Frozen String Literal Behavior is Confusing

[gfyoung]

According to the docs, my understanding was that in order for the cop to be satisfied, we had to add # frozen_string_literal: true to the top of the file.

However, after finding what appeared to be a false negative in one of my projects, I found #7241 and its corresponding fix in #7251 that highlighted that adding # frozen_string_literal: false to the top can still pass this cop. Two points:

• I think the docs should be updated to reflect that false is accepted in the magic comment
• Is there a way to configure the cop so that it enforces # frozen_string_literal: true only?

Rubocop Info

0.75.0 (using Parser 2.6.5.0, running on ruby 2.6.5 x86_64-linux)

[koic]

Good points.

I think the docs should be updated to reflect that false is accepted in the magic comment

Yeah, it will be easy to understand if it is documented.

Is there a way to configure the cop so that it enforces # frozen_string_literal: true only?

AFAIK, there is no such configure option, so enhancement will be necessary.

[gfyoung]

@koic : Thanks for your reply!

I imagine a docs change is something people wouldn't object to.

The enhancement would be nice. Would be curious to know if people have any objections.

[olliebennett]

I assumed the Style/FrozenStringLiteralComment cop already checked for true, to be honest.

For reference, current RuboCop default configuration options are;

  EnforcedStyle: always
  SupportedStyles:
    # `always` will always add the frozen string literal comment to a file
    # regardless of the Ruby version or if `freeze` or `<<` are called on a
    # string literal. If you run code against multiple versions of Ruby, it is
    # possible that this will create errors in Ruby 2.3.0+.
    - always
    # `never` will enforce that the frozen string literal comment does not
    # exist in a file.
    - never

Would it make sense to add a new SupportedStyle of alwaystrue?

    # `alwaystrue` will enforce that the frozen string literal comment exists
    # and is set to `true` (i.e. enforcing that string literals _are_ frozen). 
    - alwaystrue

I think defaulting to always (the current default) is most intuitive. If someone explicitly uses frozen_string_literal: false they are intentionally not freezing strings, rather than accidentally not freezing strings. An alwaystrue configuration option hopefully solves everyone's use-cases, while maintaining backwards compatibility.

[gfyoung]

@olliebennett : My thoughts exactly. I wanted to see if there were any maintainer objections though on this before proceeding with an enhancement.