X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?a=blobdiff_plain;f=pctb%2Focr.c;h=f6104452dc1c1854e6ccd19aac77882caae0aba9;hb=ac65228e40fa375c829b46607fb4941ff11376e9;hp=927b93a97894e41949799a773adc0644a298f217;hpb=5235400ea0618cdc6e4585f9b032edbc000ad8e8;p=ypp-sc-tools.db-test.git

diff --git a/pctb/ocr.c b/pctb/ocr.c
index 927b93a..f610445 100644
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -26,46 +26,49 @@
  */
 
 #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;
 
+typedef struct {
+  OcrReader *rd;
+  OcrCellType ct;
+  int w;
+  Pixcol *cols;
+  int x;
+  unsigned ctxmap;
+} FindCharArgs;
+typedef struct {
+  DatabaseNode *match;
+  int rx;
+} FindCharResults;
+
+enum { ct_Lower, ct_Upper, ct_Word, ct_Digit };
 static const char *context_names[]= {
-  "Lower",
-  "Upper",
-  "Digit"
+  "Lower",  /*  bit 0, value 001 */
+  "Upper",  /*  bit 1, value 002 */
+  "Word",   /*  bit 2, value 004 */
+  "Digit",  /*  bit 3, value 010 */
 };
 struct OcrCellTypeInfo {
   /* bitmaps of indices into context_names: */
   unsigned initial, nextword, midword;
   int space_spaces;
   const char *name;
-};
-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"
+  int (*findchar_select)(const FindCharArgs *fca,
+			 const FindCharResults results[]);
 };
 
-
 #define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
 
 struct OcrReader {
@@ -81,26 +84,35 @@ DEBUG_DEFINE_DEBUGF(ocr)
 
 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);
+
 static void readdb(OcrReader *rd) {
+  int ctxi;
+  
+  for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
+    cleardb_node(&rd->contexts[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; 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);
   
   if (!dbfile_open(dbfname))
     goto x;
@@ -126,23 +138,15 @@ 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(nchrs<MAXGLYPHCHRS);
-      if (c=='\\') {
-	unsigned cr;
-	c= fgetc(dbfile); sysassert(!ferror(dbfile)); dbassert(c=='x');
-	dbassert( dbfile_scanf("%2x", &cr) == 1);
-	assert(cr>0 && cr<=255);
-	c= cr;
-      }
+      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 (;;) {
@@ -159,7 +163,10 @@ 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) {
@@ -176,9 +183,20 @@ 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();
@@ -261,17 +279,117 @@ 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(" 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[ctxi].rx;
+  return results[ctxi].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= {
+  010,010,010,
+  .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=  014, /* Digit|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[]) {
   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;
   debugf("OCR h=%d w=%d",rd->h,w);
   for (x=0; x<w; x++) debugf(" %"PSPIXCOL(PRIx),cols[x]);
@@ -290,7 +408,7 @@ 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;
     }
@@ -300,81 +418,33 @@ 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->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, 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);
@@ -385,6 +455,13 @@ 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;