The Font Engineering with Extensibility (FEE) language is implemented via a Python library, fontFeatures
. This format improves over Adobe FEA (.fea
) in several important ways, and compiles to FEA. In future, FEE also will compile to raw GPOS/GSUB binary tables.
Class definition is a time consuming element of writing FEA code. FEE allows regular expressions to be used to define classes; each glyph name in the font is tested against the regular expression, and the glyphs added:
DefineClass @sc = /\.sc$/;
Ran as:
fee2fea tests/LibertinusSans-Regular.otf test.fee
Results in:
@sc = [parenleft.sc parenright.sc bracketleft.sc bracketright.sc ...];
Simple replacement can be done as easily as:
DefineClass @sc = /\.sc$/;
DefineClass @sc_able = @sc~sc;
Feature smcp {
Substitute @sc_able -> @sc;
};
Quite complex classes can be built. All the glyphs which have a smallcap and alternate form:
DefineClass @sc_and_alt_able = /.*/ and hasglyph(/$/ .alt) and hasglyph(/$/ .sc);
Returning:
@sc_and_alt_able = [h y germandbls];
FEE can even do substitutions impossible in FEA. For example:
DefineClass @digits = U+0031=>U+0039; # this is range(U+0031, U+0039) inclusive
DefineClass @encircled = U+2474=>U+247C;
# Un-CJK parenthesize
Feature ss01 {
Substitute @encircled -> parenleft @digits parenright;
};
Gives us:
feature ss01 {
lookup Routine_1 {
sub uni2474 by parenleft one parenright;
sub uni2475 by parenleft two parenright;
sub uni2476 by parenleft three parenright;
sub uni2477 by parenleft four parenright;
sub uni2478 by parenleft five parenright;
sub uni2479 by parenleft six parenright;
sub uni247A by parenleft seven parenright;
sub uni247B by parenleft eight parenright;
sub uni247C by parenleft nine parenright;
} Routine_1;
} ss01;
FEE can do much more; see the plugins documentation. Writing your own plugins is as simple as defining its grammar, verb, and adding a class with an action()
method.
OpenType fonts are "programmed" using features, which are normally authored in Adobe's feature file format. This like source code to a computer program: it's a user-friendly, but computer-unfriendly, way to represent the features.
Inside a font, the features are compiled in an efficient internal format. This is like the binary of a computer program: computers can use it, but they can't do else anything with it, and people can't read it.
The purpose of this library is to provide a middle ground for representing features in a machine-manipulable format, kind of like the abstract syntax tree of a computer programmer. This is so that:
- features can be represented in a structured human-readable and machine-readable way, analogous to the XML files of the Unified Font Object format.
- features can be more directly authored by programs (such as font editors), rather than them having to output AFDKO feature file format.
- features can be easily manipulated by programs - for example, features from two files merged together, or lookups moved between languages.
How is this different from fontTool's
feaLib
? I'm glad you asked.feaLib
translates between the Adobe feature file format and a abstract syntax tree representing elements of the feature file - not representing the feature data. The AST is still "source equivalent". For example, when you code anaalt
feature in feature file format, you might include code likefeature salt
to include lookups from another feature. But what's actually meant by that is a set of lookups.fontFeatures
allows you to manipulate meaning, not description.
fontFeatures consists of the following components:
fontFeatures
itself, which is an abstract representation of the different layout operations inside a font.fontFeatures.feaLib
(included as a mixin) which translates between Adobe feature syntax and fontFeatures representation.fontFeatures.ttLib
, which translates between OpenType binary fonts and fontFeatures representation. (Currently only OTF ->fontFeatures
is partially implemented; there is nofontFeatures
-> OTF compiler yet.)fontFeatures.feeLib
which parses a new, extensible format called FEE for font engineering.fontFeatures.fontDameLib
which translate FontDame text files into fontFeatures objects.
And the following utilities:
fee2fea
: translates a FEE file into Adobe feature syntax.otf2fea
: translates an OTF file into Adobe features syntax.txt2fea
: translates a FontDame txt file into Adobe features syntax.mergeFee
: takes an existing font, adds FEE rules to it, and writes it out again.