Skip RBS rewrite when no markers are present

[mattkubej]

Summary

Avoid running Spoom's RBS-to-Sorbet-sig translator for typed Ruby files that cannot contain runtime-rewriteable RBS syntax.

Tapioca currently calls Spoom::Sorbet::Translate.rbs_comments_to_sorbet_sigs for every typed Ruby file loaded through the RBS rewriter. A large portion of typed files in Core do not contain RBS signatures or supported RBS annotations, so the translator often parses source only to return it unchanged.

This keeps the existing typed: gate and adds a conservative marker gate. Translation now runs only when source contains one of:

• #:
• #|
• # @abstract
• # @interface
• # @sealed
• # @final
• # @requires_ancestor:
• # @override
• # @overridable
• # @without_runtime

False positives are safe because they still use the existing translator. False negatives would be unsafe, so the annotation list is intentionally aligned with Spoom's currently supported runtime-rewrite annotations.

[paracycle]

This seems to be scanning the source string multiple times, can we turn the search into a Regexp.union and do a single scan through the source?

[paracycle]

I think something like:

SEARCH_REGEXP = Regexp.union(TYPED_FILE_PATTERN, Regexp.escape("#:"), Regexp.escape("#|"), *RBS_ANNOTATION_MARKERS.map { Regexp.escape(it) })

def should_rewrite?
  source.index(SEARCH_REGEXP)
end

should give you the decision to rewrite or not with a single pass.

[mattkubej]

Good call on the redundant scans. I just pushed an amend.

However, I kept typed_file? as a separate stage rather than folding into the union. A single union would introduce OR semantics, so a typed file with no markers would match and I think we'd want to avoid the expensive Spoom translator execution in this case.

[paracycle]

Ah, right, I forgot that the typed check wasn't being OR'ed. Good call.

[paracycle]

Looks good to me, but have one question

[paracycle]

Are you sure that none of these have special Regexp characters, or does Regexp.union do the right thing? I am surprised that we don't need Regexp.escape here.

[mattkubej]

Just sanity checked this. Looks like Regexp.union runs each String arg through Regexp.escape before joining.

Looking at the Ruby implementation, it looks like Regexp.escape and Regexp.quote share the same C function.

• https://github.com/ruby/ruby/blob/master/re.c#L4874-L4875
  • quote and escape map to rb_reg_s_quote
• https://github.com/ruby/ruby/blob/master/re.c#L4385
  • union calls rb_reg_s_quote for string case

Check to show union's source against manually escaped and joined version:

markers = ["#:", "#|", "# @abstract", "# @interface", "# @sealed", "# @final", "# @requires_ancestor:", "# @override", "# @overridable", "# @without_runtime"]
pattern = Regexp.union(markers)
pattern.source == markers.map { |m| Regexp.escape(m) }.join("|")
# => true
pattern.source
# => "\#:|\#\||\#\ @abstract|\#\ @interface|\#\ @sealed|\#\ @final|\#\ @requires_ancestor:|\#\ @override|\#\ @overridable|\#\ @without_runtime"

Documentation doesn't explicitly state this behavior, but the example seems to highlight this escaping behavior as well: https://docs.ruby-lang.org/en/3.4/Regexp.html#method-c-union