Fix 695501: Intelligent cmap remapping.

Use the actual ranges from the cpt-to-gid cmap to optimize the
remapping of ToUnicode cmaps from cpt-to-unicode into gid-to-unicode
format.

1 files changed, 41 insertions, 55 deletions

diff --git a/source/pdf/pdf-unicode.c b/source/pdf/pdf-unicode.c
index ffe5f6b7..6b893051 100644
--- a/source/pdf/pdf-unicode.c
+++ b/source/pdf/pdf-unicode.c
@@ -2,83 +2,69 @@
 
 /* Load or synthesize ToUnicode map for fonts */
 
-static void find_min_max_cpt(pdf_cmap *cmap, unsigned int *minp, unsigned int *maxp)
+static void
+pdf_remap_cmap_range(fz_context *ctx, pdf_cmap *ucs_from_gid,
+	unsigned int cpt, unsigned int gid, unsigned int n, pdf_cmap *ucs_from_cpt)
 {
-	unsigned int min = UINT_MAX;
-	unsigned int max = 0;
-	int i;
+	unsigned int k;
+	int ucsbuf[8];
+	int ucslen;
 
-	for (i = 0; i < cmap->rlen; ++i)
+	for (k = 0; k <= n; ++k)
 	{
-		if (cmap->ranges[i].low < min)
-			min = cmap->ranges[i].low;
-		if (cmap->ranges[i].high > max)
-			max = cmap->ranges[i].high;
+		ucslen = pdf_lookup_cmap_full(ucs_from_cpt, cpt + k, ucsbuf);
+		if (ucslen == 1)
+			pdf_map_range_to_range(ctx, ucs_from_gid, gid + k, gid + k, ucsbuf[0]);
+		else if (ucslen > 1)
+			pdf_map_one_to_many(ctx, ucs_from_gid, gid + k, ucsbuf, ucslen);
 	}
+}
+
+static pdf_cmap *
+pdf_remap_cmap(fz_context *ctx, pdf_cmap *gid_from_cpt, pdf_cmap *ucs_from_cpt)
+{
+	pdf_cmap *ucs_from_gid;
+	unsigned int i, a, b, x;
 
-	for (i = 0; i < cmap->xlen; ++i)
+	ucs_from_gid = pdf_new_cmap(ctx);
+
+	if (gid_from_cpt->usecmap)
+		ucs_from_gid->usecmap = pdf_remap_cmap(ctx, gid_from_cpt->usecmap, ucs_from_cpt);
+
+	for (i = 0; i < gid_from_cpt->rlen; ++i)
 	{
-		if (cmap->xranges[i].low < min)
-			min = cmap->xranges[i].low;
-		if (cmap->xranges[i].high > max)
-			max = cmap->xranges[i].high;
+		a = gid_from_cpt->ranges[i].low;
+		b = gid_from_cpt->ranges[i].high;
+		x = gid_from_cpt->ranges[i].out;
+		pdf_remap_cmap_range(ctx, ucs_from_gid, a, x, b - a, ucs_from_cpt);
 	}
 
-	for (i = 0; i < cmap->mlen; ++i)
+	for (i = 0; i < gid_from_cpt->xlen; ++i)
 	{
-		if (cmap->mranges[i].low < min)
-			min = cmap->mranges[i].low;
-		if (cmap->mranges[i].low > max)
-			max = cmap->mranges[i].low;
+		a = gid_from_cpt->xranges[i].low;
+		b = gid_from_cpt->xranges[i].high;
+		x = gid_from_cpt->xranges[i].out;
+		pdf_remap_cmap_range(ctx, ucs_from_gid, a, x, b - a, ucs_from_cpt);
 	}
 
-	*minp = min;
-	*maxp = max;
+	/* Font encoding CMaps don't have one-to-many mappings, so we can ignore the mranges. */
+
+	pdf_sort_cmap(ctx, ucs_from_gid);
+
+	return ucs_from_gid;
 }
 
 void
 pdf_load_to_unicode(pdf_document *doc, pdf_font_desc *font,
 	char **strings, char *collection, pdf_obj *cmapstm)
 {
-	unsigned int cpt, min, max;
-	int gid;
-	int ucsbuf[8];
-	int ucslen;
-	int i;
 	fz_context *ctx = doc->ctx;
+	unsigned int cpt;
 
 	if (pdf_is_stream(doc, pdf_to_num(cmapstm), pdf_to_gen(cmapstm)))
 	{
-		pdf_cmap *gid_from_cpt = font->encoding;
 		pdf_cmap *ucs_from_cpt = pdf_load_embedded_cmap(doc, cmapstm);
-
-		font->to_unicode = pdf_new_cmap(ctx);
-
-		/* in case the code space range is much larger than the actual number of characters */
-		find_min_max_cpt(gid_from_cpt, &min, &max);
-
-		for (i = 0; i < gid_from_cpt->codespace_len; ++i)
-		{
-			unsigned int l = gid_from_cpt->codespace[i].low;
-			unsigned int h = gid_from_cpt->codespace[i].high;
-			l = l < min ? min : l > max ? max : l;
-			h = h < min ? min : h > max ? max : h;
-			for (cpt = l; cpt <= h; ++cpt)
-			{
-				gid = pdf_lookup_cmap(gid_from_cpt, cpt);
-				if (gid >= 0)
-				{
-					ucslen = pdf_lookup_cmap_full(ucs_from_cpt, cpt, ucsbuf);
-					if (ucslen == 1)
-						pdf_map_range_to_range(ctx, font->to_unicode, gid, gid, ucsbuf[0]);
-					if (ucslen > 1)
-						pdf_map_one_to_many(ctx, font->to_unicode, gid, ucsbuf, ucslen);
-				}
-			}
-		}
-
-		pdf_sort_cmap(ctx, font->to_unicode);
-
+		font->to_unicode = pdf_remap_cmap(ctx, font->encoding, ucs_from_cpt);
 		pdf_drop_cmap(ctx, ucs_from_cpt);
 		font->size += pdf_cmap_size(ctx, font->to_unicode);
 	}