X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?a=blobdiff_plain;f=pctb%2Focr.c;h=5fdc4575ec4f64540129d00055e3847478f36a4f;hb=89dfaeec1540f73ba85dbd25dd5332416f98778e;hp=fd4fcf4f8de99642291fb3e9c5410a2884c642fb;hpb=8787ee59f6840de63bac432b516a30d0dfe22c84;p=ypp-sc-tools.main.git

diff --git a/pctb/ocr.c b/pctb/ocr.c
index fd4fcf4..5fdc457 100644
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -8,22 +8,25 @@ typedef struct {
   struct DatabaseNode *then;
 } DatabaseLink;
 
-#define MAXGLYPHCHRS 3
+#define MAXGLYPHCHRS 7
 
 typedef struct DatabaseNode {
   char s[MAXGLYPHCHRS+1]; /* null-terminated; "" means no match here */
   int nlinks, alinks;
+  unsigned endsword:1;
   DatabaseLink *links;
 } DatabaseNode;
 
 static const char *context_names[]= {
   "Lower",
   "Upper",
-/*  "Digit"*/
+  "Digit"
 };
 
 #define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
 
+#define SPACE_SPACES 4
+
 struct OcrReader {
   int h;
   DatabaseNode contexts[NCONTEXTS];
@@ -35,11 +38,21 @@ static FILE *resolver;
 static pid_t resolver_pid;
 static int resolver_done;
 
+DEBUG_DEFINE_DEBUGF(ocr)
+
+#define dbassert(x)							\
+  ((x) ? (void)0 :							\
+     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);
-  eassert(s);
+  sysassert(!ferror(f));
+  dbassert(!feof(f));
+  assert(s);
   int l= strlen(lbuf);
-  eassert(l>0);  eassert(lbuf[--l]='\n');
+  dbassert(l>0);  dbassert(lbuf[--l]='\n');
   lbuf[l]= 0;
 }
 #define FGETSLINE(f,buf) (fgetsline(f,buf,sizeof(buf)))
@@ -57,26 +70,30 @@ static void readdb(OcrReader *rd) {
   char chrs[MAXGLYPHCHRS+1];
   Pixcol cv;
   int r,j,ctxi;
-  int h;
+  int h, endsword;
   char lbuf[100];
   FILE *db;
 
+  for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
+    cleardb_node(&rd->contexts[ctxi]);
+
   char *dbfname=0;
   asprintf(&dbfname,"%s/charset-%d.txt",get_vardir(),rd->h);
-  eassert(dbfname);
+  sysassert(dbfname);
   
-  db= fopen(dbfname,"r");  eassert(db);
+  db= fopen(dbfname,"r");
   free(dbfname);
+  if (!db) {
+    sysassert(errno==ENOENT);
+    return;
+  }
 
   FGETSLINE(db,lbuf);
-  eassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1"));
+  dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1"));
 
   r= fscanf(db, "%d", &h);
-  eassert(r==1);
-  eassert(h==rd->h);
-
-  for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
-    cleardb_node(&rd->contexts[ctxi]);
+  dbassert(r==1);
+  dbassert(h==rd->h);
 
   for (;;) {
     FGETSLINE(db,lbuf);
@@ -92,27 +109,32 @@ static void readdb(OcrReader *rd) {
 
   found_ctx:
     for (nchrs=0;;) {
-      int c= fgetc(db);  eassert(c!=EOF);
-      if (c=='\n') { eassert(nchrs); break; }
-      eassert(nchrs<MAXGLYPHCHRS);
+      int c= fgetc(db); sysassert(!ferror(db)); dbassert(c!=EOF);
+      if (c=='\n') { dbassert(nchrs); break; }
+      dbassert(nchrs<MAXGLYPHCHRS);
       if (c=='\\') {
 	unsigned cr;
-	c= fgetc(db);  eassert(c=='x');
-	r= fscanf(db, "%2x", &cr);  eassert(r==1);
+	c= fgetc(db); sysassert(!ferror(db)); dbassert(c=='x');
+	r= fscanf(db, "%2x", &cr); sysassert(!ferror(db)); dbassert(r==1);
 	assert(cr>0 && cr<=255);
 	c= cr;
       }
       chrs[nchrs++]= c;
     }
+    endsword= 0;
+    if (nchrs>1 && chrs[nchrs-1]==' ') {
+      endsword= 1;
+      nchrs--;
+    }
     chrs[nchrs]= 0;
 
