X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~yarrgweb/git?a=blobdiff_plain;f=pctb%2Focr.c;h=47597828c5f0091f1b5dce2e1592bb2e82aa4da6;hb=a16191c13cdc1eab1e43cc9662c2271481b3a9b8;hp=f174c68117c8d31e29da62fe94f153882eb8d14f;hpb=c2c039f07de178037bfbf12e4ee23c06f546de32;p=ypp-sc-tools.web-live.git diff --git a/pctb/ocr.c b/pctb/ocr.c index f174c68..4759782 100644 --- a/pctb/ocr.c +++ b/pctb/ocr.c @@ -44,7 +44,7 @@ typedef struct { OcrReader *rd; OcrCellType ct; int w; - Pixcol *cols; + const Pixcol *cols; int x; unsigned ctxmap; } FindCharArgs; @@ -55,8 +55,8 @@ typedef struct { #define FOR_EACH_CONTEXT(EACH) \ EACH(Word) \ - EACH(Upper) \ EACH(Lower) \ + EACH(Upper) \ EACH(Digit) #define FEC_ENUM(Context) ct_##Context, @@ -126,8 +126,11 @@ static void readdb1(OcrReader *rd, const char *which, int local) { if (!dbfile_open(dbfname)) goto x; + if (local) + o_flags |= ff_charset_havelocal; + FGETSLINE; - dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v1")); + dbassert(!strcmp(lbuf,"# ypp-sc-tools pctb font v2")); dbassert( dbfile_scanf("%d", &h) == 1); dbassert(h==rd->h); @@ -161,12 +164,20 @@ static void readdb1(OcrReader *rd, const char *which, int local) { for (;;) { FGETSLINE; if (!lbuf[0]) { dbassert(current != &rd->contexts[ctxi]); break; } - char *ep; - cv= strtoul(lbuf,&ep,16); dbassert(!*ep); - dbassert(!(cv & ~((1UL << rd->h)-1))); + dbassert( strlen(lbuf) == rd->h ); + FILLZERO(cv); + int y; + for (y=0; ynlinks; j++) - if (current->links[j].col == cv) { + if (!pixcol_cmp(&cv, ¤t->links[j].col)) { current= current->links[j].then; goto found_link; } @@ -225,12 +236,11 @@ static void cu_pr_ctxmap(FILE *resolver, unsigned ctxmap) { fprintf(resolver,"}"); } -static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], +static void callout_unknown(OcrReader *rd, int w, const Pixcol cols[], int unk_l, int unk_r, unsigned unk_ctxmap) { int c,i, x,y; const OcrResultGlyph *s; const char *p; - Pixcol pv; FILE *resolver= resolve_start(); if (!resolver || !(o_flags & ff_editcharset)) @@ -258,20 +268,12 @@ static void callout_unknown(OcrReader *rd, int w, Pixcol cols[], fputc('\n',resolver); fprintf(resolver, - "/* XPM */\n" - "static char *t[] = {\n" - "/* columns rows colors chars-per-pixel */\n" - "\"%d %d 2 1\",\n" - "\" c black\",\n" - "\"o c white\",\n", - w,rd->h); - for (y=0, pv=1; yh; y++, pv<<=1) { - fputc('"',resolver); + "P2\n%d %d %d\n", w, rd->h, AAMAXVAL); + for (y=0; yh; y++) { for (x=0; x fca->w) break; Pixcol cv= fca->cols[x]; - debugf(" cv=%"PSPIXCOL(PRIx),cv); + debugf(" cv="PIXCOL_PRFMT, PIXCOL_PRVAL(cv)); for (i=0; inlinks; i++) - if (current->links[i].col == cv) + if (!pixcol_cmp(&cv, ¤t->links[i].col)) goto found; /* not found */ debugf(" ?"); @@ -363,12 +365,30 @@ static DatabaseNode *findchar(const FindCharArgs *fca, static int findchar_select_text(const FindCharArgs *fca, const FindCharResults results[]) { - if (fca->ctxmap != 017) return -1; + + dbassert(! results[ct_Digit].match ); /* digits are supposedly unambiguous */ + + switch (fca->ctxmap) { - dbassert(! results[ct_Digit].match ); - if (results[ct_Word].match) return ct_Word; - if (results[ct_Lower].rx > results[ct_Upper].rx) return ct_Lower; - if (results[ct_Upper].rx > results[ct_Lower].rx) return ct_Upper; +#define RETURN_IF_LONGER(this,that) do{ \ + if (results[ct_##this].rx > results[ct_##that].rx) \ + return ct_##this; \ + }while(0) + + case ctf_Digit | ctf_Upper | ctf_Lower | ctf_Word: + /* Start of word. Prefer Word match; failing that, take the longest */ + if (results[ct_Word].match) return ct_Word; + RETURN_IF_LONGER(Lower,Upper); + RETURN_IF_LONGER(Upper,Lower); + break; + + case ctf_Digit | ctf_Upper | ctf_Lower: + /* Mid-word. Prefer longer match; failing that, match lower. */ + RETURN_IF_LONGER(Upper,Lower); + return ct_Lower; + } + + /* oh well */ return -1; } @@ -381,7 +401,7 @@ const struct OcrCellTypeInfo ocr_celltype_number= { const struct OcrCellTypeInfo ocr_celltype_text= { .initial= ctf_Digit | ctf_Upper, .nextword= ctf_Digit | ctf_Upper | ctf_Lower | ctf_Word, - .midword= ctf_Digit | ctf_Lower, + .midword= ctf_Digit | ctf_Upper | ctf_Lower, .space_spaces= 4, .name= "text", .findchar_select= findchar_select_text @@ -390,7 +410,8 @@ const struct OcrCellTypeInfo ocr_celltype_text= { const char *ocr_celltype_name(OcrCellType ct) { return ct->name; } -OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { +OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, + const Pixcol cols[]) { int nspaces; int x; @@ -407,7 +428,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { fca.ctxmap= ct->initial; rd->nresults=0; debugf("OCR h=%d w=%d",rd->h,w); - for (x=0; x=w) break; - if (!cols[x]) { + if (!pixcol_nonzero(&cols[x])) { nspaces++; x++; if (nspaces == ct->space_spaces) { @@ -454,7 +475,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) { } else { int rx; debugf(" || UNKNOWN"); - for (rx=x; rx