WIP new dictionary arrangements

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

diff --git a/pctb/ocr.c b/pctb/ocr.c
index 927b93a97894e41949799a773adc0644a298f217..2c3a16772ec8ac0a90f5a9be420c9db188c1e6da 100644 (file)
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -26,18 +26,17 @@
  */
 
 #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;
 
@@ -81,26 +80,35 @@ 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 +134,15 @@ 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 +159,10 @@ 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 +179,20 @@ 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();
@@ -332,8 +346,8 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
 
       found:
        current= current->links[i].then;
-       if (current->s[0]) {
-         debugf(" \"%s\"%s",current->s,current->endsword?"_":"");
+       if (current->match) {
+         debugf(" \"%s\"%s",current->str,current->endsword?"_":"");
          bestmatch= current;
          bestmatch_rx= x;
        } else {
@@ -344,7 +358,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
       }
       
       if (bestmatch) {
-       if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) {
+       if (uniquematch && strcmp(bestmatch->str, uniquematch->str)) {
          debugf( " ambiguous");
          uniquematch= 0;
          break;
@@ -356,9 +370,9 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
 
     if (uniquematch) {
       debugf(" || YES");
-      add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap);
+      add_result(rd, uniquematch->str, lx, uniquematch_rx, ctxmap);
       x= uniquematch_rx+1;
-      if (uniquematch->s[0]) ctxmap= ct->midword;
+      if (uniquematch->match) ctxmap= ct->midword;
       else debugf(" (empty)");
       if (uniquematch->endsword) {
        nspaces= ct->space_spaces;
@@ -385,6 +399,13 @@ 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;