A tiny followup for #1641’s oopsies

[ParadoxV5]

• chomp!: The ? suffix in (?string? separator) is already covered by (nil).
• scrub: Just like scrub!, scrub(nil, &nil) is a thing.

---

Question: What shall the convension around ?Regexp | string | nil vs. ?Regexp | string? be?

[ParadoxV5]

re-ping: #1641

[soutaro]

This case is necessary.
Passing a node of type String? will be a type error without the ?.

[ParadoxV5]

This is a problem with the type checkers unable to break string and nil from string? to different paths and then merge them back to self? | nil i.e. self?

[soutaro]

It may be true. Type checkers can do more about this.
But at least for now, RBS is for static type checking and having string? is more common for static type checking (I believe.)

[ParadoxV5]

But at least for now

How about this: I will revert it for now, but we should also remind type checker maintainers about these sort of cases.

[ParadoxV5]

soutaro/steep#333

[soutaro]

Revert the type of chomp!. 🙏

[ParadoxV5]

No comments? Then how did L648 pass CI

rbs/core/string.rbs

Line 648 in a4681af

def self.try_convert: (String object) -> String # technically will return `object` unchanged.

[soutaro]

@ParadoxV5 You uncovered another bug. 🐞

https://github.com/ruby/rbs/blob/master/lib/rbs/writer.rb#L224 writes lines, from the beginning of end to the end of the method type of the last overload. The comment here is after the end of the method type of the last overload.

Having a comment line between the two overloads would pass CI.

  def chomp!: (nil) -> nil
            # | (?string separator) -> self? # https://github.com/ruby/rbs/pull/1672#discussion_r1423324796
            | (?string? separator) -> self?

[soutaro]

👍