X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?p=ypp-sc-tools.db-test.git;a=blobdiff_plain;f=pctb%2Focr.c;h=b7934c983e2f407a362183c0e5aad6756653165c;hp=6a91935769ecfc7da0e70681ea80db627c49327e;hb=1d0e4ca6694c19fadd97452eed1451248b156997;hpb=a5c969e80989ea2c678a99b46e7db08ae553e59d

diff --git a/pctb/ocr.c b/pctb/ocr.c
index 6a91935..b7934c9 100644
--- 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;
 
@@ -75,103 +74,79 @@ struct OcrReader {
   int aresults, nresults;
 };
 
-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);
-  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 (dbfile_getsline(lbuf,sizeof(lbuf),__FILE__,__LINE__))
 
 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 r,j,ctxi;
+  int j,ctxi;
   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);
-  sysassert(dbfname);
+  char *dbfname= masprintf("%s/#%s-char%d#.txt",
+			   get_vardir(), which, rd->h);
   
-  db= fopen(dbfname,"r");
-  free(dbfname);
-  if (!db) {
-    sysassert(errno==ENOENT);
-    return;
-  }
+  if (!dbfile_open(dbfname))
+    goto x;
 
-  FGETSLINE(db,lbuf);
+  FGETSLINE;
   dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1"));
 
-  r= fscanf(db, "%d", &h);
-  dbassert(r==1);
+  dbassert( dbfile_scanf("%d", &h) == 1);
   dbassert(h==rd->h);
 
   for (;;) {
-    FGETSLINE(db,lbuf);
-    if (!lbuf || lbuf[0]=='#') continue;
+    FGETSLINE;
+    if (!lbuf[0] || lbuf[0]=='#') continue;
     if (!strcmp(lbuf,".")) break;
 
     for (ctxi=0; ctxi<NCONTEXTS; ctxi++)
       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:
     for (nchrs=0;;) {
-      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); sysassert(!ferror(db)); dbassert(c=='x');
-	r= fscanf(db, "%2x", &cr); sysassert(!ferror(db)); dbassert(r==1);
-	assert(cr>0 && cr<=255);
-	c= cr;
-      }
+      int c= fgetc(dbfile); sysassert(!ferror(dbfile)); dbassert(c!=EOF);
+      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 (;;) {
-      FGETSLINE(db,lbuf);
+      FGETSLINE;
       if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; }
       char *ep;
       cv= strtoul(lbuf,&ep,16);  dbassert(!*ep);
@@ -184,7 +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) {
@@ -201,15 +179,27 @@ 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;
+      }
+    }
   }
-  sysassert(!ferror(db));
-  sysassert(!fclose(db));
+ x:
+  dbfile_close();
+  free(dbfname);
 }
 
-static void cu_pr_ctxmap(unsigned ctxmap) {
+static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) {
   fprintf(resolver,"{");
   const char *spc="";
   int ctxi;
@@ -223,41 +213,25 @@ static void cu_pr_ctxmap(unsigned ctxmap) {
 
 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,i, x,y;
+  int c,i, x,y;
   const OcrResultGlyph *s;
   const char *p;
-  char cb;
   Pixcol pv;
+
+  FILE *resolver= resolve_start();
+  if (!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) );
-    sysassert(! pipe(donepipe) );
-    resolver_pid= fork();
-    sysassert(resolver_pid!=-1);
-    if (!resolver_pid) {
-      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: */
-      sysassert( dup2(donepipe[1],4) ==4 );
-      execlp("./yppsc-ocr-resolver", "yppsc-ocr-resolver",
-	     DEBUGP(callout) ? "--debug" : "--noop-arg",
-	     "--automatic-1",
-	     (char*)0);
-      sysassert(!"execlp ocr-resolver failed");
-    }
-    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);
-  cu_pr_ctxmap(unk_ctxmap);
+  fprintf(resolver,
+	  "char\n"
+	  "%d %d ",unk_l,unk_r);
+  cu_pr_ctxmap(resolver,unk_ctxmap);
   for (i=0, s=rd->results; i<rd->nresults; i++, s++) {
     if (!strcmp(s->s," ")) continue;
     fprintf(resolver," %d %d ",s->l,s->r);
-    cu_pr_ctxmap(s->ctxmap);
+    cu_pr_ctxmap(resolver,s->ctxmap);
     fprintf(resolver," ");
     for (p=s->s; (c= *p); p++) {
       if (c=='\\') fprintf(resolver,"\\%c",c);
@@ -282,28 +256,8 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[],
     fputs("\",\n",resolver);
   }
   fputs("};\n",resolver);
-  sysassert(!ferror(resolver));
-  sysassert(!fflush(resolver));
-
-  sysassert(resolver);
-
-  int r;
-  for (;;) {
-    r= read(resolver_done,&cb,1);
-    if (r==-1) { sysassert(errno==EINTR); continue; }
-    break;
-  }
-
-  if (r==0) {
-    waitpid_check_exitstatus(resolver_pid, "character resolver");
-    fclose(resolver);
-    close(resolver_done);
-    resolver= 0;
-  } else {
-    assert(r==1);
-    sysassert(cb==0);
-  }
 
+  resolve_finish();
   readdb(rd);
 }
 
@@ -321,17 +275,89 @@ static void add_result(OcrReader *rd, const char *s, int l, int r,
 }
 
 
+typedef struct {
+  OcrReader *rd;
+  int w;
+  Pixcol *cols;
+  int x;
+  unsigned ctxmap;
+} FindCharArgs;
+
+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_unique(const FindCharArgs *fca, int *match_rx) {
+  DatabaseNode *uniquematch= 0;
+  int ctxi;
+    
+  debugf("OCR  lx=%d ct_state=%x  ", fca->x, fca->ctxmap);
+  for (ctxi=0; ctxi<NCONTEXTS; ctxi++) {
+    if (!(fca->ctxmap & (1u << ctxi))) continue;
+    debugf(" || %s",context_names[ctxi]);
+
+    DatabaseNode *match=
+      findchar_1ctx(fca, &fca->rd->contexts[ctxi], match_rx);
+    if (!match) continue;
+
+    if (uniquematch && strcmp(match->str, uniquematch->str)) {
+      debugf( " ambiguous");
+      uniquematch= 0;
+    } else {
+      uniquematch= match;
+    }
+  }
+
+  return uniquematch;
+}
+
 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]);
@@ -350,7 +376,7 @@ 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;
     }
@@ -360,81 +386,33 @@ 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_unique(&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);
@@ -445,6 +423,13 @@ 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;