X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.db-test.git;a=blobdiff_plain;f=pctb%2Focr.c;h=fe264e8110ce8ffce6cbb6be0282c004151dec31;hp=bb9576769e10a1d633ed6ebb7786d75e3bfcf585;hb=cf8e9b2a56d12305cd61f8d4a9284d4319265215;hpb=8b3b006869bfdc6c2ddbf58d6709a73433abe2f9 diff --git a/pctb/ocr.c b/pctb/ocr.c index bb95767..fe264e8 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -1,30 +1,83 @@ /* - */ + * 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" +#include "convert.h" typedef struct { Pixcol col; struct DatabaseNode *then; } DatabaseLink; -#define MAXGLYPHCHRS 3 - typedef struct DatabaseNode { - char s[MAXGLYPHCHRS+1]; /* null-terminated; "" means no match here */ + char *str; int nlinks, alinks; + unsigned match:1, defined:1, endsword:1; DatabaseLink *links; } DatabaseNode; +typedef struct { + OcrReader *rd; + int w; + Pixcol *cols; + int x; + unsigned ctxmap; +} FindCharArgs; +typedef struct { + DatabaseNode *match; + int rx; +} FindCharResults; + static const char *context_names[]= { "Lower", "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])) -#define SPACE_SPACES 3 +#define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0])) struct OcrReader { int h; @@ -33,92 +86,83 @@ struct OcrReader { int aresults, nresults; }; -static FILE *resolver; -static pid_t resolver_pid; -static int resolver_done; +DEBUG_DEFINE_DEBUGF(ocr) -static void fgetsline(FILE *f, char *lbuf, size_t lbufsz) { - char *s= fgets(lbuf,lbufsz,f); - eassert(s); - int l= strlen(lbuf); - eassert(l>0); eassert(lbuf[--l]='\n'); - lbuf[l]= 0; -} -#define FGETSLINE(f,buf) (fgetsline(f,buf,sizeof(buf))) +#define FGETSLINE (dbfile_getsline(lbuf,sizeof(lbuf),__FILE__,__LINE__)) static void cleardb_node(DatabaseNode *n) { int i; - n->s[0]= 0; + free(n->str); n->str=0; + n->defined=n->match=n->endsword= 0; for (i=0; inlinks; i++) cleardb_node(n->links[i].then); } +static void readdb1(OcrReader *rd, const char *which); + static void readdb(OcrReader *rd) { + int ctxi; + + for (ctxi=0; ctxicontexts[ctxi]); + + readdb1(rd, "master"); + readdb1(rd, "local"); +} + +static void readdb1(OcrReader *rd, const char *which) { int nchrs; DatabaseNode *current, *additional; - char chrs[MAXGLYPHCHRS+1]; + char chrs[100]; Pixcol cv; - int r,j,ctxi; - int h; + int j,ctxi; + 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); + char *dbfname= masprintf("%s/#%s-char%d#.txt", + get_vardir(), which, rd->h); - db= fopen(dbfname,"r"); - free(dbfname); - if (!db) { - eassert(errno==ENOENT); - return; - } + if (!dbfile_open(dbfname)) + goto x; - FGETSLINE(db,lbuf); - eassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); + FGETSLINE; + dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); - r= fscanf(db, "%d", &h); - eassert(r==1); - eassert(h==rd->h); + dbassert( dbfile_scanf("%d", &h) == 1); + dbassert(h==rd->h); for (;;) { - FGETSLINE(db,lbuf); - if (!lbuf || lbuf[0]=='#') continue; + FGETSLINE; + if (!lbuf[0] || lbuf[0]=='#') continue; if (!strcmp(lbuf,".")) break; for (ctxi=0; ctxi0 && cr<=255); - c= cr; - } + int c= fgetc(dbfile); sysassert(!ferror(dbfile)); dbassert(c!