Performance/RegexpMatch: Does not consider ::Regexp.last_match(1)

[DannyBen]

I noticed a couple of issues with this cop.

1. It does not recognize ::Regexp.last_match(1) as use of MatchData
2. It gets confused and assumes use of MatchData if a following unrelated line mentions MatchData-like statements.

---

Expected behavior

Expectation 1: ::Regexp.last_match(1) should be detected as MatchData use

Expecting the cop detect ::Regexp.last_match(1) as use of MatchData and not trigger an offense (in the same way it detects Regexp.last_match(1) or $1).

# Expected not to fail
::Regexp.last_match(1) if string =~ /hello (.*)/

Expectation 2: Offense should be reported

The first line below does not use MatchData, the cop is expected to report an offense. Removing the second (unrelated) line makes it work as expected.

p 'hello' if string =~ /hello (.*)/
p $1 if string =~ /hello (.*)/

Actual behavior

This code triggers an offense where it shouldn't:

p ::Regexp.last_match(1) if string =~ /hello (.*)/

This code does not trigger an offense, where it should:

p 'hello' if string =~ /hello (.*)/
p $1 if string =~ /hello (.*)/

Steps to reproduce the problem

# .rubocop.yml

require: rubocop-performance

AllCops:
  NewCops: enable
  TargetRubyVersion: 3.0

# test.rb

string = 'hello bob'

# FAIL: Offense is reported on the last line. The first two lines pass as
#       expected
p $1 if string =~ /hello (.*)/
p Regexp.last_match(1) if string =~ /hello (.*)/
p ::Regexp.last_match(1) if string =~ /hello (.*)/

# FAIL: Offense is NOT reported on first line since we use $1 in the
#       unrelated next line
p 'hello' if string =~ /hello (.*)/
p $1 if string =~ /hello (.*)/

then run rubocop --only Performance and get this output (trimmed for readability):

➜ rubocop --only Performance
test.rb:7:29: C: [Correctable] Performance/RegexpMatch: Use match? instead of =~ when MatchData is not used.
p ::Regexp.last_match(1) if string =~ /hello (.*)/
                            ^^^^^^^^^^^^^^^^^^^^^^

Additional consideration

Note that normally I am using $1 and not Regexp.last_match (and definitely not ::Regexp.last_match), but the native Style/PerlBackrefs cop autocorrected my code to ::Regexp.last_match.

The point being - if one cop is saying "use ::Regexp.last_match", then the other cop should be aware of it.

RuboCop version

$ rubocop -V
1.39.0 (using Parser 3.1.2.1, rubocop-ast 1.23.0, running on ruby 3.1.0) [x86_64-linux]
  - rubocop-performance 1.15.1

[fatkodima]

This code does not trigger an offense, where it should:

p 'hello' if string =~ /hello (.)/
p $1 if string =~ /hello (.)/

For the second case it should not trigger an offense, because of $1.

[DannyBen]

For the second case it should not trigger an offense, because of $1.

Consider the two lines together. The first should trigger an offense, but it does not since there is an unrelated $1 beneath it. Switching the order of the lines does trigger. Or is it by design that any subsequent line with $1 is considered as part of the "block"?

[fatkodima]

Or is it by design that any subsequent line with $1 is considered as part of the "block"?

Seems like yes.

[DannyBen]

Seems like yes.

I understand the reasoning.
But the first issue is a bug, yes? the ::Regexp.last_match(1).

[fatkodima]

Yes, I opened a PR with the fix.