Custom RuboCop rule to catch .join("|") in interpolated regexes?

[jneen]

I'm not a huge expert on RuboCop, but as I am cleaning up a bunch of huge regexes, I'm wondering if it's possible to have a custom cop to catch this kind of situation?

Basically anything that looks like:

%r/#{something.join('|')}/

[jneen]

@larouxn thoughts on whether this is feasible?

[larouxn]

👋 To confirm, is the idea is we want to catch and prevent usage of .join("|") and the like to ensure regexps are explicit/clear? Makes sense to me. As for feasibility, it should be doable. For reference, a few resources.

• https://evilmartians.com/chronicles/custom-cops-for-rubocop-an-emergency-service-for-your-codebase
• https://docs.rubocop.org/rubocop/development.html

I may be able to put something together for this pretty soon if all sounds good.

[jneen]

Yes, generally speaking - or at least not huge. Often we see things like this:

HUGE_ARRAY = [...]
rule %r/#{HUGE_ARRAY.join('|')}\b/, ...

especially for lexers ported from Pygments. The suggested refactor is:

HUGE_SET = Set.new [...]
rule %r/\w+/ do |m|
  if HUGE_SET.include?(m[0])
    # ...
  else
    # ...
  end
end

[jneen]

(and, ideally, move the huge set to a lazy { ... } loaded file - I suppose we could also have a limit on set/array literal size for lexer files)

[larouxn]

Have opened a PR to add a no join pipe in regexp cop here. Lemme know what you think.

• Add custom NoBuildingAlternationPatternInRegexp RuboCop cop #2224