FEA parser and glyph names containing hyphens

Consider this feature file,

@A_CLASS = [ pre-post ];

I get a traceback when trying to parse it with this snippet,

from fontTools.feaLib.parser import Parser

with open('file.fea') as fea_file:
    parser = Parser(fea_file)
    doc = parser.parse()
    print(doc.asFea())

Traceback (most recent call last):
  File "fea_parse.py", line 7, in <module>
    doc = parser.parse()
  File "fontTools/feaLib/parser.py", line 61, in parse
    statements.append(self.parse_glyphclass_definition_())
  File "fontTools/feaLib/parser.py", line 221, in parse_glyphclass_definition_
    glyphs = self.parse_glyphclass_(accept_glyphname=False)
  File "fontTools/feaLib/parser.py", line 294, in parse_glyphclass_
    start, limit = self.split_glyph_range_(glyph, location)
  File "fontTools/feaLib/parser.py", line 259, in split_glyph_range_
    "into a range of known glyphs" % name, location)
fontTools.feaLib.error.FeatureLibError: file.fea:2:14: "pre-post" is not a glyph in the font, and it can not be split into a range of known glyphs

However, all is fine if I initialize the parser with the glyphNames argument,

parser = Parser(fea_file, glyphNames=['pre-post'])

Is this by design, or does the parser not fully support glyph names that contain hyphens?

[justvanrossum]

It seems odd that the parser needs to know ahead of time which tokens may be glyph names, but I’ll take it. Thanks for confirming it.

I think it’s a logical consequence of changing the language to allow hyphens so late in the game. The hyphen had a meaning before (a range) and now there’s an ambiguity that can only be resolved in a bw compatible way with knowledge of the glyph set.

[brawer]

Personally, I wouldn’t have strong opinions, but (as with similar requests) I don’t think we should make incompatible changes without very good reasons. So, my process recommendation would be:

1. Make a concrete proposal for changing the spec;
2. get consensus with Adobe (in past discussions, they’ve always been very reasonable);
3. only then, change the implementation.

Otherwise, you’ll fragment the file format, which would hurt everyone.