     current= &rd->contexts[ctxi];
     for (;;) {
       FGETSLINE(db,lbuf);
-      if (!lbuf[0]) { eassert(current != &rd->contexts[ctxi]); break; }
+      if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; }
       char *ep;
-      cv= strtoul(lbuf,&ep,16);  eassert(!*ep);
-      eassert(!(cv & ~((1UL << rd->h)-1)));
+      cv= strtoul(lbuf,&ep,16);  dbassert(!*ep);
+      dbassert(!(cv & ~((1UL << rd->h)-1)));
       
       for (j=0; j<current->nlinks; j++)
 	if (current->links[j].col == cv) {
@@ -120,16 +142,15 @@ static void readdb(OcrReader *rd) {
 	  goto found_link;
 	}
 
-      additional= malloc(sizeof(*additional)); eassert(additional);
+      additional= mmalloc(sizeof(*additional));
       additional->s[0]= 0;
       additional->nlinks= additional->alinks= 0;
       additional->links= 0;
       if (current->nlinks==current->alinks) {
 	current->alinks++;
 	current->alinks<<=1;
-	current->links= realloc(current->links,
-				sizeof(*current->links) * current->alinks);
-	eassert(current->links);
+	current->links= mrealloc(current->links,
+				 sizeof(*current->links) * current->alinks);
       }
       current->links[current->nlinks].col= cv;
       current->links[current->nlinks].then= additional;
@@ -139,54 +160,64 @@ static void readdb(OcrReader *rd) {
     found_link:;
     }
 
-    eassert(!current->s[0]);
+    dbassert(!current->s[0]);
     strcpy(current->s, chrs);
+    current->endsword= endsword;
+  }
+  sysassert(!ferror(db));
+  sysassert(!fclose(db));
+}
+
+static void cu_pr_ctxmap(unsigned ctxmap) {
+  fprintf(resolver,"{");
+  const char *spc="";
+  int ctxi;
+  for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
+    if (!(ctxmap & (1u << ctxi))) continue;
+    fprintf(resolver,"%s%s",spc,context_names[ctxi]);
+    spc=" ";
   }
-  eassert(!ferror(db));
-  eassert(!fclose(db));
-}      
+  fprintf(resolver,"}");
+}
 
 static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
 			    int unk_l, int unk_r, unsigned unk_ctxmap) {
-  int jobpipe[2],donepipe[2], c, r,i, x,y;
+  int jobpipe[2],donepipe[2], c,i, x,y;
   const OcrResultGlyph *s;
   const char *p;
   char cb;
   Pixcol pv;
   
   if (!resolver) {
-    r= pipe(jobpipe);  eassert(!r);
-    r= pipe(donepipe);  eassert(!r);
+    sysassert(! pipe(jobpipe) );
+    sysassert(! pipe(donepipe) );
     resolver_pid= fork();
-    eassert(resolver_pid!=-1);
+    sysassert(resolver_pid!=-1);
     if (!resolver_pid) {
-      r= dup2(jobpipe[0],0); eassert(r==0);
-      r= close(jobpipe[1]); eassert(!r);
-      r= close(donepipe[0]); eassert(!r);
+      sysassert( dup2(jobpipe[0],0) ==0 );
+      sysassert(! close(jobpipe[1]) );
+      sysassert(! close(donepipe[0]) );
       /* 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: */
-      r= dup2(donepipe[1],4); eassert(r==4);
-      execlp("./show-thing.tcl", "./show-thing.tcl",
-	     "--automatic","1",(char*)0);
-      eassert(!"execlp failed");
+      sysassert( dup2(donepipe[1],4) ==4 );
+      execlp("./yppsc-ocr-resolver", "yppsc-ocr-resolver",
+	     DEBUGP(callout) ? "--debug" : "--noop-arg",
+	     "--automatic-1",
+	     (char*)0);
+      sysassert(!"execlp failed");
     }
-    r= close(jobpipe[0]); eassert(!r);
-    r= close(donepipe[1]); eassert(!r);
-    resolver= fdopen(jobpipe[1],"w"); eassert(resolver);
+    sysassert(! close(jobpipe[0]) );
+    sysassert(! close(donepipe[1]) );
+    resolver= fdopen(jobpipe[1],"w"); sysassert(resolver);
     resolver_done= donepipe[0];
   }
-  fprintf(resolver,"%d %d {",unk_l,unk_r);
-  const char *spc="";
-  int ctxi;
-  for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
-    if (!(unk_ctxmap & (1u << ctxi))) continue;
-    fprintf(resolver,"%s%s",spc,context_names[ctxi]);
-    spc=" ";
-  }
-  fprintf(resolver,"}");
+  fprintf(resolver,"%d %d ",unk_l,unk_r);
+  cu_pr_ctxmap(unk_ctxmap);
   for (i=0, s=rd->results; i<rd->nresults; i++, s++) {
     if (!strcmp(s->s," ")) continue;
-    fprintf(resolver," %d %d %s ",s->l,s->r,context_names[s->ctx]);
+    fprintf(resolver," %d %d ",s->l,s->r);
+    cu_pr_ctxmap(s->ctxmap);
+    fprintf(resolver," ");
     for (p=s->s; (c= *p); p++) {
       if (c=='\\') fprintf(resolver,"\\%c",c);
       else if (c>=33 && c<=126) fputc(c,resolver);
@@ -210,69 +241,93 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
     fputs("\",\n",resolver);
   }
   fputs("};\n",resolver);
-  eassert(!ferror(resolver));
-  eassert(!fflush(resolver));
+  sysassert(!ferror(resolver));
+  sysassert(!fflush(resolver));
 
