Merge pull request #679 from googlefonts/kern-splitter-v3

Split kerning by script, not by direction (third attempt)

Lib/ufo2ft/constants.py

@@ -38,6 +38,8 @@
 
 UNICODE_VARIATION_SEQUENCES_KEY = "public.unicodeVariationSequences"
 
+COMMON_SCRIPT = "Zyyy"
+
 INDIC_SCRIPTS = [
     "Beng",  # Bengali
     "Cham",  # Cham

Lib/ufo2ft/featureCompiler.py

@@ -281,8 +281,14 @@ def setupFeatures(self):
             if self.featureWriters:
                 featureFile = parseLayoutFeatures(self.ufo, self.feaIncludeDir)
 
+                path = self.ufo.path
                 for writer in self.featureWriters:
-                    writer.write(self.ufo, featureFile, compiler=self)
+                    try:
+                        writer.write(self.ufo, featureFile, compiler=self)
+                    except FeatureLibError:
+                        if path is None:
+                            self._write_temporary_feature_file(featureFile.asFea())
+                        raise
 
                 # stringify AST to get correct line numbers in error messages
                 self.features = featureFile.asFea()
@@ -316,15 +322,16 @@ def buildTables(self):
                 addOpenTypeFeaturesFromString(self.ttFont, self.features, filename=path)
             except FeatureLibError:
                 if path is None:
-                    # if compilation fails, create temporary file for inspection
-                    data = self.features.encode("utf-8")
-                    with NamedTemporaryFile(delete=False) as tmp:
-                        tmp.write(data)
-                    logger.error(
-                        "Compilation failed! Inspect temporary file: %r", tmp.name
-                    )
+                    self._write_temporary_feature_file(self.features)
                 raise
 
+    def _write_temporary_feature_file(self, features: str) -> None:
+        # if compilation fails, create temporary file for inspection
+        data = features.encode("utf-8")
+        with NamedTemporaryFile(delete=False) as tmp:
+            tmp.write(data)
+        logger.error("Compilation failed! Inspect temporary file: %r", tmp.name)
+
 
 class MtiFeatureCompiler(BaseFeatureCompiler):
     """Compile OpenType layout tables from MTI feature files using

Lib/ufo2ft/featureWriters/baseFeatureWriter.py

@@ -2,6 +2,8 @@
 from collections import OrderedDict, namedtuple
 from types import SimpleNamespace
 
+from fontTools import unicodedata
+
 from ufo2ft.constants import OPENTYPE_CATEGORIES_KEY
 from ufo2ft.errors import InvalidFeaturesData
 from ufo2ft.featureWriters import ast
@@ -122,8 +124,12 @@ def shouldContinue(self):
     def write(self, font, feaFile, compiler=None):
         """Write features and class definitions for this font to a feaLib
         FeatureFile object.
-        Returns True if feature file was modified, False if no new features
-        were generated.
+
+        The main entry point for the FeatureCompiler to any of the
+        FeatureWriters.
+
+        Returns True if feature file was modified, False if no new features were
+        generated.
         """
         self.setContext(font, feaFile, compiler=compiler)
         try:
@@ -386,3 +392,34 @@ def getGDEFGlyphClasses(self):
             frozenset(marks),
             frozenset(components),
         )
+
+    def guessFontScripts(self):
+        """Returns a set of scripts the font probably supports.
+
+        This is done by:
+
+        1. Looking at all defined codepoints in a font and remembering the
+           script of any of the codepoints if it is associated with just one
+           script. This would remember the script of U+0780 THAANA LETTER HAA
+           (Thaa) but not U+061F ARABIC QUESTION MARK (multiple scripts).
+        2. Adding explicitly declared `languagesystem` scripts on top.
+        """
+        font = self.context.font
+        glyphSet = self.context.glyphSet
+        feaFile = self.context.feaFile
+        single_scripts = set()
+
+        # First, detect scripts from the codepoints.
+        for glyph in font:
+            if glyph.name not in glyphSet or glyph.unicodes is None:
+                continue
+            for codepoint in glyph.unicodes:
+                scripts = unicodedata.script_extension(chr(codepoint))
+                if len(scripts) == 1:
+                    single_scripts.update(scripts)
+
+        # Then, add explicitly declared languagesystems on top.
+        feaScripts = ast.getScriptLanguageSystems(feaFile)
+        single_scripts.update(feaScripts.keys())
+
+        return single_scripts