Analyzer::Ruby::Grape

Grape::API is absent from almost every .rb file in a large repo, so String#includes? scanned each one end to end. One precompiled matcher via content_matches? is the same test in a single pass.

class <Child> < <Super> declaration matcher (single capture: Super).

Full form capturing both Child and Super for the inheritance graph.

A route or prefix statement whose options continue on the next line (namespace 'conans/:name', / requirements: {...} do).

%w[a b].each do |x|, [:a, :b].each do |x| and { "a" => 1, "b" => 2 }.each do |x, y| over a literal receiver.

mount Child / mount Child => "/path" (capture 2: the mount path).

build_grape_index's whole-tree pre-pass gate OR-ed two String#includes? scans over every file's content as a standalone boolean gate. Folded into one precompiled union so it costs a single PCRE2 match instead of up to two naive substring scans per file.

resource source_type.pluralize do outside a static loop: the name is only known at runtime, so the segment becomes a {source_type} placeholder instead of vanishing from the path.

Crystal recompiles an interpolated regex literal (/^#{verb}.../) on every evaluation — ~11000x slower than a precompiled one — so matching the verb DSL inside the per-line loop used to recompile 14 regexes for every line of every Grape file. Build the per-verb patterns once at load time instead. _WITH_PATH matches get '/users' do (capturing the path literal/symbol) and the path-less get urgency: :low do (options only); _BARE_DO matches the path-less get do form.

Unrolling is multiplicative and noir scans untrusted repos, so it is bounded (real Grape loops are a handful of literals, one level deep).

Instance-side view of the same declaration. The per-file rescues live on this base class, which has no way to name the analyzer that is running inside them, so a skipped file could not be attributed to a tech. Deriving it from analyzer_for keeps the name written exactly once.