/*
- */
+ * 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 <ijackson@chiark.greenend.org.uk>
+ *
+ * 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 <http://www.gnu.org/licenses/>.
+ *
+ * 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 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, local:1;
DatabaseLink *links;
} DatabaseNode;
-static const char *context_names[]= {
- "Lower",
- "Upper",
- "Digit"
+typedef struct {
+ OcrReader *rd;
+ OcrCellType ct;
+ int w;
+ const Pixcol *cols;
+ int x;
+ unsigned ctxmap;
+} FindCharArgs;
+typedef struct {
+ DatabaseNode *match;
+ int rx;
+} FindCharResults;
+
+#define FOR_EACH_CONTEXT(EACH) \
+ EACH(Word) \
+ EACH(Lower) \
+ EACH(Upper) \
+ EACH(Digit)
+
+#define FEC_ENUM(Context) ct_##Context,
+#define FEC_BIT(Context) ctf_##Context = 1 << ct_##Context,
+enum {
+ FOR_EACH_CONTEXT(FEC_ENUM)
+ FOR_EACH_CONTEXT(FEC_BIT)
};
-#define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
+#define FEC_STRINGS(Context) #Context,
+static const char *context_names[]= { FOR_EACH_CONTEXT(FEC_STRINGS) };
+
+struct OcrCellTypeInfo {
+ /* bitmaps of indices into context_names: */
+ unsigned initial, nextword, midword;
+ int space_spaces;
+ const char *name;
+ int (*findchar_select)(const FindCharArgs *fca,
+ const FindCharResults results[]);
+};
-#define SPACE_SPACES 4
+#define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
struct OcrReader {
int h;
DatabaseNode contexts[NCONTEXTS];
+ char *result;
+ int lresult, aresult;
OcrResultGlyph *results;
int aresults, nresults;
};
-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);
- sysassert(!ferror(f));
- dbassert(!feof(f));
- assert(s);
- int l= strlen(lbuf);
- dbassert(l>0); dbassert(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; i<n->nlinks; i++)
cleardb_node(n->links[i].then);
}
+static void readdb1(OcrReader *rd, const char *which, int local);
+
static void readdb(OcrReader *rd) {
+ int ctxi;
+
+ for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
+ cleardb_node(&rd->contexts[ctxi]);
+
+ readdb1(rd, "master", 0);
+ readdb1(rd, "local", 1);
+}
+
+static void readdb1(OcrReader *rd, const char *which, int local) {
int nchrs;
DatabaseNode *current, *additional;
- char chrs[MAXGLYPHCHRS+1];
+ char chrs[100];
Pixcol cv;
- int r,j,ctxi;
+ int j,ctxi;
int h, endsword;
char lbuf[100];
- FILE *db;
-
- for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
- cleardb_node(&rd->contexts[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);
- db= fopen(dbfname,"r");
- free(dbfname);
- if (!db) {
- sysassert(errno==ENOENT);
- return;
- }
+ if (!dbfile_open(dbfname))
+ goto x;
- FGETSLINE(db,lbuf);
- dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1"));
+ if (local)
+ o_flags |= ff_charset_havelocal;
- r= fscanf(db, "%d", &h);
- dbassert(r==1);
+ FGETSLINE;
+ dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v3 depth=" STRING(AADEPTH)));
+
+ 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; ctxi<NCONTEXTS; ctxi++)
if (!strcmp(lbuf,context_names[ctxi]))
goto found_ctx;
/* not found, just skip */
