Mixin grouping

[marcandre]

This is 3 fixes for the price of one.

First, this cop was doing multiple replacement rules for the same range. Only the first one was the correct one, so the tests were passing only because Rubocop silently swallows the ClobberingError of the erroneous subsequent replacements.

Second, the replacement was nuking anything in between two mixins

Thirdly, mixins with receivers could also create invalid code.

Note: Part of investigating parser's rewritter...

[bbatsov]

Apart from the failing build, the changes look good to me.

[marcandre]

Hopefully build will pass now.

FWIW, I find it illuminating that my natural coding style seems to conflict so badly with rubocop's conventions, even though until this moment I considered myself an adept of the Ruby Style Guide.

I must admit that my interest in fixing errors in Rubocop went down because of this.

[bbatsov]

FWIW, I find it illuminating that my natural coding style seems to conflict so badly with rubocop's conventions, even though until this moment I considered myself an adept of the Ruby Style Guide.

I didn't notice anything particularly off in your style? I assumed the failure was coming from some misalignment or some long line.

I must admit that my interest in fixing errors in Rubocop went down because of this.

I'm sad to hear this, but you know how it is - consistency within a codebase is important. Everything in RuboCop is configurable, what you're actually disliking are likely just own preferences (the project defaults) here.

[marcandre]

I'm sad to hear this, but you know how it is - consistency within a codebase is important. Everything in RuboCop is configurable, what you're actually disliking are likely just own preferences (the project defaults) here.

Indeed 😄 I'm now officially strongly against the settings of the ABC metric, the method length metric and the line length metric.

It looks like I might not give the same importance to style over strict correctness. If I had to choose between which one of these metrics or #5094's validations that "are not needed, they are restricted unnecessarily", or if it's OK to silently swallow exceptions, my choice is clear.

BTW, I must applaud the responsive team, impressively quick feedback! 👏