You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In an ongoing effort to make this project's format database less hostile
towards ports and third-party integrations, a dubious novelty introduced
in v2 has been removed: the supposed "fuzzy" matching of language names.
This feature attempted to maximise the perceived "accuracy" of generated
expressions by substituting non-alphabetic sequences with ones that were
more likely to match something unambiguous:
1. Word boundaries would match `[\W_ \t]?` instead of a space or dash,
2. Zeroes would match the letter "O" (and vice versa),
4. Punctuation in the original string was made optional, and
3. Words with medial caps were blindly assumed to be camelCased, which
resulted in poor handling of names with non-standard capitalisation
(e.g., "ECLiPSe").
This "promiscuous" matching is problematic for the import script used by
the VSCode port, which enumerates the possibilities of finite regexes to
generate a flat list of names for JSON encoding. Moreover, its relevance
in Atom is questionable at best; at worst, hopelessly counter-intuitive.
Aside from eliminating the aforementioned behaviour, this commit updates
the compiler logic to retain an icon-name's original case when possible,
even though the resulting expression is case-insensitive. This gives one
more permutation of capitalisation to use in VSCode, which matches names
case-sensitively.
0 commit comments