- for (;;) { FGETSLINE(db,lbuf); if (!lbuf[0]) break; }
+ for (;;) { FGETSLINE; if (!lbuf[0]) break; }
continue;
found_ctx:
for (nchrs=0;;) {
- int c= fgetc(db); sysassert(!ferror(db)); dbassert(c!=EOF);
- if (c=='\n') { dbassert(nchrs); break; }
- dbassert(nchrs<MAXGLYPHCHRS);
- if (c=='\\') {
- unsigned cr;
- c= fgetc(db); sysassert(!ferror(db)); dbassert(c=='x');
- r= fscanf(db, "%2x", &cr); sysassert(!ferror(db)); dbassert(r==1);
- assert(cr>0 && cr<=255);
- c= cr;
- }
+ int c= fgetc(dbfile); sysassert(!ferror(dbfile)); dbassert(c!=EOF);
+ if (c=='\n') break; /* forces no match */
+ dbassert(nchrs<sizeof(chrs));
chrs[nchrs++]= c;
}
endsword= 0;
- if (nchrs>1 && chrs[nchrs-1]==' ') {
+ if (nchrs>0 && chrs[nchrs-1]==' ') {
endsword= 1;
nchrs--;
}
- chrs[nchrs]= 0;
current= &rd->contexts[ctxi];
for (;;) {
- FGETSLINE(db,lbuf);
+ 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)));
+ dbassert( strlen(lbuf) == rd->h );
+ FILLZERO(cv);
+ int y;
+ for (y=0; y<h; y++) {
+ char pb[2];
+ pb[0]= lbuf[y];
+ pb[1]= 0;
+ char *ep;
+ unsigned pv= strtoul(pb,&ep,16); dbassert(!*ep);
+ pixcol_p_add(&cv, y, pv);
+ }
for (j=0; j<current->nlinks; j++)
- if (current->links[j].col == cv) {
+ if (!pixcol_cmp(&cv, ¤t->links[j].col)) {
current= current->links[j].then;
goto found_link;
}
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) {
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;
+ current->local= local;
+
+ if (nchrs) {
+ current->str= mmalloc(nchrs+1);
+ memcpy(current->str, chrs, nchrs);
+ current->str[nchrs]= 0;
+ current->match= 1;
+ current->endsword= endsword;
+ }
+ }
}
- sysassert(!ferror(db));
- sysassert(!fclose(db));
+ x:
+ dbfile_close();
+ free(dbfname);
}
-static void cu_pr_ctxmap(unsigned ctxmap) {
+typedef struct Rejection Rejection;
+struct Rejection {
+ struct Rejection *next;
+ const char *fname;
+ int lno;
+ pcre *re;
+};
+
+Rejection *rejections;
+
+static void load_rejections(const char *which) {
+ char lbuf[1000];
+ char *fname= masprintf("%s/#%s-reject#.txt", get_vardir(), which);
+ int c, lno=0;
+ Rejection *rej;
+
+ if (!dbfile_open(fname)) { free(fname); return; }
+
+ while ((c= fgetc(dbfile))!=EOF) {
+ ungetc(c,dbfile);
+ lno++;
+ dbfile_getsline(lbuf,sizeof(lbuf),fname,lno);
+
+ if (!lbuf[0] || isspace(lbuf[0] || lbuf[0]=='#'))
+ continue;
+
+ rej= mmalloc(sizeof(*rej));
+ rej->next= rejections;
+ rej->fname= fname;
+ rej->lno= lno;
+
+ const char *err;
+ int erroffset;
+ rej->re= pcre_compile(lbuf, PCRE_NO_AUTO_CAPTURE|PCRE_UTF8,
+ &err, &erroffset, 0);
+ if (!rej->re) {
+ char *what= masprintf("invalid regexp at offset %d: %s\n",
+ erroffset, err);
+ dbfile_assertfail(fname, lno, what);
+ }
+ debugf("OCR LOADED REJECTION %s:%d `%s' %p\n", fname,lno, lbuf, rej->re);
+
+ rejections= rej;
+ }
+ sysassert(feof(dbfile));
+ dbfile_close();
+}
+
+static int should_reject(OcrReader *rd) {
+ static int rejections_loaded;
+ int ovector[30];
+
+ if (!rejections_loaded) {
+ fetch_with_rsync("reject");
+ load_rejections("master");
+ load_rejections("local");
+ rejections_loaded=1;
+ }
+
+ debugf("[OCR REJECTION `%s'%d", rd->result, rd->lresult);
