pdf: write only ToUnicode mappings needed by the font, trimming anything out of [firstChar, lastCha…

src/pdf/SkPDFFont.cpp

Index: src/pdf/SkPDFFont.cpp
===================================================================
--- src/pdf/SkPDFFont.cpp	(revision 11153)
+++ src/pdf/SkPDFFont.cpp	(working copy)
@@ -342,7 +342,9 @@
 
 }  // namespace
 
-static void append_tounicode_header(SkDynamicMemoryWStream* cmap) {
+static void append_tounicode_header(SkDynamicMemoryWStream* cmap,
+                                    uint16_t firstGlyphID,
+                                    uint16_t lastGlyphID) {
     // 12 dict begin: 12 is an Adobe-suggested value. Shall not change.
     // It's there to prevent old version Adobe Readers from malfunctioning.
     const char* kHeader =
@@ -365,17 +367,20 @@
 
     // The CMapName must be consistent to /CIDSystemInfo above.
     // /CMapType 2 means ToUnicode.
-    // We specify codespacerange from 0x0000 to 0xFFFF because we convert our
-    // code table from unsigned short (16-bits). Codespace range just tells the
-    // PDF processor the valid range. It does not matter whether a complete
-    // mapping is provided or not.
-    const char* kTypeInfo =
+    // Codespace range just tells the PDF processor the valid range.
+    const char* kTypeInfoHeader =
         "/CMapName /Adobe-Identity-UCS def\n"
         "/CMapType 2 def\n"
-        "1 begincodespacerange\n"
-        "<0000> <FFFF>\n"
-        "endcodespacerange\n";
-    cmap->writeText(kTypeInfo);
+        "1 begincodespacerange\n";
+    cmap->writeText(kTypeInfoHeader);
+    
+    // e.g.     "<0000> <FFFF>\n"
+    SkString range;
+    range.appendf("<%04X> <%04X>\n", firstGlyphID, lastGlyphID);
+    cmap->writeText(range.c_str());
+    
+    const char* kTypeInfoFooter = "endcodespacerange\n";
+    cmap->writeText(kTypeInfoFooter);
 }
 
 static void append_cmap_footer(SkDynamicMemoryWStream* cmap) {
@@ -469,11 +474,15 @@
 // ( see caller in tests/ToUnicode.cpp )
 void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
                           const SkPDFGlyphSet* subset,
-                          SkDynamicMemoryWStream* cmap);
+                          SkDynamicMemoryWStream* cmap,
+                          uint16_t firstGlyphID,
+                          uint16_t lastGlyphID);
 
 void append_cmap_sections(const SkTDArray<SkUnichar>& glyphToUnicode,
                           const SkPDFGlyphSet* subset,
-                          SkDynamicMemoryWStream* cmap) {
+                          SkDynamicMemoryWStream* cmap,
+                          uint16_t firstGlyphID,
+                          uint16_t lastGlyphID) {
     if (glyphToUnicode.isEmpty()) {
         return;
     }
@@ -485,7 +494,8 @@
     bool rangeEmpty = true;
     const int count = glyphToUnicode.count();
 
-    for (int i = 0; i < count + 1; ++i) {
+    for (int i = SkMax32(firstGlyphID - 1, 0);
+          i < SkMax32(count, lastGlyphID) + 1; ++i) {
         bool inSubset = i < count && (subset == NULL || subset->has(i));
         if (!rangeEmpty) {
             // PDF spec requires bfrange not changing the higher byte,
@@ -498,14 +508,23 @@
                 glyphToUnicode[i] == currentRangeEntry.fUnicode + i -
                                          currentRangeEntry.fStart;
             if (!inSubset || !inRange) {
-                if (currentRangeEntry.fEnd > currentRangeEntry.fStart) {
-                    bfrangeEntries.push(currentRangeEntry);
-                } else {
+                BFRange range = {(uint16_t)SkMax32(firstGlyphID,

[vandebo (ex-Chrome), 2013/09/09 16:04:50]

With the loop changed, I don't think you need any of the changes here.

[edisonn, 2013/09/09 16:48:41]

I need it, I start from firstGlypthID - 1

[vandebo (ex-Chrome), 2013/09/09 17:13:11]

I don't see why you start at first - 1, please explain.

[edisonn, 2013/09/09 17:32:11]

Correct, it is not needed.

+                                                   currentRangeEntry.fStart),
+                                 (uint16_t)SkMin32(lastGlyphID,
+                                                   currentRangeEntry.fEnd),
+                                 currentRangeEntry.fUnicode};
+                if (range.fEnd > range.fStart) {
+                    bfrangeEntries.push(range);
+                } else if (range.fStart == range.fEnd &&
+                           firstGlyphID <= range.fStart &&
+                           range.fEnd <= lastGlyphID) {
                     BFChar* entry = bfcharEntries.append();
-                    entry->fGlyphId = currentRangeEntry.fStart;
-                    entry->fUnicode = currentRangeEntry.fUnicode;
+                    entry->fGlyphId = range.fStart;
+                    entry->fUnicode = range.fUnicode;
                 }
                 rangeEmpty = true;
+            } else if (i > lastGlyphID && !inSubset && !inRange) {

[vandebo (ex-Chrome), 2013/09/09 16:04:50]

or here

[edisonn, 2013/09/09 16:48:41]

This one i need, we go to max(lastGlyphID, count)

[vandebo (ex-Chrome), 2013/09/09 17:13:11]

Isn't this break redundant with the loop condition?

[edisonn, 2013/09/09 17:32:11]

no, the loop goes to max(last, count), while this if breaks if we exceeded last,
and there is no rand not subset

Without the if, the loop with continue to cycle doing nothing until max(count,
last)

+                break;
             }
         }
         if (inSubset) {
@@ -526,10 +545,13 @@
 
 static SkPDFStream* generate_tounicode_cmap(
         const SkTDArray<SkUnichar>& glyphToUnicode,
-        const SkPDFGlyphSet* subset) {
+        const SkPDFGlyphSet* subset,
+        uint16_t firstGlyphID,
+        uint16_t lastGlyphID) {
     SkDynamicMemoryWStream cmap;
-    append_tounicode_header(&cmap);
-    append_cmap_sections(glyphToUnicode, subset, &cmap);
+    append_tounicode_header(&cmap, firstGlyphID, lastGlyphID);
+    append_cmap_sections(glyphToUnicode, subset, &cmap,
+                         firstGlyphID, lastGlyphID);
     append_cmap_footer(&cmap);
     SkAutoTUnref<SkMemoryStream> cmapStream(new SkMemoryStream());
     cmapStream->setData(cmap.copyToData())->unref();
@@ -1015,7 +1037,8 @@
         return;
     }
     SkAutoTUnref<SkPDFStream> pdfCmap(
-        generate_tounicode_cmap(fFontInfo->fGlyphToUnicode, subset));
+        generate_tounicode_cmap(fFontInfo->fGlyphToUnicode, subset,
+                                firstGlyphID(), lastGlyphID()));
     addResource(pdfCmap.get());
     insert("ToUnicode", new SkPDFObjRef(pdfCmap.get()))->unref();
 }