From 21d9063f3d9f1283ec51ca5e588cfc2fa403f4a5 Mon Sep 17 00:00:00 2001 From: Stefan Ziegler <33447587+sz5000@users.noreply.github.com> Date: Tue, 21 Apr 2026 16:45:11 +0200 Subject: [PATCH] TTFSubsetter extension MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit two new public method added, addCustomCmapEntry and addCustomCmap TTFSubsetter currently only writes a single Windows Unicode BMP cmap subtable (platform 3, encoding 1) and only when addAll() has been called. There is no way for callers to inject additional cmap subtables — for example a Mac Roman subtable (platform 1, encoding 0) or a Windows Symbol subtable (platform 3, encoding 0). This limitation makes it impossible to correctly re-subset TrueType fonts that were originally subsetted by Ghostscript using its TT_BIAS=0xF000 strategy, where the font's Mac Roman cmap is the primary rendering cmap used by viewers. --- .../org/apache/fontbox/ttf/TTFSubsetter.java | 317 ++++++++++++++---- 1 file changed, 248 insertions(+), 69 deletions(-) diff --git a/fontbox/src/main/java/org/apache/fontbox/ttf/TTFSubsetter.java b/fontbox/src/main/java/org/apache/fontbox/ttf/TTFSubsetter.java index 1f633bf8d85..66cc39f4ff4 100755 --- a/fontbox/src/main/java/org/apache/fontbox/ttf/TTFSubsetter.java +++ b/fontbox/src/main/java/org/apache/fontbox/ttf/TTFSubsetter.java @@ -67,6 +67,13 @@ public final class TTFSubsetter private String prefix; private boolean hasAddedCompoundReferences; + /** + * Custom cmap entries added via {@link #addCustomCmapEntry(int, int, int, int)}. + * Each entry is {@code [platformId, platformEncodingId, charCode, gid]}. + * During {@link #buildCmapTable()} these are translated to new GIDs. + */ + private final List customCmapEntries = new java.util.ArrayList<>(); + /** * Creates a subsetter for the given font. * @@ -746,67 +753,191 @@ private int getNewGlyphId(Integer oldGid) private byte[] buildCmapTable() throws IOException { - if (ttf.getCmap() == null || uniToGID.isEmpty() - || keepTables != null && !keepTables.contains(CmapTable.TAG)) + boolean hasUnicode = ttf.getCmap() != null && !uniToGID.isEmpty(); + boolean hasCustom = !customCmapEntries.isEmpty(); + + if ((!hasUnicode && !hasCustom) + || (keepTables != null && !keepTables.contains(CmapTable.TAG))) { return null; } - ByteArrayOutputStream bos = new ByteArrayOutputStream(64); + // Group custom entries by (platformId, platformEncodingId). + // Each group becomes its own Format 4 subtable. + // Use a LinkedHashMap so insertion order is preserved (stable output). + Map> customSubtables = new LinkedHashMap<>(); + for (int[] entry : customCmapEntries) + { + int platformId = entry[0]; + int platformEncodingId = entry[1]; + int charCode = entry[2]; + int gid = entry[3]; + + // Only include entries whose glyph is actually in the subset. + if (!glyphIds.contains(gid)) + { + continue; + } + + long key = ((long) platformId << 16) | (platformEncodingId & 0xFFFF); + customSubtables + .computeIfAbsent(key, k -> new TreeMap<>()) + .put(charCode, gid); + } + + // Total number of subtables: 1 for the standard Windows Unicode BMP (if present) + // plus one per distinct custom (platform, encoding) pair. + int numSubtables = (hasUnicode ? 