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=b7934c983e2f407a362183c0e5aad6756653165c;hp=be479f83b8d626442899c77f08ffdab049e26747;hb=1d0e4ca6694c19fadd97452eed1451248b156997;hpb=cd6a4f773c32f73aff27f97e8994c6b7c1019bb7 diff --git a/pctb/ocr.c b/pctb/ocr.c index be479f8..b7934c9 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -1,78 +1,175 @@ /* - */ + * 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 OCRDatabaseNode *then; -} OCRDatabaseLink; - -#define MAXGLYPHCHRS 3 + struct DatabaseNode *then; +} DatabaseLink; -typedef struct OCRDatabaseNode { - char s[MAXGLYPHCHRS+1]; /* null-terminated; "" means no match here */ +typedef struct DatabaseNode { + char *str; int nlinks, alinks; - OCRDatabaseLink *links; -} OCRDatabaseNode; + unsigned match:1, defined:1, endsword:1; + DatabaseLink *links; +} DatabaseNode; + +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])) + +struct OcrReader { + int h; + DatabaseNode contexts[NCONTEXTS]; + OcrResultGlyph *results; + int aresults, nresults; +}; + +DEBUG_DEFINE_DEBUGF(ocr) + +#define FGETSLINE (dbfile_getsline(lbuf,sizeof(lbuf),__FILE__,__LINE__)) + +static void cleardb_node(DatabaseNode *n) { + int i; + free(n->str); n->str=0; + n->defined=n->match=n->endsword= 0; + for (i=0; inlinks; i++) + cleardb_node(n->links[i].then); +} -#define N_OCR_CONTEXTS 2 +static void readdb1(OcrReader *rd, const char *which); -static OCRDatabaseNode ocr_contexts[N_OCR_CONTEXTS]; -static FILE *db; -static OcrResultGlyph *results; -static int aresults, nresults; +static void readdb(OcrReader *rd) { + int ctxi; + + for (ctxi=0; ctxicontexts[ctxi]); -static FILE *resolver; -static pid_t resolver_pid; -static int resolver_done; + readdb1(rd, "master"); + readdb1(rd, "local"); +} -static void ocr_readdb(void) { - int ctx,nchrs; - OCRDatabaseNode *current, *additional; - char chrs[MAXGLYPHCHRS+1]; +static void readdb1(OcrReader *rd, const char *which) { + int nchrs; + DatabaseNode *current, *additional; + char chrs[100]; Pixcol cv; - int r,i,j; + int j,ctxi; + int h, endsword; + char lbuf[100]; + + char *dbfname= masprintf("%s/#%s-char%d#.txt", + get_vardir(), which, rd->h); + + if (!dbfile_open(dbfname)) + goto x; - assert(!db); - db= fopen("database","r"); eassert(db); + FGETSLINE; + dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); + + dbassert( dbfile_scanf("%d", &h) == 1); + dbassert(h==rd->h); for (;;) { - r= fscanf(db, "%d %d", &ctx, &nchrs); - if (r==EOF) break; - eassert(r==2); - eassert(ctx>=0 && ctx0 && nchrs<=MAXGLYPHCHRS); - - for (i=0; i0 && c<=255); - chrs[i]= c; + FGETSLINE; + if (!lbuf[0] || lbuf[0]=='#') continue; + if (!strcmp(lbuf,".")) break; + + for (ctxi=0; ctxi0 && chrs[nchrs-1]==' ') { + endsword= 1; + nchrs--; } - chrs[nchrs]= 0; - int twidth; - r= fscanf(db, "%d", &twidth); eassert(r==1); - current= &ocr_contexts[ctx]; - for (i=0; icontexts[ctxi]; + for (;;) { + FGETSLINE; + if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; } + char *ep; + cv= strtoul(lbuf,&ep,16); dbassert(!*ep); + dbassert(!(cv & ~((1UL << rd->h)-1))); + for (j=0; jnlinks; j++) if (current->links[j].col == cv) { current= current->links[j].then; 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; @@ -82,47 +179,60 @@ static void ocr_readdb(void) { 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(feof(db)); -} - -static void callout_unknown(int w, int h, Pixcol cols[], - int unk_l, int unk_r, int unk_ctx, - const OcrResultGlyph *sofar, int nsofar) { - int jobpipe[2],donepipe[2], c, r,i, x,y; + x: + dbfile_close(); + free(dbfname); +} + +static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) { + fprintf(resolver,"{"); + const char *spc=""; + int ctxi; + for (ctxi=0; ctxi= 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 %d",unk_l,unk_r,unk_ctx); - for (i=0, s=sofar; iresults; inresults; i++, s++) { if (!strcmp(s->s," ")) continue; - fprintf(resolver," %d %d %d ",s->l,s->r,s->ctx); + fprintf(resolver," %d %d ",s->l,s->r); + cu_pr_ctxmap(resolver,s->ctxmap); + fprintf(resolver," "); for (p=s->s; (c= *p); p++) { if (c=='\\') fprintf(resolver,"\\%c",c); else if (c>=33 && c<=126) fputc(c,resolver); @@ -138,79 +248,122 @@ static void