database.\n" \ + " Requirement not met: %s:%d: %s", __FILE__,__LINE__, #x)) + static void fgetsline(FILE *f, char *lbuf, size_t lbufsz) { + errno=0; char *s= fgets(lbuf,lbufsz,f); - eassert(s); + sysassert(!ferror(f)); + dbassert(!feof(f)); + assert(s); int l= strlen(lbuf); - eassert(l>0); eassert(lbuf[--l]='\n'); + dbassert(l>0); dbassert(lbuf[--l]='\n'); lbuf[l]= 0; } #define FGETSLINE(f,buf) (fgetsline(f,buf,sizeof(buf))) @@ -57,26 +111,30 @@ static void readdb(OcrReader *rd) { char chrs[MAXGLYPHCHRS+1]; Pixcol cv; int r,j,ctxi; - int h; + int h, endsword; char lbuf[100]; FILE *db; + for (ctxi=0; ctxicontexts[ctxi]); + char *dbfname=0; asprintf(&dbfname,"%s/charset-%d.txt",get_vardir(),rd->h); - eassert(dbfname); + sysassert(dbfname); - db= fopen(dbfname,"r"); eassert(db); + db= fopen(dbfname,"r"); free(dbfname); + if (!db) { + sysassert(errno==ENOENT); + return; + } FGETSLINE(db,lbuf); - eassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); + dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); r= fscanf(db, "%d", &h); - eassert(r==1); - eassert(h==rd->h); - - for (ctxi=0; ctxicontexts[ctxi]); + dbassert(r==1); + dbassert(h==rd->h); for (;;) { FGETSLINE(db,lbuf); @@ -92,27 +150,32 @@ static void readdb(OcrReader *rd) { found_ctx: for (nchrs=0;;) { - int c= fgetc(db); eassert(c!=EOF); - if (c=='\n') { eassert(nchrs); break; } - eassert(nchrs0 && cr<=255); c= cr; } chrs[nchrs++]= c; } + endsword= 0; + if (nchrs>1 && chrs[nchrs-1]==' ') { + endsword= 1; + nchrs--; + } chrs[nchrs]= 0; current= &rd->contexts[ctxi]; for (;;) { FGETSLINE(db,lbuf); - if (!lbuf[0]) { eassert(current != &rd->contexts[ctxi]); break; } + if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; } char *ep; - cv= strtoul(lbuf,&ep,16); eassert(!*ep); - eassert(!(cv & ~((1UL << rd->h)-1))); + cv= strtoul(lbuf,&ep,16); dbassert(!*ep); + dbassert(!(cv & ~((1UL << rd->h)-1))); for (j=0; jnlinks; j++) if (current->links[j].col == cv) { @@ -120,16 +183,15 @@ static void readdb(OcrReader *rd) { goto found_link; } - additional= malloc(sizeof(*additional)); eassert(additional); + additional= mmalloc(sizeof(*additional)); additional->s[0]= 0; additional->nlinks= additional->alinks= 0; additional->links= 0; if (current->nlinks==current->alinks) { current->alinks++; current->alinks<<=1; - current->links= realloc(current->links, - sizeof(*current->links) * current->alinks); - eassert(current->links); + current->links= mrealloc(current->links, + sizeof(*current->links) * current->alinks); } current->links[current->nlinks].col= cv; current->links[current->nlinks].then= additional; @@ -139,11 +201,12 @@ static void readdb(OcrReader *rd) { found_link:; } - eassert(!current->s[0]); + dbassert(!current->s[0]); strcpy(current->s, chrs); + current->endsword= endsword; } - eassert(!ferror(db)); - eassert(!fclose(db)); + sysassert(!ferror(db)); + sysassert(!fclose(db)); } static void cu_pr_ctxmap(unsigned ctxmap) { @@ -160,31 +223,33 @@ static void cu_pr_ctxmap(unsigned ctxmap) { static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], int unk_l, int unk_r, unsigned unk_ctxmap) { - int jobpipe[2],donepipe[2], c, r,i, x,y; + int jobpipe[2],donepipe[2], c,i, x,y; const OcrResultGlyph *s; const char *p; char cb; Pixcol pv; if (!resolver) { - r= pipe(jobpipe); eassert(!r); - r= pipe(donepipe); eassert(!r); + sysassert(! pipe(jobpipe) ); + sysassert(! pipe(donepipe) ); resolver_pid= fork(); - eassert(resolver_pid!=-1); + sysassert(resolver_pid!=-1); if (!resolver_pid) { - r= dup2(jobpipe[0],0); eassert(r==0); - r= close(jobpipe[1]); eassert(!r); - r= close(donepipe[0]); eassert(!r); + sysassert( dup2(jobpipe[0],0) ==0 ); + sysassert(! close(jobpipe[1]) ); + sysassert(! close(donepipe[0]) ); /* we know donepipe[1] is >= 4 and we have dealt with all the others * so we aren't in any danger of overwriting some other fd 4: */ - r= dup2(donepipe[1],4); eassert(r==4); - execlp("./show-thing.tcl", "./show-thing.tcl", - "--automatic","1",(char*)0); - eassert(!"execlp failed"); + sysassert( dup2(donepipe[1],4) ==4 ); + execlp("./yppsc-ocr-resolver", "yppsc-ocr-resolver", + DEBUGP(callout) ? "--debug" : "--noop-arg", + "--automatic-1", + (char*)0); + sysassert(!"execlp ocr-resolver failed"); } - r= close(jobpipe[0]); eassert(!r); - r= close(donepipe[1]); eassert(!r); - resolver= fdopen(jobpipe[1],"w"); eassert(resolver); + sysassert(! close(jobpipe[0]) ); + sysassert(! close(donepipe[1]) ); + resolver= fdopen(jobpipe[1],"w"); sysassert(resolver); resolver_done= donepipe[0]; } fprintf(resolver,"%d %d ",unk_l,unk_r); @@ -217,38 +282,26 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], fputs("\",\n",resolver); } fputs("};\n",resolver); - eassert(!ferror(resolver)); - eassert(!fflush(resolver)); + sysassert(!ferror(resolver)); + sysassert(!fflush(resolver)); - eassert(resolver); + sysassert(resolver); + int r; for (;;) { r= read(resolver_done,&cb,1); - if (r==-1) { eassert(errno==EINTR); continue; } + if (r==-1) { sysassert(errno==EINTR); continue; } break; } if (r==0) { - pid_t pid; - for (;;) { - pid= waitpid(resolver_pid, &r, 0); - if (pid==-1) { eassert(errno==EINTR); continue; } - break; - } - eassert(pid==resolver_pid); - if (WIFEXITED(r)) { - eassert(!WEXITSTATUS(r)); - fclose(resolver); - close(resolver_done); - resolver= 0; - } else if (WIFSIGNALED(r)) { - eassert(!"resolver child died due to signal"); - } else { - eassert(!"weird wait status"); - } + waitpid_check_exitstatus(resolver_pid, "character resolver"); + fclose(resolver); + close(resolver_done); + resolver= 0; } else { - eassert(r==1); - eassert(cb==0); + assert(r==1); + sysassert(cb==0); } readdb(rd); @@ -258,8 +311,7 @@ static void add_result(OcrReader *rd, const char *s, int l, int r, unsigned ctxmap) { if (rd->nresults >= rd->aresults) { rd->aresults++; rd->aresults<<=1; - rd->results= realloc(rd->results,sizeof(*rd->results)*rd->aresults); - eassert(rd->results); + rd->results= mrealloc(rd->results, sizeof(*rd->results)*rd->aresults); } rd->results[rd->nresults].s= s; rd->results[rd->nresults].l= l; @@ -268,17 +320,8 @@ static void add_result(OcrReader *rd, const char *s, int l, int r, rd->nresults++; } -struct OcrCellTypeInfo { - unsigned initial, nextword, midword; -}; -const struct OcrCellTypeInfo ocr_celltype_number= { - 4,4,4 -}; -const struct OcrCellTypeInfo ocr_celltype_text= { - .initial=2 /* Uppercase */, - .nextword=3 /* Either */, - .midword=1 /* Lower only */ -}; + +const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { int nspaces; @@ -290,9 +333,9 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { nspaces=- w; ctxmap= ct->initial; rd->nresults=0; - fprintf(debug,"OCR h=%d w=%d",rd->h,w); - for (x=0; xh,w); + for (x=0; xspace_spaces) { + debugf("OCR x=%x nspaces=%d space\n",x,nspaces); ctxmap= ct->nextword; } continue; } + + /* something here, so we need to add the spaces */ + if (nspaces >= ct->space_spaces) + add_result(rd," ",x-nspaces,x+1,0); nspaces=0; /* find character */ @@ -320,7 +366,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { DatabaseNode *uniquematch= 0; int uniquematch_rx=-1; - fprintf(debug,"OCR lx=%d ctxmap=%x ",lx,ctxmap); + debugf("OCR lx=%d ctxmap=%x ",lx,ctxmap); for (ctxi=0; ctxicontexts[ctxi];; @@ -329,29 +375,29 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { x= lx; if (!(ctxmap & (1u << ctxi))) continue; - fprintf(debug," || %s",context_names[ctxi]); + debugf(" || %s",context_names[ctxi]); for (;;) { debug_flush(); - fprintf(debug," | x=%d",x); + debugf(" | x=%d",x); if (x>w) break; Pixcol cv= cols[x]; - fprintf(debug," cv=%"PSPIXCOL(PRIx),cv); + debugf(" cv=%"PSPIXCOL(PRIx),cv); for (i=0; inlinks; i++) if (current->links[i].col == cv) goto found; /* not found */ - fprintf(debug," ?"); + debugf(" ?"); break; found: current= current->links[i].then; if (current->s[0]) { - fprintf(debug," \"%s\"",current->s); + debugf(" \"%s\"%s",current->s,current->endsword?"_":""); bestmatch= current; bestmatch_rx= x; } else { - fprintf(debug," ..."); + debugf(" ..."); } x++; @@ -359,7 +405,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { if (bestmatch) { if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) { - fprintf(debug, " ambiguous"); + debugf( " ambiguous"); uniquematch= 0; break; } @@ -369,22 +415,29 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { } if (uniquematch) { - fprintf(debug," || YES\n"); + debugf(" || YES"); add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap); x= uniquematch_rx+1; - ctxmap= ct->midword; + if (uniquematch->s[0]) ctxmap= ct->midword; + else debugf(" (empty)"); + if (uniquematch->endsword) { + nspaces= ct->space_spaces; + debugf("_"); + ctxmap= ct->nextword; + } + debugf("\n"); } else { int rx; - fprintf(debug," || UNKNOWN"); + debugf(" || UNKNOWN"); for (rx=lx; rxnresults); + debugf("OCR finished %d glyphs\n",rd->nresults); debug_flush(); return rd->results; } @@ -392,7 +445,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { OcrReader *ocr_init(int h) { OcrReader *rd; - rd= malloc(sizeof(*rd)); eassert(rd); + rd= mmalloc(sizeof(*rd)); memset(rd,0,sizeof(*rd)); rd->h= h; readdb(rd);