-  eassert(resolver);
+  sysassert(resolver);
 
+  int r;
   for (;;) {
     r= read(resolver_done,&cb,1);
-    if (r==-1) { eassert(errno==EINTR); continue; }
+    if (r==-1) { sysassert(errno==EINTR); continue; }
     break;
   }
 
   if (r==0) {
     pid_t pid;
+    int st;
     for (;;) {
-      pid= waitpid(resolver_pid, &r, 0);
-      if (pid==-1) { eassert(errno==EINTR); continue; }
+      pid= waitpid(resolver_pid, &st, 0);
+      if (pid==-1) { sysassert(errno==EINTR); continue; }
       break;
     }
-    eassert(pid==resolver_pid);
-    if (WIFEXITED(r)) {
-      eassert(!WEXITSTATUS(r));
+    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(r)) {
-      eassert(!"resolver child died due to signal");
+    } else if (WIFSIGNALED(st)) {
+      fatal("character resolver died due to signal %s%s",
+	    strsignal(WTERMSIG(st)), WCOREDUMP(st)?" (core dumped)":"");
     } else {
-      eassert(!"weird wait status");
+      fatal("character resolver gave strange wait status %d",st);
     }
   } else {
-    eassert(r==1);
-    eassert(cb==0);
+    assert(r==1);
+    sysassert(cb==0);
   }
 
   readdb(rd);
 }
 
-static void add_result(OcrReader *rd, const char *s, int l, int r, int ctx) {
+static void add_result(OcrReader *rd, const char *s, int l, int r,
+		       unsigned ctxmap) {
   if (rd->nresults >= rd->aresults) {
     rd->aresults++; rd->aresults<<=1;
-    rd->results= realloc(rd->results,sizeof(*rd->results)*rd->aresults);
-    eassert(rd->results);
+    rd->results= mrealloc(rd->results, sizeof(*rd->results)*rd->aresults);
   }
   rd->results[rd->nresults].s= s;
   rd->results[rd->nresults].l= l;
   rd->results[rd->nresults].r= r;
-  rd->results[rd->nresults].ctx= ctx;
+  rd->results[rd->nresults].ctxmap= ctxmap;
   rd->nresults++;
 }
 
-OcrResultGlyph *ocr(OcrReader *rd, int w, Pixcol cols[]) {
-  int nspaces=-w;
-  unsigned ctxmap=2; /* uppercase */
+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; }
+
+OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
+  int nspaces;
+  unsigned ctxmap;
   int ctxi, i, x;
 
-  rd->nresults=0;
+ restart:
 
-  fprintf(debug,"OCR h=%d w=%d",rd->h,w);
-  for (x=0; x<w; x++) fprintf(debug," %"PSPIXCOL(PRIx),cols[x]);
-  fprintf(debug,"\n");
+  nspaces=- w;
+  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]);
+  debugf("\n");
   debug_flush();
 
- restart:
   x=0;
   for (;;) {
     debug_flush();
@@ -283,22 +338,25 @@ OcrResultGlyph *ocr(OcrReader *rd, int w, Pixcol cols[]) {
     if (!cols[x]) {
       nspaces++;
       x++;
-      if (nspaces==3) {
-	fprintf(debug,"OCR  x=%x nspaces=%d space\n",x,nspaces);
-	add_result(rd," ",x-nspaces,x+1,0);
-	ctxmap=3; /* either */
+      if (nspaces==SPACE_SPACES) {
+	debugf("OCR  x=%x nspaces=%d space\n",x,nspaces);
+	ctxmap= ct->nextword;
       }
       continue;
     }
+
+    /* something here, so we need to add the spaces */
+    if (nspaces>=SPACE_SPACES)
+      add_result(rd," ",x-nspaces,x+1,0);
     nspaces=0;
 
