Merge pull request #378 from googlefonts/colrv1-handle-unknown-flags

Colrv1 handle unknown flags and don't include region indicators in noflags build

colrv1/noflags.toml

@@ -54,32 +54,6 @@ srcs = [
   "../svg/emoji_u1f198.svg", 
   "../svg/emoji_u1f199.svg", 
   "../svg/emoji_u1f19a.svg", 
-  "../svg/emoji_u1f1e6.svg", 
-  "../svg/emoji_u1f1e7.svg", 
-  "../svg/emoji_u1f1e8.svg", 
-  "../svg/emoji_u1f1e9.svg", 
-  "../svg/emoji_u1f1ea.svg", 
-  "../svg/emoji_u1f1eb.svg", 
-  "../svg/emoji_u1f1ec.svg", 
-  "../svg/emoji_u1f1ed.svg", 
-  "../svg/emoji_u1f1ee.svg", 
-  "../svg/emoji_u1f1ef.svg", 
-  "../svg/emoji_u1f1f0.svg", 
-  "../svg/emoji_u1f1f1.svg", 
-  "../svg/emoji_u1f1f2.svg", 
-  "../svg/emoji_u1f1f3.svg", 
-  "../svg/emoji_u1f1f4.svg", 
-  "../svg/emoji_u1f1f5.svg", 
-  "../svg/emoji_u1f1f6.svg", 
-  "../svg/emoji_u1f1f7.svg", 
-  "../svg/emoji_u1f1f8.svg", 
-  "../svg/emoji_u1f1f9.svg", 
-  "../svg/emoji_u1f1fa.svg", 
-  "../svg/emoji_u1f1fb.svg", 
-  "../svg/emoji_u1f1fc.svg", 
-  "../svg/emoji_u1f1fd.svg", 
-  "../svg/emoji_u1f1fe.svg", 
-  "../svg/emoji_u1f1ff.svg", 
   "../svg/emoji_u1f201.svg", 
   "../svg/emoji_u1f202.svg", 
   "../svg/emoji_u1f21a.svg", 

colrv1_postproc.py

@@ -5,11 +5,13 @@ For now substantially based on copying from a correct bitmap build.
 """
 from absl import app
 import functools
+from fontTools.feaLib.builder import addOpenTypeFeaturesFromString
 from fontTools import ttLib
 from fontTools.ttLib.tables import _g_l_y_f as glyf
 from fontTools.ttLib.tables import otTables as ot
 import map_pua_emoji
 from nototools import add_vs_cmap
+from nototools import font_data
 from nototools import unicode_data
 from pathlib import Path
 
@@ -95,12 +97,13 @@ def _add_cmap_entries(colr_font, codepoint, glyph_name):
         print(f"Map 0x{codepoint:04x} to {glyph_name}, format {table.format}")
 
 
-def _map_missing_flag_tag_chars_to_empty_glyphs(colr_font):
+FLAG_TAGS = set(range(0xE0030, 0xE0039 + 1)) | set(range(0xE0061, 0xE007A + 1))
-    # Add all tag characters used in flags
+CANCEL_TAG = 0xE007F
-    tag_cps = set(range(0xE0030, 0xE0039 + 1)) | set(range(0xE0061, 0xE007A + 1))
 
-    # Cancel tag
+
-    tag_cps |= {0xE007F}
+def _map_missing_flag_tag_chars_to_empty_glyphs(colr_font):
+    # Add all tag characters used in flags + cancel tag
+    tag_cps = FLAG_TAGS | {CANCEL_TAG}
 
     # Anything already cmap'd is fine
     tag_cps -= set(_Cmap(colr_font).keys())
@@ -143,28 +146,6 @@ def _Cmap(ttfont):
     return functools.reduce(_Reducer, unicode_cmaps, {})
 
 
-def _map_empty_flag_tag_to_black_flag(colr_font):
-    # fontchain_lint wants direct support for empty flag tags
-    # so map them to the default flag to match cbdt behavior
-
-    # if the emoji font starts using extensions this code will require revision
-
-    cmap = _Cmap(colr_font)
-    black_flag_glyph = cmap[0x1F3F4]
-    cancel_tag_glyph = cmap[0xE007F]
-    lookup_list = colr_font["GSUB"].table.LookupList
-    liga_set = _ligaset_for_glyph(lookup_list, black_flag_glyph)
-    assert liga_set is not None, "There should be existing ligatures using black flag"
-
-    # Map black flag + cancel tag to just black flag
-    # Since this is the ligature set for black flag, component is just cancel tag
-    # Since we only have one component its safe to put our rule at the front
-    liga = ot.Ligature()
-    liga.Component = [cancel_tag_glyph]
-    liga.LigGlyph = black_flag_glyph
-    liga_set.insert(0, liga)
-
-
 def _add_vertical_layout_tables(cbdt_font, colr_font):
     upem_scale = colr_font["head"].unitsPerEm / cbdt_font["head"].unitsPerEm
 
@@ -201,6 +182,112 @@ def _add_vertical_layout_tables(cbdt_font, colr_font):
         vmtx.metrics[gn] = height, 0
 
