X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?a=blobdiff_plain;f=pctb%2Focr.c;h=f4bc77d56118ae8b4caebfce168a3336cd2eecc2;hb=74e4e249f2c3e848592984cb193aded6a77a341d;hp=fd4cba9fe841e1c6a439f34edc741eae31d44c5c;hpb=f2c5d45ada202f2ad2640b260cac2fd7bb83eb3e;p=ypp-sc-tools.db-live.git

diff --git a/pctb/ocr.c b/pctb/ocr.c
index fd4cba9..f4bc77d 100644
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -46,10 +46,27 @@ static const char *context_names[]= {
   "Upper",
   "Digit"
 };
+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"
+};
 
-#define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
 
-#define SPACE_SPACES 4
+#define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
 
 struct OcrReader {
   int h;
@@ -69,17 +86,7 @@ DEBUG_DEFINE_DEBUGF(ocr)
      fatal("Error in character set database.\n"				\
 	   " Requirement not met: %s:%d: %s", __FILE__,__LINE__, #x))
 
-static void fgetsline(FILE *f, char *lbuf, size_t lbufsz) {
-  errno=0;
-  char *s= fgets(lbuf,lbufsz,f);
-  sysassert(!ferror(f));
-  dbassert(!feof(f));
-  assert(s);
-  int l= strlen(lbuf);
-  dbassert(l>0);  dbassert(lbuf[--l]='\n');
-  lbuf[l]= 0;
-}
-#define FGETSLINE(f,buf) (fgetsline(f,buf,sizeof(buf)))
+#define FGETSLINE (fgetsline(db,lbuf,sizeof(lbuf)))
 
 static void cleardb_node(DatabaseNode *n) {
   int i;
@@ -112,7 +119,7 @@ static void readdb(OcrReader *rd) {
     return;
   }
 
-  FGETSLINE(db,lbuf);
+  FGETSLINE;
   dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1"));
 
   r= fscanf(db, "%d", &h);
@@ -120,7 +127,7 @@ static void readdb(OcrReader *rd) {
   dbassert(h==rd->h);
 
   for (;;) {
-    FGETSLINE(db,lbuf);
+    FGETSLINE;
     if (!lbuf || lbuf[0]=='#') continue;
     if (!strcmp(lbuf,".")) break;
 
@@ -128,7 +135,7 @@ static void readdb(OcrReader *rd) {
       if (!strcmp(lbuf,context_names[ctxi]))
 	goto found_ctx;
     /* not found, just skip */
-    for (;;) { FGETSLINE(db,lbuf); if (!lbuf[0]) break; }
+    for (;;) { FGETSLINE; if (!lbuf[0]) break; }
     continue;
 
   found_ctx:
@@ -154,7 +161,7 @@ static void readdb(OcrReader *rd) {
 
     current= &rd->contexts[ctxi];
     for (;;) {
-      FGETSLINE(db,lbuf);
+      FGETSLINE;
       if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; }
       char *ep;
       cv= strtoul(lbuf,&ep,16);  dbassert(!*ep);
@@ -211,6 +218,11 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
   const char *p;
   char cb;
   Pixcol pv;
+
+  if (!o_resolver)
+    fatal("OCR failed - unrecognised characters or ligatures.\n"
+	  "Character set database needs to be updated or augmented.\n"
+	  "See README.charset.\n");
   
   if (!resolver) {
     sysassert(! pipe(jobpipe) );
@@ -224,11 +236,11 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
       /* we know donepipe[1] is >= 4 and we have dealt with all the others
        * so we aren't in any danger of overwriting some other fd 4: */
       sysassert( dup2(donepipe[1],4) ==4 );
-      execlp("./yppsc-ocr-resolver", "yppsc-ocr-resolver",
+      execlp(o_resolver, o_resolver,
 	     DEBUGP(callout) ? "--debug" : "--noop-arg",
 	     "--automatic-1",
 	     (char*)0);
-      sysassert(!"execlp failed");
+      sysassert(!"execlp ocr-resolver failed");
     }
     sysassert(! close(jobpipe[0]) );
     sysassert(! close(donepipe[1]) );
@@ -278,27 +290,10 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
   }
 
   if (r==0) {
-    pid_t pid;
-    int st;
-    for (;;) {
-      pid= waitpid(resolver_pid, &st, 0);
-      if (pid==-1) { sysassert(errno==EINTR); continue; }
-      break;
-    }
-    sysassert(pid==resolver_pid);
-    if (WIFEXITED(st)) {
-      if (WEXITSTATUS(st))
-	fatal("character resolver failed with nonzero exit status %d",
-	      WEXITSTATUS(st));
-      fclose(resolver);
-      close(resolver_done);
-      resolver= 0;
-    } else if (WIFSIGNALED(st)) {
-      fatal("character resolver died due to signal %s%s",
-	    strsignal(WTERMSIG(st)), WCOREDUMP(st)?" (core dumped)":"");
-    } else {
-      fatal("character resolver gave strange wait status %d",st);
-    }
+    waitpid_check_exitstatus(resolver_pid, "character resolver");
+    fclose(resolver);
+    close(resolver_done);
+    resolver= 0;
   } else {
     assert(r==1);
     sysassert(cb==0);
@@ -320,20 +315,6 @@ static void add_result(OcrReader *rd, const char *s, int l, int r,
   rd->nresults++;
 }
 
-struct OcrCellTypeInfo {
-  unsigned initial, nextword, midword;
-  const char *name;
-};
-const struct OcrCellTypeInfo ocr_celltype_number= {
-  4,4,4,
-  .name= "number"
-};
-const struct OcrCellTypeInfo ocr_celltype_text= {
-  .initial=2, /* Uppercase */
-  .nextword=3, /* Either */
-  .midword=1, /* Lower only */
-  .name= "text"
-};
 
 const char *ocr_celltype_name(OcrCellType ct) { return ct->name; }
 
@@ -362,7 +343,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
     if (!cols[x]) {
       nspaces++;
       x++;
-      if (nspaces==SPACE_SPACES) {
+      if (nspaces == ct->space_spaces) {
 	debugf("OCR  x=%x nspaces=%d space\n",x,nspaces);
 	ctxmap= ct->nextword;
       }
@@ -370,7 +351,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
     }
 
     /* something here, so we need to add the spaces */
-    if (nspaces>=SPACE_SPACES)
+    if (nspaces >= ct->space_spaces)
       add_result(rd," ",x-nspaces,x+1,0);
     nspaces=0;
 
@@ -435,7 +416,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
       if (uniquematch->s[0]) ctxmap= ct->midword;
       else debugf(" (empty)");
       if (uniquematch->endsword) {
-	nspaces= SPACE_SPACES;
+	nspaces= ct->space_spaces;
 	debugf("_");
 	ctxmap= ct->nextword;
       }