[cheesepaulo]

Ran 0.1.2 against a private app going from Rails 7.1 to 7.2. fixture_path_assignment matched 7 lines in its spec helpers, all a local variable:

fixture_path = build_fixture_path(route, mock_data_file)

Ruby reads a bare name = x as a local variable, never a call to the setter, so the removed fixture_path= always has a receiver. The pattern goes from \bfixture_path\s*= to \.fixture_path\s*=(?!=):

Line	Before	After
self.fixture_path = x	hit	hit
config.fixture_path = "spec/fixtures"	hit	hit
ActiveSupport::TestCase.fixture_path = x	hit	hit
fixture_path = build(route)	hit	—
@fixture_path = x	hit	—
self.fixture_path == x	hit	—
config.fixture_paths = [x]	—	—

Two lines go into NEAR_MISSES (the local and the instance variable). With the old pattern, test_replacement_apis_are_not_flagged fails on both; with the new one the suite passes (97 runs, 0 failures). On the app, the 7 hits go to 0 and nothing else in the report changes (blockers 3 → 2).

🤖 Generated with Claude Code

[aslam]

Thanks, this is right. A bare fixture_path = x can only ever be a local variable, so making the pattern require a receiver drops the false positives and loses no real hits. I ran the suite on this branch and it passes.

One small optional change: obj.fixture_path => 1 and obj.fixture_path =~ /a/ still match. Changing the lookahead to (?![=~>]) excludes both:

pattern: "\\.fixture_path\\s*=(?![=~>])"

Neither is likely in real code, though.