1 : 0) + customSubtables.size(); + + ByteArrayOutputStream bos = new ByteArrayOutputStream(256); DataOutputStream out = new DataOutputStream(bos); // cmap header writeUint16(out, 0); // version - writeUint16(out, 1); // numberSubtables + writeUint16(out, numSubtables); + + // Each encoding record is 8 bytes: platformID(2) + encodingID(2) + offset(4). + // Offsets are relative to the start of the cmap table. + // Header = 4 bytes, encoding records = 8 * numSubtables bytes. + int headerSize = 4 + 8 * numSubtables; + + // Build each subtable as raw bytes so we can compute offsets up front. + List subtableBytes = new java.util.ArrayList<>(); + + if (hasUnicode) + { + subtableBytes.add(buildFormat4Subtable(uniToGID)); + } + for (SortedMap codeToGid : customSubtables.values()) + { + // Translate gid → newGid for each entry. + SortedMap codeToNewGid = new TreeMap<>(); + codeToGid.forEach((code, gid) -> + { + int newGid = getNewGlyphId(gid); + if (newGid > 0) + { + codeToNewGid.put(code, newGid); + } + }); + subtableBytes.add(buildFormat4SubtableNewGids(codeToNewGid)); + } + + // Write encoding records (offsets relative to start of cmap table). + int subtableOffset = headerSize; + + if (hasUnicode) + { + writeUint16(out, CmapTable.PLATFORM_WINDOWS); + writeUint16(out, CmapTable.ENCODING_WIN_UNICODE_BMP); + writeUint32(out, subtableOffset); + subtableOffset += subtableBytes.get(0).length; + } + + int subtableIndex = hasUnicode ? 1 : 0; + for (long key : customSubtables.keySet()) + { + int platformId = (int) (key >> 16); + int platformEncodingId = (int) (key & 0xFFFF); + writeUint16(out, platformId); + writeUint16(out, platformEncodingId); + writeUint32(out, subtableOffset); + subtableOffset += subtableBytes.get(subtableIndex++).length; + } + + // Write subtable data. + for (byte[] subtable : subtableBytes) + { + out.write(subtable); + } + + return bos.toByteArray(); + } + + /** + * Builds a Format 4 cmap subtable from a {@code unicode → gid} map (the existing + * path used for the standard Windows Unicode BMP subtable). + */ + private byte[] buildFormat4Subtable(SortedMap uniToGIDMap) + throws IOException + { + // Translate gid → newGid inline. + SortedMap codeToNewGid = new TreeMap<>(); + for (Map.Entry e : uniToGIDMap.entrySet()) + { + codeToNewGid.put(e.getKey(), getNewGlyphId(e.getValue())); + } + return buildFormat4SubtableNewGids(codeToNewGid); + } - // encoding record - writeUint16(out, CmapTable.PLATFORM_WINDOWS); // platformID - writeUint16(out, CmapTable.ENCODING_WIN_UNICODE_BMP); // platformSpecificID - writeUint32(out, 12); // offset 4 * 2 + 4 + /** + * Builds a Format 4 cmap subtable from a {@code charCode → newGid} map. + * The map must be sorted by charCode ascending. + */ + private byte[] buildFormat4SubtableNewGids(SortedMap codeToNewGid) + throws IOException + { + ByteArrayOutputStream bos = new ByteArrayOutputStream(64); + DataOutputStream out = new DataOutputStream(bos); + + if (codeToNewGid.isEmpty()) + { + // Write a minimal valid Format 4 subtable with only the sentinel segment. + int segCount = 1; + int searchRange = 2; + writeUint16(out, 4); // format + writeUint16(out, 8 * 2 + segCount * 4 * 2); // length + writeUint16(out, 0); // language + writeUint16(out, segCount * 2); // segCountX2 + writeUint16(out, searchRange); + writeUint16(out, 0); // entrySelector + writeUint16(out, 0); // rangeShift + writeUint16(out, 0xffff); // endCode[0] + writeUint16(out, 0); // reservedPad + writeUint16(out, 0xffff); // startCode[0] + writeUint16(out, 1); // idDelta[0] + writeUint16(out, 0); // idRangeOffset[0] + return bos.toByteArray(); + } - // build Format 4 subtable (Unicode BMP) - Iterator> it = uniToGID.entrySet().iterator(); + Iterator> it = codeToNewGid.entrySet().iterator(); Entry lastChar = it.next(); Entry prevChar = lastChar; - int lastGid = getNewGlyphId(lastChar.getValue()); + int lastGid = lastChar.getValue(); - // +1 because .notdef is missing in uniToGID - int[] startCode = new int[uniToGID.size()+1]; - int[] endCode = new int[startCode.length]; - int[] idDelta = new int[startCode.length]; + int[] startCode = new int[codeToNewGid.size() + 1]; + int[] endCode = new int[startCode.length]; + int[] idDelta = new int[startCode.length]; int segCount = 0; - while(it.hasNext()) + + while (it.hasNext()) { Entry curChar2Gid = it.next(); - int curGid = getNewGlyphId(curChar2Gid.getValue()); + int curGid = curChar2Gid.getValue(); - // todo: need format Format 12 for non-BMP if (curChar2Gid.getKey() > 0xFFFF) { - throw new UnsupportedOperationException("non-BMP Unicode character"); + throw new UnsupportedOperationException("non-BMP character code in custom cmap"); } - if (curChar2Gid.getKey() != prevChar.getKey()+1 || - curGid - lastGid != curChar2Gid.getKey() - lastChar.getKey()) + if (curChar2Gid.getKey() != prevChar.getKey() + 1 + || curGid - lastGid != curChar2Gid.getKey() - lastChar.getKey()) { if (lastGid != 0) { - // don't emit ranges, which map to GID 0, the - // undef glyph is emitted a the very last segment startCode[segCount] = lastChar.getKey(); - endCode[segCount] = prevChar.getKey(); - idDelta[segCount] = lastGid - lastChar.getKey(); + endCode[segCount] = prevChar.getKey(); + idDelta[segCount] = lastGid - lastChar.getKey(); segCount++; } else if (!lastChar.getKey().equals(prevChar.getKey())) { - // shorten ranges which start with GID 0 by one startCode[segCount] = lastChar.getKey() + 1; - endCode[segCount] = prevChar.getKey(); - idDelta[segCount] = lastGid - lastChar.getKey(); + endCode[segCount] = prevChar.getKey(); + idDelta[segCount] = lastGid - lastChar.getKey(); segCount++; } - lastGid = curGid; + lastGid = curGid; lastChar = curChar2Gid; } prevChar = curChar2Gid; @@ -814,51 +945,30 @@ else if (!lastChar.getKey().equals(prevChar.getKey())) // trailing segment startCode[segCount] = lastChar.getKey(); - endCode[segCount] = prevChar.getKey(); - idDelta[segCount] = lastGid -lastChar.getKey(); + endCode[segCount] = prevChar.getKey(); + idDelta[segCount] = lastGid - lastChar.getKey(); segCount++; - // GID 0 + // sentinel segment (GID 0 / 0xFFFF) startCode[segCount] = 0xffff; - endCode[segCount] = 0xffff; - idDelta[segCount] = 1; + endCode[segCount] = 0xffff; + idDelta[segCount] = 1; segCount++; - // write format 4 subtable - int searchRange = 2 * (int)Math.pow(2, log2(segCount)); - writeUint16(out, 4); // format - writeUint16(out, 8 * 2 + segCount * 4*2); // length - writeUint16(out, 0); // language - writeUint16(out, segCount * 2); // segCountX2 - writeUint16(out, searchRange); // searchRange - writeUint16(out, log2(searchRange / 2)); // entrySelector - writeUint16(out, 2 * segCount - searchRange); // rangeShift - - // endCode[segCount] - for (int i = 0; i < segCount; i++) - { - writeUint16(out, endCode[i]); - } - - // reservedPad - writeUint16(out, 0); - - // startCode[segCount] - for (int i = 0; i < segCount; i++) - { - writeUint16(out, startCode[i]); - } - - // idDelta[segCount] - for (int i = 0; i < segCount; i++) - { - writeUint16(out, idDelta[i]); - } - - for (int i = 0; i < segCount; i++) - { - writeUint16(out, 0); - } + int searchRange = 2 * (int) Math.pow(2, log2(segCount)); + writeUint16(out, 4); // format + writeUint16(out, 8 * 2 + segCount * 4 * 2); // length + writeUint16(out, 0); // language + writeUint16(out, segCount * 2); // segCountX2 + writeUint16(out, searchRange); + writeUint16(out, log2(searchRange / 2)); // entrySelector + writeUint16(out, 2 * segCount - searchRange); // rangeShift + + for (int i = 0; i < segCount; i++) writeUint16(out, endCode[i]); + writeUint16(out, 0); // reservedPad + for (int i = 0; i < segCount; i++) writeUint16(out, startCode[i]); + for (int i = 0; i < segCount; i++) writeUint16(out, idDelta[i]); + for (int i = 0; i < segCount; i++) writeUint16(out, 0); // idRangeOffset return bos.toByteArray(); } @@ -1157,9 +1267,78 @@ private int log2(int num) return (int) Math.floor(Math.log(num) / Math.log(2)); } + /** + * Adds the given glyph IDs directly to the subset, bypassing the Unicode cmap lookup. + * Use this for glyphs that have no standard Unicode mapping, such as glyphs in + * symbol or PUA-encoded fonts. The GIDs are renumbered during subsetting; use + * {@link #getGIDMap()} afterwards to obtain the old-to-new mapping. + * + * @param allGlyphIds the set of glyph IDs to include in the subset + */ public void addGlyphIds(Set allGlyphIds) { glyphIds.addAll(allGlyphIds); } + /** + * Adds a single entry to a custom cmap subtable that will be written into the subset TTF. + * Use this when the font uses a non-Unicode cmap encoding that must be preserved in the + * subset — for example a Mac Roman subtable (platform 1, encoding 0) or a Windows Symbol + * subtable (platform 3, encoding 0) where character codes do not correspond to standard + * Unicode codepoints. + * + *

The {@code gid} is automatically remapped to the renumbered GID that the subsetter + * assigns during {@link #writeToStream(OutputStream)}. If the glyph is not included in the + * subset (via {@link #addGlyphIds(Set)} or {@link #add(int)}), the entry is silently + * ignored. + * + *

Multiple calls with the same {@code platformId} and {@code platformEncodingId} are + * accumulated into a single Format 4 subtable. Multiple calls with different platform or + * encoding values produce separate subtables. + * + *

Example — preserve a Mac Roman cmap (platform 1, encoding 0): + * 

+     * CmapSubtable macCmap = ttf.getCmap()
+     *         .getSubtable(CmapTable.PLATFORM_MACINTOSH, CmapTable.ENCODING_MAC_ROMAN);
+     * for (int code : usedCodes) {
+     *     int gid = macCmap.getGlyphId(code);
+     *     if (gid > 0) {
+     *         subsetter.addGlyphIds(Collections.singleton(gid));
+     *         subsetter.addCustomCmapEntry(CmapTable.PLATFORM_MACINTOSH,
+     *                                      CmapTable.ENCODING_MAC_ROMAN, code, gid);
+     *     }
+     * }
+     *
+ * + * @param platformId the platform identifier (e.g. {@link CmapTable#PLATFORM_MACINTOSH} + * or {@link CmapTable#PLATFORM_WINDOWS}) + * @param platformEncodingId the platform-specific encoding identifier + * (e.g. {@link CmapTable#ENCODING_MAC_ROMAN} or + * {@link CmapTable#ENCODING_WIN_SYMBOL}) + * @param charCode the character code to map from, as it appears in the font's cmap + * @param gid the glyph ID in the original (pre-subset) font + */ + public void addCustomCmapEntry(int platformId, int platformEncodingId, + int charCode, int gid) + { + customCmapEntries.add(new int[] { platformId, platformEncodingId, charCode, gid }); + } + + /** + * Adds all entries from a character-code-to-GID map as custom cmap entries for the given + * platform and encoding. This is a convenience wrapper around repeated calls to + * {@link #addCustomCmapEntry(int, int, int, int)}; see that method for full details. + * + * @param platformId the platform identifier (e.g. {@link CmapTable#PLATFORM_MACINTOSH}) + * @param platformEncodingId the platform-specific encoding identifier + * (e.g. {@link CmapTable#ENCODING_MAC_ROMAN}) + * @param codeToGid map of character code to glyph ID in the original font + */ + public void addCustomCmap(int platformId, int platformEncodingId, + Map codeToGid) + { + codeToGid.forEach((code, gid) -> + addCustomCmapEntry(platformId, platformEncodingId, code, gid)); + } + }