Pass and use possible as well as actual contexts

[ypp-sc-tools.db-test.git] / pctb / ocr.c

diff --git a/pctb/ocr.c b/pctb/ocr.c
index fe264e8110ce8ffce6cbb6be0282c004151dec31..6448ee44d5e72bb275e17459a2863e3be68d7d2c 100644 (file)
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -42,6 +42,7 @@ typedef struct DatabaseNode {
 
 typedef struct {
   OcrReader *rd;
+  OcrCellType ct;
   int w;
   Pixcol *cols;
   int x;
@@ -52,30 +53,30 @@ typedef struct {
   int rx;
 } FindCharResults;
 
-static const char *context_names[]= {
-  "Lower",
-  "Upper",
-  "Digit"
+#define FOR_EACH_CONTEXT(EACH)                 \
+  EACH(Word)                                   \
+  EACH(Upper)                                  \
+  EACH(Lower)                                  \
+  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;
   int space_spaces;
   const char *name;
+  int (*findchar_select)(const FindCharArgs *fca,
+                        const FindCharResults results[]);
 };
-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]))
 
@@ -231,7 +232,7 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
   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");
@@ -243,7 +244,9 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
   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,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);
@@ -274,7 +277,7 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
 }
 
 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);
@@ -282,6 +285,7 @@ static void add_result(OcrReader *rd, const char *s, int l, int r,
   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++;
 }
@@ -322,7 +326,8 @@ static DatabaseNode *findchar_1ctx(const FindCharArgs *fca,
   return bestmatch;
 }  
 
-static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) {
+static DatabaseNode *findchar(const FindCharArgs *fca,
+                             int *match_rx, int *match_rctxi) {
   FindCharResults results[NCONTEXTS];
   int ctxi, match=-1, nmatches=0;
 
@@ -340,17 +345,48 @@ static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) {
     nmatches++;
   }
   if (nmatches==1) {
-    debugf( " unambiguous");
+    debugf(" unique");
   } else {
-    match=-1;
+    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[ctxi].rx;
-  return results[ctxi].match;
+  *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[]) {
+  if (fca->ctxmap != 017) return -1;
+
+  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;
+  return -1;
+}
+
+const struct OcrCellTypeInfo ocr_celltype_number= {
+  ctf_Digit, ctf_Digit, ctf_Digit,
+  .space_spaces= 5,
+  .name= "number",
+  .findchar_select= 0
+};
+const struct OcrCellTypeInfo ocr_celltype_text= {
+  .initial=  ctf_Digit | ctf_Upper,
+  .nextword= ctf_Digit | ctf_Upper | ctf_Lower | ctf_Word,
+  .midword=  ctf_Digit | 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[]) {
@@ -359,6 +395,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
 
   FindCharArgs fca;
   fca.rd= rd;
+  fca.ct= ct;
   fca.w= w;
   fca.cols= cols;
   fca.x= -1;
@@ -392,17 +429,18 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
 
     /* something here, so we need to add the spaces */
     if (nspaces >= ct->space_spaces)
-      add_result(rd," ",x-nspaces,x+1,0);
+      add_result(rd," ",x-nspaces,x+1,-1,0);
     nspaces=0;
 
     fca.x= x;
 
     int match_rx=-1;
-    DatabaseNode *match= findchar(&fca, &match_rx);
+    int match_ctxi=-1;
+    DatabaseNode *match= findchar(&fca, &match_rx, &match_ctxi);
     
     if (match) {
       debugf(" || YES");
-      add_result(rd, match->str, x, match_rx, fca.ctxmap);
+      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)");
@@ -423,7 +461,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
     }
 
   }
-  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;