callout_unknown(int w, int h, Pixcol cols[], "\"%d %d 2 1\",\n" "\" c black\",\n" "\"o c white\",\n", - w,h); - for (y=0, pv=1; yh); + for (y=0, pv=1; yh; y++, pv<<=1) { fputc('"',resolver); for (x=0; xnresults >= rd->aresults) { + rd->aresults++; rd->aresults<<=1; + rd->results= mrealloc(rd->results, sizeof(*rd->results)*rd->aresults); + } + rd->results[rd->nresults].s= s; + rd->results[rd->nresults].l= l; + rd->results[rd->nresults].r= r; + rd->results[rd->nresults].ctxmap= ctxmap; + rd->nresults++; +} + + +typedef struct { + OcrReader *rd; + int w; + Pixcol *cols; + int x; + unsigned ctxmap; +} FindCharArgs; + +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 (;;) { - r= read(resolver_done,&cb,1); - if (r==-1) { eassert(errno==EINTR); continue; } + 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; - } - if (r==0) { - pid_t pid; - for (;;) { - pid= waitpid(resolver_pid, &r, 0); - if (pid==-1) { eassert(errno==EINTR); continue; } - break; + found: + current= current->links[i].then; + if (current->match) { + debugf(" \"%s\"%s",current->str,current->endsword?"_":""); + bestmatch= current; + *matchx_r= x; + } else { + debugf(" ..."); } - 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"); + + x++; + } + return bestmatch; +} + +static DatabaseNode *findchar_unique(const FindCharArgs *fca, int *match_rx) { + DatabaseNode *uniquematch= 0; + int ctxi; + + debugf("OCR lx=%d ct_state=%x ", fca->x, fca->ctxmap); + for (ctxi=0; ctxictxmap & (1u << ctxi))) continue; + debugf(" || %s",context_names[ctxi]); + + DatabaseNode *match= + findchar_1ctx(fca, &fca->rd->contexts[ctxi], match_rx); + if (!match) continue; + + if (uniquematch && strcmp(match->str, uniquematch->str)) { + debugf( " ambiguous"); + uniquematch= 0; } else { - eassert(!"weird wait status"); + uniquematch= match; } - } else { - eassert(r==1); - eassert(cb==0); } - fclose(db); - db= 0; - ocr_readdb(); + return uniquematch; } -static void add_result(const char *s, int l, int r, int ctx) { - if (nresults >= aresults) { - aresults++; aresults<<=1; - results= realloc(results,sizeof(*results)*aresults); - eassert(results); - } - results[nresults].s= s; - results[nresults].l= l; - results[nresults].r= r; - results[nresults].ctx= ctx; - nresults++; -} +const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } -OcrResultGlyph *ocr(int w, int h, Pixcol cols[]) { - int nspaces=-w; - int ctx=1,i, x; +OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { + int nspaces; + int x; - nresults=0; - assert(db); + FindCharArgs fca; + fca.rd= rd; + fca.w= w; + fca.cols= cols; + fca.x= -1; - fprintf(debug,"OCR h=%d w=%d",w,h); - for (x=0; xinitial; + rd->nresults=0; + debugf("OCR h=%d w=%d",rd->h,w); + for (x=0; xspace_spaces) { + debugf("OCR x=%x nspaces=%d space\n",x,nspaces); + fca.ctxmap= ct->nextword; } continue; } - nspaces=0; - - /* find character */ - OCRDatabaseNode *current=0, *bestmatch=0; - int lx=x; - int bestmatch_rx=-1; - current= &ocr_contexts[ctx]; - fprintf(debug,"OCR lx=%d ctx=%d ",lx,ctx); - for (;;) { - debug_flush(); - fprintf(debug,"| x=%d",x); - if (x>w) break; - Pixcol cv= cols[x]; - fprintf(debug," cv=%"PSPIXCOL(PRIx),x); - for (i=0; inlinks; i++) - if (current->links[i].col == cv) - goto found; - /* not found */ - fprintf(debug," ?"); - break; + /* something here, so we need to add the spaces */ + if (nspaces >= ct->space_spaces) + add_result(rd," ",x-nspaces,x+1,0); + nspaces=0; - found: - current= current->links[i].then; - if (current->s[0]) { - fprintf(debug," \"%s\"",current->s); - bestmatch=current; bestmatch_rx=x; - } else { - fprintf(debug," ..."); + fca.x= x; + + int match_rx=-1; + DatabaseNode *match= findchar_unique(&fca, &match_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; } - x++; - } - - if (bestmatch) { - fprintf(debug," YES\n"); - add_result(bestmatch->s, lx, bestmatch_rx, ctx); - x= bestmatch_rx+1; - ctx= 0; + debugf("\n"); } else { int rx; - fprintf(debug," UNKNOWN"); - for (rx=lx; rxnresults); debug_flush(); - return results; + return rd->results; } -void ocr_init(void) { - ocr_readdb(); +OcrReader *ocr_init(int h) { + OcrReader *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); + return rd; }