Namespaces for cop names

[jonas054]

Based on the discussion in rubocop/rubocop-rspec#3

I'm working on a solution for namespaces and would like to discuss it. These are the ideas:

1. Cop names can be prefixed with a namespace in lower case, so Encoding becomes style/Encoding, etc. This is done in .yml configuration files, in arguments to the --only option, and in inline comments, e.g., #rubocop:disable style/ClassLength.
2. Output from --display-cop-names and --show-cops will include the namespace qualifiers.
3. The namespace qualifiers are optional. There's a namespace lookup that will find the right namespace if none was given. If the cop name is ambiguous, execution will stop with an error message.
4. No warning will be printed for missing namespace when a single namespace is found for the cop name.

[bbatsov]

All points sound good to me. I also wonder if we shouldn't make it possible to use snake_case for cop names and config options, as I don't think the our pervasive use of CamelCase (or rather PascalCase) is particularly popular (or common) in the Ruby community.

[geniou]

Sounds good to me, just two things:

I think it would be good to be consistent by the writing of the namespace and the cop name, so both should be CamelCase, or both snake_case - in my opinion.

Also I think the new namespace should be added to the generated todo file.

[yujinakayama]

I agree with @geniou.

[bbatsov]

My personal preference would be to use snake_case. The only reason we started with CamelCase names in the beginning was it was easier to just use the class names directly in the config. @jonas054 would correct me if I'm missing something. :-) Of course, this would be a serious change, so if we go this way we should keep supporting old naming scheme at least for one release.

[nevir]

+1 to consistent snake_case OR CamelCaps usage.

IMO, in the YAML file, the namespaces should be actual YAML keys: (organization, DRY). Downside being that you can no longer just append cops to the bottom of the file.

Also, it might be worth supporting both snake_case, and CamelCaps:

• snake_case is great for human-facing configuration (and just general Ruby consistency)
• CamelCaps is great for tools/scripts (plus, Rubocop owns the snake_case <-> CamelCaps encoding that way)
• Or at least exposing snake/camel conversion utils as part of the public surface area

[jonas054]

@geniou Namespaces will be added to .rubocop_todo.yml. I forgot to mention that.

@nevir I'm not sold on the idea of having namespace keys with cop name keys under it. We would lose the uniformity of name references. Being able to put style/LineLength in .rubocop.yml, after --only, and after # rubocop:disable is a great advantage. You're right that it would be more dry, though.

About case... I think I was influenced by our internal use of cop types :lint and :rails, and perhaps also by conventions from C++, when I chose to have lower case namespaces with cop names still in CamelCase.

I'll change the namespaces to CamelCase, but I really don't feel like changing the cop names at this point, or supporting two naming styles. The gain would be small IMO.

[bbatsov]

Let's go with @jonas054's idea for now. Btw, I think that the namespace can also be a cop key.