OCR rejection infrastructure

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

diff --git a/pctb/ocr.c b/pctb/ocr.c
index 47597828c5f0091f1b5dce2e1592bb2e82aa4da6..4b9fa505b156afa3a001f3635ad0efeca5de2c84 100644 (file)
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -83,6 +83,8 @@ struct OcrCellTypeInfo {
 struct OcrReader {
   int h;
   DatabaseNode contexts[NCONTEXTS];
+  char *result;
+  int lresult, aresult;
   OcrResultGlyph *results;
   int aresults, nresults;
 };
@@ -224,6 +226,84 @@ static void readdb1(OcrReader *rd, const char *which, int local) {
   free(dbfname);
 }
 
+typedef struct Rejection Rejection;
+struct Rejection {
+  struct Rejection *next;
+  const char *fname;
+  int lno;
+  pcre *re;
+};
+
+Rejection *rejections;
+
+static void load_rejections(const char *which) {
+  char lbuf[1000];
+  char *fname= masprintf("%s/#%s-reject#.txt", get_vardir(), which);
+  int c, lno=0;
+  Rejection *rej;
+  
+  if (!dbfile_open(fname)) { free(fname); return; }
+  
+  while ((c= fgetc(dbfile))!=EOF) {
+    ungetc(c,dbfile);
+    lno++;
+    dbfile_getsline(lbuf,sizeof(lbuf),fname,lno);
+
+    if (!lbuf[0] || isspace(lbuf[0] || lbuf[0]=='#'))
+      continue;
+
+    rej= mmalloc(sizeof(*rej));
+    rej->next= rejections;
+    rej->fname= fname;
+    rej->lno= lno;
+
+    const char *err;
+    int erroffset;
+    rej->re= pcre_compile(lbuf, PCRE_NO_AUTO_CAPTURE|PCRE_UTF8,
+                         &err, &erroffset, 0);
+    if (!rej->re) {
+      char *what= masprintf("invalid regexp at offset %d: %s\n",
+                           erroffset, err);
+      dbfile_assertfail(fname, lno, what);
+    }
+    debugf("OCR LOADED REJECTION %s:%d `%s' %p\n", fname,lno, lbuf, rej->re);
+
+    rejections= rej;
+  }
+  sysassert(feof(dbfile));
+  dbfile_close();
+}
+  
+static int should_reject(OcrReader *rd) {
+  static int rejections_loaded;
+  int ovector[30];
+
+  if (!rejections_loaded) {
+    fetch_with_rsync("reject");
+    load_rejections("master");
+    load_rejections("local");
+    rejections_loaded=1;
+  }
+
+  debugf("[OCR REJECTION `%s'%d", rd->result, rd->lresult);
+  Rejection *rej;
+  for (rej=rejections; rej; rej=rej->next) {
+    debugf(" (%p)", rej);
+
+    int res= pcre_exec(rej->re, 0, rd->result,rd->lresult, 0,
+                      0, ovector, ARRAYSIZE(ovector));
+    if (res==PCRE_ERROR_NOMATCH) continue;
+    sysassert(res>0);
+
+    debugf(" MATCH]\n");
+    fprintf(stderr,"Rejecting OCR result `%s' (due to %s:%d)\n",
+           rd->result, rej->fname,rej->lno);
+    return 1;
+  }
+  debugf(" OK]");
+  return 0;
+}
+
 static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) {
   fprintf(resolver,"{");
   const char *spc="";
@@ -243,7 +323,10 @@ static void callout_unknown(OcrReader *rd, int w, const Pixcol cols[],
   const char *p;
 
   FILE *resolver= resolve_start();
-  if (!resolver || !(o_flags & ff_editcharset))
+  if (!resolver ||
+      !((o_flags & ff_charset_edit) ||
+       ((o_flags & ff_charset_allowedit) &&
+        (o_flags & ff_charset_havelocal))))
     fatal("OCR failed - unrecognised characters or ligatures.\n"
          "Character set database needs to be updated or augmented.\n"
          "See README.charset.\n");
@@ -291,6 +374,18 @@ static void add_result(OcrReader *rd, const char *s, int l, int r,
   rd->results[rd->nresults].match= match;
   rd->results[rd->nresults].ctxmap= ctxmap;
   rd->nresults++;
+
+  if (!s) return; /* just the sentinel for the caller */
+
+  int sl= strlen(s);
+  int newlresult= rd->lresult + sl;
+  if (newlresult >= rd->aresult) {
+    rd->aresult= (newlresult << 1) + 1;
+    rd->result= mrealloc(rd->result, rd->aresult);
+  }
+  memcpy(rd->result + rd->lresult, s, sl);
+  rd->lresult= newlresult;
+  rd->result[rd->lresult]= 0;
 }
 
 
@@ -427,6 +522,8 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w,
   nspaces=- w;
   fca.ctxmap= ct->initial;
   rd->nresults=0;
+  rd->lresult=0;
+  rd->result[0]=0;
   debugf("OCR h=%d w=%d",rd->h,w);
   for (x=0; x<w; x++) debugf(" "PIXCOL_PRFMT, PIXCOL_PRVAL(cols[x]));
   debugf("\n");
@@ -463,6 +560,10 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w,
     if (match) {
       debugf(" || YES");
       add_result(rd, match->str, x, match_rx, match_ctxi, fca.ctxmap);
+      if (should_reject(rd)) {
+       callout_unknown(rd, w,cols, match_rx+1,match_rx, 0);
+       goto restart;
+      }
       x= match_rx+1;
       if (match->match) fca.ctxmap= ct->midword;
       else debugf(" (empty)");
@@ -503,7 +604,12 @@ OcrReader *ocr_init(int h) {
 
   rd= mmalloc(sizeof(*rd));
   memset(rd,0,sizeof(*rd));
+
   rd->h= h;
+
+  rd->aresult= 10;
+  rd->result= mmalloc(rd->aresult);
+
   readdb(rd);
   return rd;
 }