v2.33.2

• Allow to reverseDirection=True even when convertCubics=False (#770).

v2.33.1

• Added convertCubics=True to compileInterpolatableTTFs/compileVariableTTF methods, to allow disabling cu2qu altogether (similar to the existing option that was already there for compileTTF for statics).
• OutlineTTFCompiler: ensure that only quadratic curves are present in TrueType glyphs if glyphDataFormat=0 was not explicitly set to 1 (#769).

v2.33.0

• Support public.openTypePostUnderlinePosition lib key meant to clarify ambiguous meaning of underline position in post vs CFF table (unified-font-object/ufo-spec#217, #760).
• [instructionCompiler] Make logging less noisy (#767).
• Added allQuadratic=True compile option; when False, tells cu2qu to keep mix of cubic and quadratic curves (whichever has fewer points) as per glyf v1 draft spec proposal (#768, https://github.com/harfbuzz/boring-expansion-spec/blob/main/glyf1.md).

v2.32.0

• Raise error when multiple glyphs claim the same Unicode codepoint. When expanding color layers to glyphs, ensure any codepoint is stripped to prevent unintentional clashes (#739).
• Use fontTools.cu2qu instead of standalone cu2qu module, no changes, just one less dependency (#756).
• Added compile option for static/variable TTF outputs to dropImpliedOnCurves to omit on-curve points that are halfway between two consecutive off-curve points, slightly reducing glyf table size with no visible differences (#753).
• [constants] Updated Indic / USE script lists (#755).
• Require fonttools >= 4.40.0

v2.31.1

• [featureWriter.ast] Have iterClassDefinitions search not just top-level but inside feature and lookup blocks as well (#732).
• Added autoUseMyMetrics=True compile option (for TrueType output only) to allow disabling the automatic setting of USE_MY_METRICS glyf component flags, which is not needed for fonts without hinted metrics (#741).
• [kernFeatureWriter] Only use exported glyphs when generating mark filtering sets (#740).

v2.31.0

• Dropped support for Python 3.7 which reached EOL. We now require Python >= 3.8.
• Take into account extra substitutions found in DesignSpace rules when classifying glyphs by various unicode properties in feature writers (#731).
• Added support for compiling TrueType instructions and flags stored in the UFO's public.truetype.* lib keys (#654).
• Automatically compute COLRv1 ClipBoxes if none are explicitly provided. This is enabled by default for static builds only. You can set colrAutoClipBoxes=True for interpolatable/variable builds if safe to do so (#728).
• Improved "no default source" error (#677).
• Warn about mis-cased # Automatic Code insertion marker comments; also warn if there are no insertion markers but kerning.plist is non empty and there are manual kern/dist features in the FEA code (#719).
• Respect openTypeNameRecords overrides for the predefined name IDs (#727).
• Reorder class definitions in feature writers such that they appear before the lookups that use them (#725).
• Renamed dottedCircleFilter module to dottedCirlce, was wrongly named; old kept for backward compat (#716).
• [kernFeatureWriter] Avoid overwriting classDefs (#705).
  Register kern lookups for all defined languages (#711).
  Correctly split mark and base kerning pairs (#713).
  Don't skip over spacing marks when kerning (#720).
• Fixed FEA include dir detection when in current dir (#689)
• Automatically decompose components when 2x2 transforms differ (#698)
• [kernFeatureWriter] Split kerning by script, not horizontal direction, which should be faster to compile and produce smaller GPOS. The old kernFeatureWriter is still kept in a separate module (#679, #636, #697).

v2.31.0b3

• Dropped support for Python 3.7 which reached EOL. We now require Python >= 3.8.
• Added support for compiling TrueType instructions and flags stored in the UFO's public.truetype.* lib keys (#654).
• Automatically compute COLRv1 ClipBoxes if none are explicitly provided. This is enabled by default for static builds only. You can set colrAutoClipBoxes=True for interpolatable/variable builds if safe to do so (#728).
• Improved "no default source" error (#677).
• Warn about mis-cased # Automatic Code insertion marker comments; also warn if there are no insertion markers but kerning.plist is non empty and there are manual kern/dist features in the FEA code (#719).
• Respect openTypeNameRecords overrides for the predefined name IDs (#727).
• Reorder class definitions in feature writers such that they appear before the lookups that use them (#725).
• Renamed dottedCircleFilter module to dottedCirlce, was wrongly named; old kept for backward compat (#716).
• [kernFeatureWriter] Avoid overwriting classDefs (#705).
  Register kern lookups for all defined languages (#711).
  Correctly split mark and base kerning pairs (#713).
  Don't skip over spacing marks when kerning (#720).
• Fixed FEA include dir detection when in current dir (#689)
• Automatically decompose components when 2x2 transforms differ (#698)
• Copy old KernFeatureWriter from ufo2ft 2.30.0 into separate module (#697)

v2.31.0b2

Make it possible to kern Hiragana and Katakana against each other (#683)

v2.31.0b1

• Split kerning by script, not direction (#679). This is a rework of #636 and should now properly split up kerning into per-script lookups without losing any. The major upside is that multi-script fonts with a lot of kerning will compile faster and smaller, because lookups will overflow less often and become denser.
• As a side-effect, kerning will now also be generated such that the shaper will actually apply it when the designer did not explicitly declare an appropriate languagesystem in the font's features. Before, kerning for Greek glyphs was effectively dropped unless the designer declared languagesystem grek dflt;.
• Another side-effect is that kerning between some common glyphs like increment and asterisk will now always be applied, where before due to circumstances it might have been stuck in e.g. the LTR instead of the Common lookup and therefore not applied if the script wasn't set correctly.

v2.30.0

• Improve mark feature writer for characters that can be used in multiple scripts and some of them require mark feature, while others require abvm/blwm features (#579).