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=2c3a16772ec8ac0a90f5a9be420c9db188c1e6da;hp=fafdba8eec2ac74f54c4ffc05f096f62c0c5d479;hb=0487d41d9c7796c6efebb8235c9aa39fe580ccf9;hpb=8360f12a6a457b73ebb18dbeedbb15e6ed91318b diff --git a/pctb/ocr.c b/pctb/ocr.c index fafdba8..2c3a167 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -26,18 +26,17 @@ */ #include "ocr.h" +#include "convert.h" typedef struct { Pixcol col; struct DatabaseNode *then; } DatabaseLink; -#define MAXGLYPHCHRS 7 - typedef struct DatabaseNode { - char s[MAXGLYPHCHRS+1]; /* null-terminated; "" means no match here */ + char *str; int nlinks, alinks; - unsigned endsword:1; + unsigned match:1, defined:1, endsword:1; DatabaseLink *links; } DatabaseNode; @@ -75,36 +74,41 @@ struct OcrReader { int aresults, nresults; }; -static FILE *resolver; -static pid_t resolver_pid; -static int resolver_done; - DEBUG_DEFINE_DEBUGF(ocr) #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 j,ctxi; int h, endsword; char lbuf[100]; - for (ctxi=0; ctxicontexts[ctxi]); - - char *dbfname=0; - asprintf(&dbfname,"%s/charset-%d.txt",get_vardir(),rd->h); - sysassert(dbfname); + char *dbfname= masprintf("%s/#%s-char%d#.txt", + get_vardir(), which, rd->h); if (!dbfile_open(dbfname)) goto x; @@ -130,23 +134,15 @@ static void readdb(OcrReader *rd) { found_ctx: for (nchrs=0;;) { int c= fgetc(dbfile); sysassert(!ferror(dbfile)); dbassert(c!=EOF); - if (c=='\n') { dbassert(nchrs); break; } - dbassert(nchrs0 && cr<=255); - c= cr; - } + if (c=='\n') break; /* forces no match */ + dbassert(nchrs1 && chrs[nchrs-1]==' ') { + if (nchrs>0 && chrs[nchrs-1]==' ') { endsword= 1; nchrs--; } - chrs[nchrs]= 0; current= &rd->contexts[ctxi]; for (;;) { @@ -163,7 +159,10 @@ static void readdb(OcrReader *rd) { } additional= mmalloc(sizeof(*additional)); - additional->s[0]= 0; + 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) { @@ -180,16 +179,27 @@ static void readdb(OcrReader *rd) { found_link:; } - dbassert(!current->s[0]); - strcpy(current->s, chrs); - current->endsword= endsword; + 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; + } + } } 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; @@ -203,46 +213,25 @@ 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,i, x,y; + int c,i, x,y; const OcrResultGlyph *s; const char *p; - char cb; Pixcol pv; - if (!o_resolver) + 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) { - sysassert(! pipe(jobpipe) ); - sysassert(! pipe(donepipe) ); - resolver_pid= fork(); - sysassert(resolver_pid!=-1); - if (!resolver_pid) { - 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: */ - sysassert( dup2(donepipe[1],4) ==4 ); - execlp(o_resolver, o_resolver, - DEBUGP(callout) ? "--debug" : "--noop-arg", - "--automatic-1", - (char*)0); - sysassert(!"execlp ocr-resolver failed"); - } - 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); - 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); @@ -267,28 +256,8 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], fputs("\",\n",resolver); } fputs("};\n",resolver); - sysassert(!ferror(resolver)); - sysassert(!fflush(resolver)); - - sysassert(resolver); - - int r; - for (;;) { - r= read(resolver_done,&cb,1); - if (r==-1) { sysassert(errno==EINTR); continue; } - break; - } - - if (r==0) { - waitpid_check_exitstatus(resolver_pid, "character resolver"); - fclose(resolver); - close(resolver_done); - resolver= 0; - } else { - assert(r==1); - sysassert(cb==0); - } + resolve_finish(); readdb(rd); } @@ -377,8 +346,8 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { found: current= current->links[i].then; - if (current->s[0]) { - debugf(" \"%s\"%s",current->s,current->endsword?"_":""); + if (current->match) { + debugf(" \"%s\"%s",current->str,current->endsword?"_":""); bestmatch= current; bestmatch_rx= x; } else { @@ -389,7 +358,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { } if (bestmatch) { - if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) { + if (uniquematch && strcmp(bestmatch->str, uniquematch->str)) { debugf( " ambiguous"); uniquematch= 0; break; @@ -401,9 +370,9 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { if (uniquematch) { debugf(" || YES"); - add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap); + add_result(rd, uniquematch->str, lx, uniquematch_rx, ctxmap); x= uniquematch_rx+1; - if (uniquematch->s[0]) ctxmap= ct->midword; + if (uniquematch->match) ctxmap= ct->midword; else debugf(" (empty)"); if (uniquematch->endsword) { nspaces= ct->space_spaces; @@ -430,6 +399,13 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { 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;