Silence parser warning and patch version mismatch when loading parser/current

[navidemad]

Description

This pull request adds a factory class RubyParserFactory to temporarily disable verbose mode when loading the "parser/current" library. The purpose is to suppress compatibility warnings that are displayed when the library is loaded with a different Ruby version than the one it recognizes.

Example warning for the release of Ruby 3.4.1 :

warning: parser/current is loading parser/ruby34, which recognizes
3.4.0-compliant syntax, but you are running 3.4.1.
Please see https://github.com/whitequark/parser#compatibility-with-ruby-mri.

By disabling verbose mode, these warnings are suppressed to provide a cleaner output and avoid confusion. The verbose mode is restored after the parser instance is created to maintain the original behavior.

References

• Bump Ruby 3.4.1 whitequark/parser#1056
• marcandre/rubocop-ast@7081f7c
• Pin the version of parser/current to whatever Ruby version we're using dependabot/dependabot-core#5461
• Parser issue rubocop/rubocop#9700

Changes

• Added a new module I18n::Tasks::Scanners to encapsulate the factory class.
• Implemented the RubyParserFactory class with a create_parser method that temporarily disables verbose mode, loads the "parser/current" library, creates a new parser instance, and restores the verbose mode.
• Added descriptive comments to explain the purpose of the module and provide an example of the warning being suppressed.

Benefits

• Suppresses compatibility warnings when loading the "parser/current" library, providing a cleaner output for users.
• Encapsulates the logic for creating a parser instance within a dedicated factory class, promoting code organization and reusability.
• Maintains the original behavior of the Ruby environment by restoring the verbose mode after the parser instance is created.

Issues

• Silence parser warning and patch version mismatch when loading parser/current #612
• Bump Ruby 3.4.1 whitequark/parser#1056

Please let me know if you have any questions or feedback regarding this pull request.
Thank you for considering this contribution to the gem!

[glebm]

I think a much nicer solution to this would be to send a PR to whitequark/parser adding an argument to Parser::CurrentRuby.new that disables the warning

[navidemad]

I made whitequark/parser#1065 we will see, but yeah you are right fixing it on the parser itself it's better, than hotfixing here.

[guillaumewrobel]

They won't update their parser with Ruby >3.4 so we're stuck

[davidwessman]

LGTM, interesting to see how this will play out with rubocop and the parsing ecosystem of Ruby.

[navidemad]

I do not know why but with my PR

bundle exec i18n-tasks translate-missing --backend=openai 
bundler: failed to load command: i18n-tasks (/Users/dev/.rbenv/versions/3.4.1/lib/ruby/gems/3.4.0/bin/i18n-tasks)                                                                                                                           >  ETA: ??:??:?? 0/161 (0%)
/Users/dev/.rbenv/versions/3.4.1/lib/ruby/gems/3.4.0/gems/faraday-2.12.2/lib/faraday/response/raise_error.rb:30:in 'Faraday::Response::RaiseError#on_complete': the server responded with status 400 (Faraday::BadRequestError)

But works with the original repository, do you have an idea ?

I did a little bit by editing the OpenAi::Client to:

    def translator
      @translator ||= OpenAI::Client.new(access_token: api_key, log_errors: true) do |f|
        f.response :logger, Logger.new($stdout), bodies: true
      end
    end

I got:

INFO -- : response: {
  "error": {
    "message": "'messages' must contain the word 'json' in some form, to use 'response_format' of type 'json_object'.",
    "type": "invalid_request_error",
    "param": "messages",
    "code": null
  }
}

[navidemad]

I found the problem, the prompt has to have the word json in the prompt to be able to use response_format with json_object.
But i have no clue why it is happening on my repository fork and not on yours.

[davidwessman]

@navidemad please split out the OpenAI logging to a separate PR 🙂

[navidemad]

@davidwessman

I split the changes into three PR as requested:

• Append JSON instructions to OpenAI system prompt #615
• OpenAI: log errors #614
• Silence parser warning and patch version mismatch when loading parser/current #613

[guillaumewrobel]

hey @davidwessman do you think this PR is ready to be merged?

[davidwessman]

@navidemad can you rebase and fix conflicts?

[navidemad]

@davidwessman Done

[davidwessman]

@navidemad please fix the lint error

[navidemad]

@navidemad please fix the lint error

Done