+ Rejection *rej;
+ for (rej=rejections; rej; rej=rej->next) {
+ debugf(" (%p)", rej);
+
+ int res= pcre_exec(rej->re, 0, rd->result,rd->lresult, 0,
+ 0, ovector, ARRAYSIZE(ovector));
+ if (res==PCRE_ERROR_NOMATCH) continue;
+ sysassert(res>0);
+
+ debugf(" MATCH]\n");
+ fprintf(stderr,"Rejecting OCR result `%s' (due to %s:%d)\n",
+ rd->result, rej->fname,rej->lno);
+ return 1;
+ }
+ debugf(" OK]");
+ return 0;
+}
+
+static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) {
fprintf(resolver,"{");
const char *spc="";
int ctxi;
fprintf(resolver,"}");
}
-static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
+static void callout_unknown(OcrReader *rd, int w, const 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;
+
+ FILE *resolver= resolve_start();
+ if (!resolver ||
+ !((o_flags & ff_charset_edit) ||
+ ((o_flags & ff_charset_allowedit) &&
+ (o_flags & ff_charset_havelocal))))
+ 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("./show-thing.tcl", "./show-thing.tcl",
- DEBUGP(callout) ? "--debug" : "--noop-arg",
- "--automatic-1",
- (char*)0);
- sysassert(!"execlp 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; i<rd->nresults; i++, s++) {
if (!strcmp(s->s," ")) continue;
fprintf(resolver," %d %d ",s->l,s->r);
- cu_pr_ctxmap(s->ctxmap);
+ cu_pr_ctxmap(resolver, 1u << s->match);
+ fprintf(resolver," ");
+ cu_pr_ctxmap(resolver, s->ctxmap);
fprintf(resolver," ");
for (p=s->s; (c= *p); p++) {
if (c=='\\') fprintf(resolver,"\\%c",c);
fputc('\n',resolver);
fprintf(resolver,
- "/* XPM */\n"
- "static char *t[] = {\n"
- "/* columns rows colors chars-per-pixel */\n"
- "\"%d %d 2 1\",\n"
- "\" c black\",\n"
- "\"o c white\",\n",
- w,rd->h);
- for (y=0, pv=1; y<rd->h; y++, pv<<=1) {
- fputc('"',resolver);
+ "P2\n%d %d %d\n", w, rd->h, AAMAXVAL);
+ for (y=0; y<rd->h; y++) {
for (x=0; x<w; x++)
- fputc(cols[x] & pv ? 'o' : ' ', resolver);
- 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) {
- pid_t pid;
- int st;
- for (;;) {
- pid= waitpid(resolver_pid, &st, 0);
- if (pid==-1) { sysassert(errno==EINTR); continue; }
- break;
- }
- sysassert(pid==resolver_pid);
- if (WIFEXITED(st)) {
- if (WEXITSTATUS(st))
- fatal("character resolver failed with nonzero exit status %d",
- WEXITSTATUS(st));
- fclose(resolver);
- close(resolver_done);
- resolver= 0;
- } else if (WIFSIGNALED(st)) {
- fatal("character resolver died due to signal %s%s",
- strsignal(WTERMSIG(st)), WCOREDUMP(st)?" (core dumped)":"");
- } else {
- fatal("character resolver gave strange wait status %d",st);
- }
- } else {
- assert(r==1);
- sysassert(cb==0);
+ fprintf(resolver, " %d", pixcol_p_get(&cols[x], y));
+ fputc('\n',resolver);
}
+ resolve_finish();
readdb(rd);
}
static void add_result(OcrReader *rd, const char *s, int l, int r,
- unsigned ctxmap) {
+ int match, unsigned ctxmap) {
if (rd->nresults >= 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].match= match;
rd->results[rd->nresults].ctxmap= ctxmap;
rd->nresults++;
+
+ if (!s) return; /* just the sentinel for the caller */
+
+ int sl= strlen(s);
+ int newlresult= rd->lresult + sl;
+ if (newlresult >= rd->aresult) {
