Lint for redundant keyword_init on structs

[aldhsu]

Is your feature request related to a problem? Please describe.

Ruby 3.2 enables keyword_init on structs without specifying it. https://www.ruby-lang.org/en/news/2022/12/25/ruby-3-2-0-released/

# before 3.2
foo = Struct.new(:bar, keyword_init: true)

# equivalent after 3.2
foo = Struct.new(:bar)

Describe the solution you'd like

A lint for redundant keyword_init: true on structs.

[koic]

This change introduces a behavioral incompatibility with the Struct.keyword_init? method, so it may be worth considering whether "redundant" is an appropriate designation.

Struct.new(:bar, keyword_init: true).keyword_init? # => true
Struct.new(:bar).keyword_init? # => nil

[aldhsu]

Good catch @koic!
Do you think that behaviour makes sense for the keyword_init? method?

The default has changed. Consumers relying on !Struct.new(:bar).keyword_init? will not be aware of the change. I'm not sure what the patterns usually are for ? methods in Ruby. However, I think in my experience, the expectation is that nil will be treated as falsy rather than truth-y for ? methods.

[koic]

Yeah, RuboCop is used widely across applications, libraries, and frameworks, and user programs can include unexpected cases. Rather than focusing on whether it is a predicate method, it is important to consider the potentials for altering existing behavior, which differs from layout changes.

[aldhsu]

OK that makes sense.

consider the potentials for altering existing behavior, which differs from layout changes.

Would it make sense to have the cop disabled by default with the safety warning that keyword_init? doesn't respond the same when using keyword_init explicitly?

[koic]

Technically, for Ruby 3.2+, it seems possible to pending it by default with SafeAutoCorrect: false.