Fix call-seq deduplicate for name overlap in aliases

[tompng]

Backend fix of #1335
Fix call-seq deduplicate of Hash#value? and Hash#has_value?.
When alias names have overlap, deduplicate didn't work correctly.

[tompng]

Main change
/^\w*\.?#{ignore_union}[$\(\s]/
↓
/^(\w*\.)?#{ignore_union}([\(\s]|$)/

^\w*\.?value?[$\(\s] matches to has_value?, but we only want to ignore value? and receiver.value?
Fix is: \w*\.? → (\w*\.)?

The latter part [$\(\s], $ seems like a mistake of $ (end-of-line), not "$" (dollar character)

[tompng]

Unfortunately, with this change, receiver[arg] won't be handled here.
So I added bracket_method remove logic in line 375-379.

[kou]

The $\(\s] pattern was added by #897 .

The latter part [$\(\s], $ seems like a mistake of $ (end-of-line), not "$" (dollar character)

I think that the assumption is correct.

[matzbot]

🚀 Preview deployment available at: https://d1d6844d.rdoc-6cd.pages.dev (commit: 35d3352)

[kou]

+1

It may be better that we extract method names from "call_seq" and compare method names instead of creating (complex) regular expressions to match raw "call_seq"...

[kou]

Suggested change

	entry =~ /\s#{ignore_union}\s/
	/^(\w*\.)?#{ignore_union}([\(\s]|$)/.match?(entry) or
	\s#{ignore_union}\s/.match?(entry)

BTW:

• Can we use \A/\z not ^/$ here?
• Can we remove needless \ escape from [\(]?
• How about using (?:...) not (...) not referred groups?

[tompng]

done!

[kou]

The $\(\s] pattern was added by #897 .

The latter part [$\(\s], $ seems like a mistake of $ (end-of-line), not "$" (dollar character)

I think that the assumption is correct.

[tompng]

These docs will be improved

Source and description will be available:
https://docs.ruby-lang.org/en/master/Array.html#method-i-find_index
https://docs.ruby-lang.org/en/master/Date.html#method-i-asctime
https://docs.ruby-lang.org/en/master/Date.html#method-i-mday
https://docs.ruby-lang.org/en/master/Hash.html#method-i-has_value-3F
https://docs.ruby-lang.org/en/master/Time.html#method-i-mday

Wrong call-seq (alias is wrongly used) will be fixed:
https://docs.ruby-lang.org/en/master/Symbol.html#method-i-next
https://docs.ruby-lang.org/en/master/JSON/Ext/Generator/State.html#method-i-escape_slash
https://docs.ruby-lang.org/en/master/JSON/Ext/Generator/State.html#method-i-script_safe-3F
https://docs.ruby-lang.org/en/master/JSON/Ext/Generator/State.html#method-i-strict-3F
https://docs.ruby-lang.org/en/master/JSON/Ext/Generator/State.html#method-i-escape_slash
https://docs.ruby-lang.org/en/master/JSON/Ext/Generator/State.html#method-i-to_hash

[tompng]

It may be better that we extract method names from "call_seq"

I agree. I think call-seq's name part and params part extraction needs refactor.
But I'd like to fix this change in regexp way for now.

AnyMethod#name already have a name extraction from call-seq, but the logic seems incomplete. It doesn't handle method like foo!, foo?.

def name
    return @name if @name

    @name =
      @call_seq[/^.*?\.(\w+)/, 1] ||
      @call_seq[/^.*?(\w+)/, 1] ||
      @call_seq if @call_seq
  end

Maybe we can have a call-seq common format parser and use it in deduplicate_call_seq, name, param_list and param_seq but it will be a relatively large change.