=EOF); + if (c=='\n') break; /* forces no match */ + dbassert(nchrs0 && chrs[nchrs-1]==' ') { + endsword= 1; + nchrs--; + } current= &rd->contexts[ctxi]; for (;;) { - FGETSLINE(db,lbuf); - if (!lbuf[0]) { eassert(current != &rd->contexts[ctxi]); break; } + FGETSLINE; + 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) { @@ -126,16 +170,18 @@ static void readdb(OcrReader *rd) { goto found_link; } - additional= malloc(sizeof(*additional)); eassert(additional); - additional->s[0]= 0; + additional= mmalloc(sizeof(*additional)); + additional->str= 0; + additional->defined= 0; + additional->match= 0; + additional->endsword= 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; @@ -145,14 +191,27 @@ static void readdb(OcrReader *rd) { found_link:; } - eassert(!current->s[0]); - strcpy(current->s, chrs); + if (!current->defined) { + free(current->str); + current->str= 0; + current->defined= 1; + current->match= 0; + + if (nchrs) { + current->str= mmalloc(nchrs+1); + memcpy(current->str, chrs, nchrs); + current->str[nchrs]= 0; + current->match= 1; + current->endsword= endsword; + } + } } - eassert(!ferror(db)); - eassert(!fclose(db)); + x: + dbfile_close(); + free(dbfname); } -static void cu_pr_ctxmap(unsigned ctxmap) { +static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) { fprintf(resolver,"{"); const char *spc=""; int ctxi; @@ -166,39 +225,25 @@ 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 c,i, x,y; const OcrResultGlyph *s; const char *p; - char cb; Pixcol pv; + + FILE *resolver= resolve_start(); + if (!resolver) + fatal("OCR failed - unrecognised characters or ligatures.\n" + "Character set database needs to be updated or augmented.\n" + "See README.charset.\n"); - if (!resolver) { - r= pipe(jobpipe); eassert(!r); - r= pipe(donepipe); eassert(!r); - resolver_pid= fork(); - eassert(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); - /* 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"); - } - r= close(jobpipe[0]); eassert(!r); - r= close(donepipe[1]); eassert(!r); - resolver= fdopen(jobpipe[1],"w"); eassert(resolver); - resolver_done= donepipe[0]; - } - fprintf(resolver,"%d %d ",unk_l,unk_r); - cu_pr_ctxmap(unk_ctxmap); + fprintf(resolver, + "char\n" + "%d %d ",unk_l,unk_r); + cu_pr_ctxmap(resolver,unk_ctxmap); for (i=0, s=rd->results; inresults; i++, s++) { if (!strcmp(s->s," ")) continue; fprintf(resolver," %d %d ",s->l,s->r); - cu_pr_ctxmap(s->ctxmap); + cu_pr_ctxmap(resolver,s->ctxmap); fprintf(resolver," "); for (p=s->s; (c= *p); p++) { if (c=='\\') fprintf(resolver,"\\%c",c); @@ -223,40 +268,8 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], fputs("\",\n",resolver); } fputs("};\n",resolver); - eassert(!ferror(resolver)); - eassert(!fflush(resolver)); - - eassert(resolver); - - for (;;) { - r= read(resolver_done,&cb,1); - if (r==-1) { eassert(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"); - } - } else { - eassert(r==1); - eassert(cb==0); - } + resolve_finish(); readdb(rd); } @@ -264,8 +277,7 @@ 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; @@ -274,31 +286,91 @@ 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 */ -}; + +static DatabaseNode *findchar_1ctx(const FindCharArgs *fca, + DatabaseNode *start, int *matchx_r) { + DatabaseNode *current= start; + DatabaseNode *bestmatch= 0; + int i; + int x= fca->x; + + for (;;) { + debug_flush(); + debugf(" | x=%d",x); + if (x > fca->w) break; + Pixcol cv= fca->cols[x]; + debugf(" cv=%"PSPIXCOL(PRIx),cv); + for (i=0; inlinks; i++) + if (current->links[i].col == cv) + goto found; + /* not found */ + debugf(" ?"); + break; + + found: + current= current->links[i].then; + if (current->match) { + debugf(" \"%s\"%s",current->str,current->endsword?"