X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.db-live.git;a=blobdiff_plain;f=pctb%2Focr.c;h=fe264e8110ce8ffce6cbb6be0282c004151dec31;hp=5828bd0f830e5b3fa3f192fb421a039c5df57311;hb=cf8e9b2a56d12305cd61f8d4a9284d4319265215;hpb=6a3c0962283d32bc6e5f6c47c929baf37ddc642f

diff --git a/pctb/ocr.c b/pctb/ocr.c
index 5828bd0..fe264e8 100644
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -40,6 +40,18 @@ typedef struct DatabaseNode {
   DatabaseLink *links;
 } DatabaseNode;
 
+typedef struct {
+  OcrReader *rd;
+  int w;
+  Pixcol *cols;
+  int x;
+  unsigned ctxmap;
+} FindCharArgs;
+typedef struct {
+  DatabaseNode *match;
+  int rx;
+} FindCharResults;
+
 static const char *context_names[]= {
   "Lower",
   "Upper",
@@ -275,17 +287,86 @@ static void add_result(OcrReader *rd, const char *s, int l, int r,
 }
 
 
+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=%"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->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) {
+  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( " unambiguous");
+  } else {
+    match=-1;
+  }
+  if (match<0)
+    return 0;
+  
+  *match_rx= results[ctxi].rx;
+  return results[ctxi].match;
+}
+
 const char *ocr_celltype_name(OcrCellType ct) { return ct->name; }
 
 OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
   int nspaces;
-  unsigned ctxmap;
-  int ctxi, i, x;
+  int x;
+
+  FindCharArgs fca;
+  fca.rd= rd;
+  fca.w= w;
+  fca.cols= cols;
+  fca.x= -1;
 
  restart:
 
   nspaces=- w;
-  ctxmap= ct->initial;
+  fca.ctxmap= ct->initial;
   rd->nresults=0;
   debugf("OCR h=%d w=%d",rd->h,w);
   for (x=0; x<w; x++) debugf(" %"PSPIXCOL(PRIx),cols[x]);
@@ -304,7 +385,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
       x++;
       if (nspaces == ct->space_spaces) {
 	debugf("OCR  x=%x nspaces=%d space\n",x,nspaces);
-	ctxmap= ct->nextword;
+	fca.ctxmap= ct->nextword;
       }
       continue;
     }
@@ -314,81 +395,33 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
       add_result(rd," ",x-nspaces,x+1,0);
     nspaces=0;
 
-    /* find character */
-    int lx=x;
+    fca.x= x;
 
-    DatabaseNode *uniquematch= 0;
-    int uniquematch_rx=-1;
+    int match_rx=-1;
+    DatabaseNode *match= findchar(&fca, &match_rx);
     
-    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->match) {
-	  debugf(" \"%s\"%s",current->str,current->endsword?"_":"");
-	  bestmatch= current;
-	  bestmatch_rx= x;
-	} else {
-	  debugf(" ...");
-	}
-
-	x++;
-      }
-      
-      if (bestmatch) {
-	if (uniquematch && strcmp(bestmatch->str, uniquematch->str)) {
-	  debugf( " ambiguous");
-	  uniquematch= 0;
-	  break;
-	}
-	uniquematch= bestmatch;
-	uniquematch_rx= bestmatch_rx;
-      }
-    }
-
-    if (uniquematch) {
+    if (match) {
       debugf(" || YES");
-      add_result(rd, uniquematch->str, lx, uniquematch_rx, ctxmap);
-      x= uniquematch_rx+1;
-      if (uniquematch->match) ctxmap= ct->midword;
+      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 (uniquematch->endsword) {
+      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 && 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);
   debugf("OCR  finished %d glyphs\n",rd->nresults);
@@ -400,8 +433,8 @@ OcrReader *ocr_init(int h) {
   OcrReader *rd;
 
   if (o_flags & ff_dict_fetch) {
-    char *fetchfile= masprintf("master-char%d",rd->h);
-    progress("Updating %s",fetchfile);
+    char *fetchfile= masprintf("char%d",h);
+    progress("Updating %s...",fetchfile);
     fetch_with_rsync(fetchfile);
     free(fetchfile);
   }