+ rd->aresult= (newlresult << 1) + 1;
+ rd->result= mrealloc(rd->result, rd->aresult);
+ }
+ memcpy(rd->result + rd->lresult, s, sl);
+ rd->lresult= newlresult;
+ rd->result[rd->lresult]= 0;
+}
+
+
+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="PIXCOL_PRFMT, PIXCOL_PRVAL(cv));
+ for (i=0; i<current->nlinks; i++)
+ if (!pixcol_cmp(&cv, ¤t->links[i].col))
+ 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, int *match_rctxi) {
+ FindCharResults results[NCONTEXTS];
+ int ctxi, match=-1, nmatches=0;
+
+ debugf("OCR lx=%d ct_state=%x ", fca->x, fca->ctxmap);
+ for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
+ results[ctxi].match= 0;
+ if (!(fca->ctxmap & (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(" unique");
+ } else {
+ debugf(" ambiguous");
+ match= !fca->ct->findchar_select ? -1 :
+ fca->ct->findchar_select(fca,results);
+ debugf(" resolved %s", match<0 ? "<none>" : context_names[match]);
+ }
+ if (match<0)
+ return 0;
+
+ *match_rx= results[match].rx;
+ if (match_rctxi) *match_rctxi= match;
+ return results[match].match;
+}
+
+static int findchar_select_text(const FindCharArgs *fca,
+ const FindCharResults results[]) {
+
+ dbassert(! results[ct_Digit].match ); /* digits are supposedly unambiguous */
+
+ switch (fca->ctxmap) {
+
+#define RETURN_IF_LONGER(this,that) do{ \
+ if (results[ct_##this].rx > results[ct_##that].rx) \
+ return ct_##this; \
+ }while(0)
+
+ case ctf_Digit | ctf_Upper | ctf_Lower | ctf_Word:
+ /* Start of word. Prefer Word match; failing that, take the longest */
+ if (results[ct_Word].match) return ct_Word;
+ RETURN_IF_LONGER(Lower,Upper);
+ RETURN_IF_LONGER(Upper,Lower);
+ break;
+
+ case ctf_Digit | ctf_Upper | ctf_Lower:
+ /* Mid-word. Prefer longer match; failing that, match lower. */
+ RETURN_IF_LONGER(Upper,Lower);
+ return ct_Lower;
+ }
+
+ /* oh well */
+ return -1;
}
-struct OcrCellTypeInfo {
- unsigned initial, nextword, midword;
- const char *name;
-};
const struct OcrCellTypeInfo ocr_celltype_number= {
- 4,4,4,
- .name= "number"
+ ctf_Digit, ctf_Digit, ctf_Digit,
+ .space_spaces= 5,
+ .name= "number",
+ .findchar_select= 0
};
const struct OcrCellTypeInfo ocr_celltype_text= {
- .initial=2, /* Uppercase */
- .nextword=3, /* Either */
- .midword=1, /* Lower only */
- .name= "text"
+ .initial= ctf_Digit | ctf_Upper,
+ .nextword= ctf_Digit | ctf_Upper | ctf_Lower | ctf_Word,
+ .midword= ctf_Digit | ctf_Upper | ctf_Lower,
+ .space_spaces= 4,
+ .name= "text",
+ .findchar_select= findchar_select_text
};
+
const char *ocr_celltype_name(OcrCellType ct) { return ct->name; }
-OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
+OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w,
+ const Pixcol cols[]) {
int nspaces;
- unsigned ctxmap;
- int ctxi, i, x;
+ int x;
+
+ FindCharArgs fca;
+ fca.rd= rd;
+ fca.ct= ct;
+ fca.w= w;
+ fca.cols= cols;
+ fca.x= -1;
restart:
nspaces=- w;
- ctxmap= ct->initial;
+ fca.ctxmap= ct->initial;
rd->nresults=0;
+ rd->lresult=0;
+ rd->result[0]=0;
debugf("OCR h=%d w=%d",rd->h,w);
- for (x=0; x<w; x++) debugf(" %"PSPIXCOL(PRIx),cols[x]);