_":""); + bestmatch= current; + *matchx_r= x; + } else { + debugf(" ..."); + } + + x++; + } + return bestmatch; +} + +static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) { + FindCharResults results[NCONTEXTS]; + int ctxi, match=-1, nmatches=0; + + debugf("OCR lx=%d ct_state=%x ", fca->x, fca->ctxmap); + for (ctxi=0; ctxictxmap & (1u << ctxi))) continue; + debugf(" || %s",context_names[ctxi]); + + results[ctxi].match= findchar_1ctx(fca, &fca->rd->contexts[ctxi], + &results[ctxi].rx); + if (!results[ctxi].match) continue; + + match= ctxi; + nmatches++; + } + if (nmatches==1) { + debugf( " unambiguous"); + } else { + match=-1; + } + if (match<0) + return 0; + + *match_rx= results[ctxi].rx; + return results[ctxi].match; +} + +const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { int nspaces; - unsigned ctxmap; - int ctxi, i, x; + int x; + + FindCharArgs fca; + fca.rd= rd; + fca.w= w; + fca.cols= cols; + fca.x= -1; restart: nspaces=- w; - ctxmap= ct->initial; + fca.ctxmap= ct->initial; rd->nresults=0; - fprintf(debug,"OCR h=%d w=%d",rd->h,w); - for (x=0; xh,w); + for (x=0; xnextword; + if (nspaces == ct->space_spaces) { + debugf("OCR x=%x nspaces=%d space\n",x,nspaces); + fca.ctxmap= ct->nextword; } continue; } /* something here, so we need to add the spaces */ - if (nspaces>=SPACE_SPACES) + if (nspaces >= ct->space_spaces) add_result(rd," ",x-nspaces,x+1,0); nspaces=0; - /* find character */ - int lx=x; + fca.x= x; - DatabaseNode *uniquematch= 0; - int uniquematch_rx=-1; + int match_rx=-1; + DatabaseNode *match= findchar(&fca, &match_rx); - fprintf(debug,"OCR lx=%d ctxmap=%x ",lx,ctxmap); - - for (ctxi=0; ctxicontexts[ctxi];; - DatabaseNode *bestmatch= 0; - int bestmatch_rx=-1; - - x= lx; - if (!(ctxmap & (1u << ctxi))) continue; - fprintf(debug," || %s",context_names[ctxi]); - - for (;;) { - debug_flush(); - fprintf(debug," | x=%d",x); - if (x>w) break; - Pixcol cv= cols[x]; - fprintf(debug," cv=%"PSPIXCOL(PRIx),cv); - for (i=0; inlinks; i++) - if (current->links[i].col == cv) - goto found; - /* not found */ - fprintf(debug," ?"); - break; - - found: - current= current->links[i].then; - if (current->s[0]) { - fprintf(debug," \"%s\"",current->s); - bestmatch= current; - bestmatch_rx= x; - } else { - fprintf(debug," ..."); - } - - x++; - } - - if (bestmatch) { - if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) { - fprintf(debug, " ambiguous"); - uniquematch= 0; - break; - } - uniquematch= bestmatch; - uniquematch_rx= bestmatch_rx; + if (match) { + debugf(" || YES"); + add_result(rd, match->str, x, match_rx, fca.ctxmap); + x= match_rx+1; + if (match->match) fca.ctxmap= ct->midword; + else debugf(" (empty)"); + if (match->endsword) { + nspaces= ct->space_spaces; + debugf("_"); + fca.ctxmap= ct->nextword; } - } - - if (uniquematch) { - fprintf(debug," || YES\n"); - add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap); - x= uniquematch_rx+1; - ctxmap= ct->midword; + debugf("\n"); } else { int rx; - fprintf(debug," || UNKNOWN"); - for (rx=lx; rxnresults); + debugf("OCR finished %d glyphs\n",rd->nresults); debug_flush(); return rd->results; } @@ -401,7 +432,14 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { OcrReader *ocr_init(int h) { OcrReader *rd; - rd= malloc(sizeof(*rd)); eassert(rd); + if (o_flags & ff_dict_fetch) { + char *fetchfile= masprintf("char%d",h); + progress("Updating %s...",fetchfile); + fetch_with_rsync(fetchfile); + free(fetchfile); + } + + rd= mmalloc(sizeof(*rd)); memset(rd,0,sizeof(*rd)); rd->h= h; readdb(rd);