typedef struct DatabaseNode {
char *str;
int nlinks, alinks;
- unsigned match:1, defined:1, endsword:1;
+ unsigned match:1, defined:1, endsword:1, local:1;
DatabaseLink *links;
} DatabaseNode;
int rx;
} FindCharResults;
-enum { ct_Lower, ct_Upper, ct_Word, ct_Digit };
-static const char *context_names[]= {
- "Lower", /* bit 0, value 001 */
- "Upper", /* bit 1, value 002 */
- "Word", /* bit 2, value 004 */
- "Digit", /* bit 3, value 010 */
+#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 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;
cleardb_node(n->links[i].then);
}
-static void readdb1(OcrReader *rd, const char *which);
+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");
- readdb1(rd, "local");
+ readdb1(rd, "master", 0);
+ readdb1(rd, "local", 1);
}
-static void readdb1(OcrReader *rd, const char *which) {
+static void readdb1(OcrReader *rd, const char *which, int local) {
int nchrs;
DatabaseNode *current, *additional;
char chrs[100];
current->str= 0;
current->defined= 1;
current->match= 0;
+ current->local= local;
if (nchrs) {
current->str= mmalloc(nchrs+1);
Pixcol pv;
FILE *resolver= resolve_start();
- if (!resolver)
+ if (!resolver || !(o_flags & ff_editcharset))
fatal("OCR failed - unrecognised characters or ligatures.\n"
"Character set database needs to be updated or augmented.\n"
"See README.charset.\n");
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(resolver, 1u << s->ctxi);
+ 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);
readdb(rd);
}
-static void add_result(OcrReader *rd, const char *s, int l, int r, int ctxi) {
+static void add_result(OcrReader *rd, const char *s, int l, int r,
+ 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].ctxi= ctxi;
+ rd->results[rd->nresults].match= match;
+ rd->results[rd->nresults].ctxmap= ctxmap;
rd->nresults++;
}
static int findchar_select_text(const FindCharArgs *fca,
const FindCharResults results[]) {
- if (fca->ctxmap != 017) return -1;
+
+ dbassert(! results[ct_Digit].match ); /* digits are supposedly unambiguous */
+
+ switch (fca->ctxmap) {
- dbassert(! results[ct_Digit].match );
- if (results[ct_Word].match) return ct_Word;
- if (results[ct_Lower].rx > results[ct_Upper].rx) return ct_Lower;
- if (results[ct_Upper].rx > results[ct_Lower].rx) return ct_Upper;
+#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;
}
const struct OcrCellTypeInfo ocr_celltype_number= {
- 010,010,010,
+ ctf_Digit, ctf_Digit, ctf_Digit,
.space_spaces= 5,
.name= "number",
.findchar_select= 0
};
const struct OcrCellTypeInfo ocr_celltype_text= {
- .initial= 012, /* Digit|Upper */
- .nextword= 017, /* Digit|Upper|Lower|Word */
- .midword= 011, /* Digit|Lower */
+ .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
/* something here, so we need to add the spaces */
if (nspaces >= ct->space_spaces)
- add_result(rd," ",x-nspaces,x+1,-1);
+ add_result(rd," ",x-nspaces,x+1,-1,0);
nspaces=0;
fca.x= x;
if (match) {
debugf(" || YES");
- add_result(rd, match->str, x, match_rx, match_ctxi);
+ add_result(rd, match->str, x, match_rx, match_ctxi, fca.ctxmap);
x= match_rx+1;
if (match->match) fca.ctxmap= ct->midword;
else debugf(" (empty)");
}
}
- add_result(rd, 0,-1,-1,-1);
+ add_result(rd, 0,-1,-1,-1,0);
debugf("OCR finished %d glyphs\n",rd->nresults);
debug_flush();
return rd->results;
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[32];
+ 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);
+ }
+ }
+}