+ for (x=0; x<w; x++) debugf(" "PIXCOL_PRFMT, PIXCOL_PRVAL(cols[x]));
debugf("\n");
debug_flush();
if (x>=w)
break;
- if (!cols[x]) {
+ if (!pixcol_nonzero(&cols[x])) {
nspaces++;
x++;
- if (nspaces==SPACE_SPACES) {
+ if (nspaces == ct->space_spaces) {
debugf("OCR x=%x nspaces=%d space\n",x,nspaces);
- ctxmap= ct->nextword;
+ fca.ctxmap= ct->nextword;
}
continue;
}
/* something here, so we need to add the spaces */
- if (nspaces>=SPACE_SPACES)
- add_result(rd," ",x-nspaces,x+1,0);
+ if (nspaces >= ct->space_spaces)
+ add_result(rd," ",x-nspaces,x+1,-1,0);
nspaces=0;
- /* find character */
- int lx=x;
+ fca.x= x;
- DatabaseNode *uniquematch= 0;
- int uniquematch_rx=-1;
+ int match_rx=-1;
+ int match_ctxi=-1;
+ DatabaseNode *match= findchar(&fca, &match_rx, &match_ctxi);
- debugf("OCR lx=%d ctxmap=%x ",lx,ctxmap);
-
- for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
- DatabaseNode *current= &rd->contexts[ctxi];;
- DatabaseNode *bestmatch= 0;
- int bestmatch_rx=-1;
-
- x= lx;
- if (!(ctxmap & (1u << ctxi))) continue;
- debugf(" || %s",context_names[ctxi]);
-
- for (;;) {
- debug_flush();
- debugf(" | x=%d",x);
- if (x>w) break;
- Pixcol cv= cols[x];
- debugf(" cv=%"PSPIXCOL(PRIx),cv);
- for (i=0; i<current->nlinks; i++)
- if (current->links[i].col == cv)
- goto found;
- /* not found */
- debugf(" ?");
- break;
-
- found:
- current= current->links[i].then;
- if (current->s[0]) {
- debugf(" \"%s\"%s",current->s,current->endsword?"_":"");
- bestmatch= current;
- bestmatch_rx= x;
- } else {
- debugf(" ...");
- }
-
- x++;
- }
-
- if (bestmatch) {
- if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) {
- debugf( " ambiguous");
- uniquematch= 0;
- break;
- }
- uniquematch= bestmatch;
- uniquematch_rx= bestmatch_rx;
- }
- }
-
- if (uniquematch) {
+ if (match) {
debugf(" || YES");
- add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap);
- x= uniquematch_rx+1;
- if (uniquematch->s[0]) ctxmap= ct->midword;
+ add_result(rd, match->str, x, match_rx, match_ctxi, fca.ctxmap);
+ if (should_reject(rd)) {
+ callout_unknown(rd, w,cols, match_rx+1,match_rx, 0);
+ goto restart;
+ }
+ x= match_rx+1;
+ if (match->match) fca.ctxmap= ct->midword;
else debugf(" (empty)");
- if (uniquematch->endsword) {
- nspaces= SPACE_SPACES;
+ if (match->endsword) {
+ nspaces= ct->space_spaces;
debugf("_");
- ctxmap= ct->nextword;
+ fca.ctxmap= ct->nextword;
}
debugf("\n");
} else {
int rx;
debugf(" || UNKNOWN");
- for (rx=lx; rx<w && cols[rx]; rx++);
- debugf(" x=%d ctxmap=%x %d..%d\n",x, ctxmap, lx,rx);
+ for (rx=x; rx<w && pixcol_nonzero(&cols[rx]); rx++);
+ debugf(" x=%d ctxmap=%x %d..%d\n",x, fca.ctxmap, x,rx);
debug_flush();
- callout_unknown(rd, w,cols, lx,rx-1, ctxmap);
+ callout_unknown(rd, w,cols, x,rx-1, fca.ctxmap);
goto restart;
}
+
}
- add_result(rd, 0,-1,-1,0);
+ add_result(rd, 0,-1,-1,-1,0);
debugf("OCR finished %d glyphs\n",rd->nresults);
debug_flush();
return rd->results;
OcrReader *ocr_init(int h) {
OcrReader *rd;
+ assert(h <= OCR_MAX_H);
+
+ 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;
+
+ rd->aresult= 10;
+ rd->result= mmalloc(rd->aresult);