 
+UNKNOWN_FLAG_PUA = 0xFE82B
+BLACK_FLAG = 0x1F3F4
+REGIONAL_INDICATORS = set(range(0x1F1E6, 0x1F1FF + 1))
+
+
+def _add_fallback_subs_for_unknown_flags(colr_font):
+    """Add GSUB lookups to replace unsupported flag sequences with the 'unknown flag'.
+
+    In order to locate the unknown flag, the glyph must be mapped to 0xFE82B PUA code;
+    the latter is removed from the cmap table after the GSUB has been updated.
+    """
+    cmap = _Cmap(colr_font)
+    unknown_flag = cmap[UNKNOWN_FLAG_PUA]
+    black_flag = cmap[BLACK_FLAG]
+    cancel_tag = cmap[CANCEL_TAG]
+    flag_tags = sorted(cmap[cp] for cp in FLAG_TAGS)
+    # in the *-noflags.ttf font there are no region flags thus this list is empty
+    regional_indicators = sorted(cmap[cp] for cp in REGIONAL_INDICATORS if cp in cmap)
+
+    classes = f'@FLAG_TAGS = [{" ".join(flag_tags)}];\n'
+    if regional_indicators:
+        classes += f"""
+            @REGIONAL_INDICATORS = [{" ".join(regional_indicators)}];
+            @UNKNOWN_FLAG = [{" ".join([unknown_flag] * len(regional_indicators))}];
+        """
+    lookups = (
+        # the first lookup is a dummy that stands for the emoji sequences ligatures
+        # from the destination font; we only use it to ensure the lookup indices match.
+        # We can't leave it empty otherwise feaLib optimizes it away.
+        f"""
+        lookup placeholder {{
+            sub {unknown_flag} {unknown_flag} by {unknown_flag};
+        }} placeholder;
+        """
+        + "\n".join(
+            ["lookup delete_glyph {"]
+            + [f"    sub {g} by NULL;" for g in sorted(regional_indicators + flag_tags)]
+            + ["} delete_glyph;"]
+        )
+        + (
+            """
+            lookup replace_with_unknown_flag {
+                sub @REGIONAL_INDICATORS by @UNKNOWN_FLAG;
+            } replace_with_unknown_flag;
+            """
+            if regional_indicators
+            else "\n"
+        )
+    )
+    features = (
+        "languagesystem DFLT dflt;\n"
+        + classes
+        + lookups
+        + "feature ccmp {"
+        + f"""
+            lookup placeholder;
+            sub {black_flag} @FLAG_TAGS' lookup delete_glyph;
+            sub {black_flag} {cancel_tag} by {unknown_flag};
+        """
+        + (
+            """
+            sub @REGIONAL_INDICATORS' lookup replace_with_unknown_flag
+                @REGIONAL_INDICATORS' lookup delete_glyph;
+            """
+            if regional_indicators
+            else ""
+        )
+        + "} ccmp;"
+    )
+    # feaLib always builds a new GSUB table (can't update one in place) so we have to
+    # use an empty TTFont and then update our GSUB with the newly built lookups
+    temp_font = ttLib.TTFont()
+    temp_font.setGlyphOrder(colr_font.getGlyphOrder())
+
+    addOpenTypeFeaturesFromString(temp_font, features)
+
+    temp_gsub = temp_font["GSUB"].table
+    # sanity check
+    assert len(temp_gsub.FeatureList.FeatureRecord) == 1
+    assert temp_gsub.FeatureList.FeatureRecord[0].FeatureTag == "ccmp"
+    temp_ccmp = temp_gsub.FeatureList.FeatureRecord[0].Feature
+
+    colr_gsub = colr_font["GSUB"].table
+    ccmps = [
+        r.Feature for r in colr_gsub.FeatureList.FeatureRecord if r.FeatureTag == "ccmp"
+    ]
+    assert len(ccmps) == 1, f"expected only 1 'ccmp' feature record, found {len(ccmps)}"
+    colr_ccmp = ccmps[0]
+
+    colr_lookups = colr_gsub.LookupList.Lookup
+    assert (
+        len(colr_lookups) == 1
+    ), f"expected only 1 lookup in COLRv1's GSUB.LookupList, found {len(colr_lookups)}"
+    assert (
+        colr_lookups[0].LookupType == 4
+    ), f"expected Lookup[0] of type 4 in COLRv1, found {colr_lookups[0].LookupType}"
+
+    colr_lookups.extend(temp_gsub.LookupList.Lookup[1:])
+    colr_gsub.LookupList.LookupCount = len(colr_lookups)
+    colr_ccmp.LookupListIndex = temp_ccmp.LookupListIndex
+    colr_ccmp.LookupCount = len(colr_ccmp.LookupListIndex)
+
+    # get rid of the Unknown Flag private codepoint as no longer needed
+    font_data.delete_from_cmap(colr_font, [UNKNOWN_FLAG_PUA])
+
+
 def main(argv):
     if len(argv) != 3:
         raise ValueError(
@@ -233,12 +320,12 @@ def main(argv):
 
     _map_missing_flag_tag_chars_to_empty_glyphs(colr_font)
 
-    _map_empty_flag_tag_to_black_flag(colr_font)
-
     add_soft_light_to_flags(colr_font)
 
     _add_vertical_layout_tables(cbdt_font, colr_font)
 
+    _add_fallback_subs_for_unknown_flags(colr_font)
+
     out_file = Path(_OUTPUT_FILE[colr_file.name]).absolute()
     print("Writing", out_file)
     colr_font.save(out_file)