ruby-grape / grape / 33498797105

97%

master: 97%

Pull #2880

github

Default the versioner's pattern to nil instead of a regexp matching anything

`Versioner::Path#version_from_first_segment` runs on every request of a
path-versioned API and tested the candidate segment against `pattern`:

    return unless potential_version.match?(pattern)

`pattern` defaulted to `/.*/i`, which matches any segment including the empty
one. Unless an API supplied its own, that regexp ran on every request to answer
a question it could not answer with no.

Default it to nil, which is what the middleware is actually asking: whether a
pattern was supplied at all. "No constraint" is what `/.*/i` was standing in
for, and nil says it directly — the check becomes a nil test, and an API that
explicitly passes `pattern: /.*/i` keeps it, where inferring the answer from
the value would have discarded it as indistinguishable from the default.

`version 'v1', using: :path` cannot set this option: `Endpoint#build_stack`
does not pass one, and `DSL::Routing#version` does not accept the keyword. A
pattern reaches the middleware when an API inserts the versioner itself —

    use Grape::Middleware::Versioner::Path, versions: %w[v1], pattern: /v.+/

— which is undocumented but reachable and spec-pinned, and is unchanged: under
`/v.+/` a `/x1/...` segment is still not taken as a version, where the default
takes it.

What it saves is one regexp per request on a path-versioned API, measured in
one process at 57 ns for `seg.match?(/.*/i)` against 18 ns for the nil test.
That is roughly 0.7% of a minimal request, below what the version_throughput
harness can resolve, so no end-to-end figure is claimed. Allocation counts are
unchanged: `match?` builds no MatchData.

`Versioner::Base::DEFAULT_OPTIONS[:pattern]` and `Versioner::Base#pattern`
answer nil rather than `/.*/i` as a result. See UPGRADING.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

1236 of 1330 branches covered (92.93%)

Branch coverage included in aggregate %.

1 of 1 new or added line in 1 file covered. (100.0%)

3910 of 3978 relevant lines covered (98.29%)

27261.36 hits per line