+
readdb(rd);
return rd;
}
+
+/*---------- character set dump ----------*/
+
+static void show_recurse(const DatabaseNode *t, int *count,
+ const DatabaseNode **store_ary) {
+ if (t->defined) {
+ if (store_ary) store_ary[*count]= t;
+ (*count)++;
+ }
+ int l;
+ for (l=0; l<t->nlinks; l++)
+ show_recurse(t->links[l].then, count,store_ary);
+}
+
+static int show_char_compar(const void *av, const void *bv) {
+ const DatabaseNode *const *ap= av; const DatabaseNode *a= *ap;
+ const DatabaseNode *const *bp= bv; const DatabaseNode *b= *bp;
+ return strcmp(a->str, b->str) ?:
+ ((int)a->match - (int)b->match) ?:
+ ((int)a->endsword - (int)b->endsword) ?:
+ ((int)a->local - (int)b->local) ?:
+ 0;
+}
+
+void ocr_showcharsets(void) {
+ DIR *d;
+ struct dirent *de;
+ char found[OCR_MAX_H];
+ pcre *fnpat;
+ int matchvec[10];
+ char hbuf[10];
+ const char *pcre_err;
+ int pcre_erroffset;
+
+ memset(found,0,sizeof(found));
+
+ fnpat= pcre_compile("\\#(?:master|local)\\-char([1-9]\\d{0,2})\\#\\.txt$",
+ PCRE_ANCHORED|PCRE_DOLLAR_ENDONLY,
+ &pcre_err,&pcre_erroffset, 0);
+ debugf("pcre_compile %p %s\n",fnpat,pcre_err);
+ assert(fnpat);
+
+ sysassert( d= opendir(get_vardir()) );
+ for (;;) {
+ errno=0; de= readdir(d); if (!de) break;
+
+ int rer= pcre_exec(fnpat,0, de->d_name,strlen(de->d_name), 0,0,
+ matchvec,ARRAYSIZE(matchvec));
+ debugf("pcre_exec `%s' => %d\n", de->d_name,rer);
+
+ if (rer==PCRE_ERROR_NOMATCH || rer==PCRE_ERROR_BADUTF8) continue;
+ assert(rer==2);
+
+ rer= pcre_copy_substring(de->d_name,matchvec,rer, 1, hbuf,sizeof(hbuf));
+ debugf("pcre_copy_substring => %d\n", rer);
+ assert(rer>0);
+
+ int h= atoi(hbuf);
+ if (h >= ARRAYSIZE(found)) continue;
+
+ found[h]= 1;
+ }
+
+ int h;
+ for (h=0; h<ARRAYSIZE(found); h++) {
+ if (!found[h]) continue;
+
+ OcrReader *rd= ocr_init(h); /* we leak this but never mind */
+ progress("");
+
+ int ctxi;
+ for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
+ int nchars= 0;
+ show_recurse(&rd->contexts[ctxi], &nchars, 0);
+ const DatabaseNode **chars= mmalloc(sizeof(*chars) * nchars);
+ int chari= 0;
+ show_recurse(&rd->contexts[ctxi], &chari, chars);
+ assert(chari==nchars);
+ qsort(chars, nchars, sizeof(*chars), show_char_compar);
+
+ int local;
+ for (local=0; local<2; local++) {
+ printf("%2d %-6s %-6s ", h, context_names[ctxi],
+ local?"local":"master");
+ for (chari=0; chari<nchars; chari++) {
+ const DatabaseNode *t= chars[chari];
+ static const char accept[]=
+ "abcdefghijklmnopqrstuvwxyz"
+ "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+ "0123456789" "'!&-+.,>";
+
+ if (t->local != local) continue;
+
+ if (!t->match)
+ printf(" [nomatch]");
+ else if (!t->endsword && strspn(t->str, accept) == strlen(t->str))
+ printf(" %s",t->str);
+ else {
+ printf(" \"");
+ char *p= t->str;
+ int c;
+ while ((c=*p++)) {
+ if (c=='"' || c=='\\') printf("\\%c",c);
+ else if (c>=' ' && c<=126) putchar(c);
+ else printf("\\x%02x", (unsigned char)c);
+ }
+ if (t->endsword) putchar(' ');
+ putchar('"');
+ }
+ }
+ putchar('\n');
+ }
+ free(chars);
+ }
+ }
+}
~yarrgweb / ypp-sc-tools.db-live.git / blobdiff