     /* find character */
     int lx=x;
 
     DatabaseNode *uniquematch= 0;
-    int uniquematch_rx=-1, uniquematch_ctxi=-1;
+    int uniquematch_rx=-1;
     
-    fprintf(debug,"OCR  lx=%d ctxmap=%x  ",lx,ctxmap);
+    debugf("OCR  lx=%d ctxmap=%x  ",lx,ctxmap);
 
     for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
       DatabaseNode *current= &rd->contexts[ctxi];;
@@ -307,63 +365,69 @@ OcrResultGlyph *ocr(OcrReader *rd, int w, Pixcol cols[]) {
 
       x= lx;
       if (!(ctxmap & (1u << ctxi))) continue;
-      fprintf(debug," || %s",context_names[ctxi]);
+      debugf(" || %s",context_names[ctxi]);
 
       for (;;) {
 	debug_flush();
-	fprintf(debug," | x=%d",x);
+	debugf(" | x=%d",x);
 	if (x>w) break;
 	Pixcol cv= cols[x];
-	fprintf(debug," cv=%"PSPIXCOL(PRIx),cv);
+	debugf(" cv=%"PSPIXCOL(PRIx),cv);
 	for (i=0; i<current->nlinks; i++)
 	  if (current->links[i].col == cv)
 	    goto found;
 	/* not found */
-	fprintf(debug," ?");
+	debugf(" ?");
 	break;
 
       found:
 	current= current->links[i].then;
 	if (current->s[0]) {
-	  fprintf(debug," \"%s\"",current->s);
+	  debugf(" \"%s\"%s",current->s,current->endsword?"_":"");
 	  bestmatch= current;
 	  bestmatch_rx= x;
 	} else {
-	  fprintf(debug," ...");
+	  debugf(" ...");
 	}
 
 	x++;
       }
       
       if (bestmatch) {
-	if (uniquematch) {
-	  fprintf(debug, " ambiguous");
+	if (uniquematch && strcmp(bestmatch->s, uniquematch->s)) {
+	  debugf( " ambiguous");
 	  uniquematch= 0;
 	  break;
 	}
 	uniquematch= bestmatch;
 	uniquematch_rx= bestmatch_rx;
-	uniquematch_ctxi= ctxi;
       }
     }
 
     if (uniquematch) {
-      fprintf(debug," || YES\n");
-      add_result(rd, uniquematch->s, lx, uniquematch_rx, uniquematch_ctxi);
+      debugf(" || YES");
+      add_result(rd, uniquematch->s, lx, uniquematch_rx, ctxmap);
       x= uniquematch_rx+1;
-      ctxmap= 1; /* Lower only */
+      if (uniquematch->s[0]) ctxmap= ct->midword;
+      else debugf(" (empty)");
+      if (uniquematch->endsword) {
+	nspaces= SPACE_SPACES;
+	debugf("_");
+	ctxmap= ct->nextword;
+      }
+      debugf("\n");
     } else {
       int rx;
-      fprintf(debug," || UNKNOWN");
+      debugf(" || UNKNOWN");
       for (rx=lx; rx<w && cols[rx]; rx++);
-      fprintf(debug," x=%d ctxmap=%x %d..%d\n",x, ctxmap, lx,rx);
+      debugf(" x=%d ctxmap=%x %d..%d\n",x, ctxmap, lx,rx);
       debug_flush();
       callout_unknown(rd, w,cols, lx,rx-1, ctxmap);
       goto restart;
     }
   }
   add_result(rd, 0,-1,-1,0);
-  fprintf(debug,"OCR  finished %d glyphs\n",rd->nresults);
+  debugf("OCR  finished %d glyphs\n",rd->nresults);
   debug_flush();
   return rd->results;
 }
@@ -371,7 +435,7 @@ OcrResultGlyph *ocr(OcrReader *rd, int w, Pixcol cols[]) {
 OcrReader *ocr_init(int h) {
   OcrReader *rd;
 
-  rd= malloc(sizeof(*rd));  eassert(rd);
+  rd= mmalloc(sizeof(*rd));
   memset(rd,0,sizeof(*rd));
   rd->h= h;
   readdb(rd);