Bug #14744: Refinements modules have a superclass

• Status: Closed
• Assignee: shugo (Shugo Maeda)
• Target version: -
• ruby -v: ruby 2.5.1p57 (2018-03-29 revision 63029) [x86_64-linux]
• Backport: 2.3: UNKNOWN, 2.4: UNKNOWN, 2.5: UNKNOWN
• Tags: refine

[Eregon (Benoit Daloze)]

$ ruby -e 'module M; refine Array do; p self; p self.class; p ancestors; end; end'
#<refinement:Array@M>
Module
[#<refinement:Array@M>, Array, Enumerable, Object, Kernel, BasicObject]

So the refinement module (self in the refine Array do block) is a Module, but looking at its ancestors it has Array as a "superclass".

Is that expected?
I thought modules can never have a superclass in Ruby.

[shevegen (Robert A. Heiler)]

I thought modules can never have a superclass in Ruby.

This may be a bug, I think; other modules do not seem to have
anything but themselves if you call .ancestors() on them either.

I am not absolutely certain, but perhaps either way it could be
mentioned in the documentation too, whether modules (including
refinements) can lead to modules having superclasses.

[shugo (Shugo Maeda)]

• Assignee set to matz (Yukihiro Matsumoto)

So the refinement module (self in the refine Array do block) is a Module, but looking at its ancestors it has Array as a "superclass".

Is that expected?

The superclass of a refinement module is for implementing super in refined methods.
However, it may be better to hidden from reflection APIs.

What do you think, Matz?

[Eregon (Benoit Daloze)]

shugo (Shugo Maeda) wrote:

The superclass of a refinement module is for implementing super in refined methods.
However, it may be better to hidden from reflection APIs.

I see, thank you for the explanation.

I'm not sure it's good to hide though, as I noticed this interesting case by seeing code like:

module R
  refine Array do
    p instance_method(:sum)
  end
end

I think either super should be implemented differently to not rely on setting a Module's superclass,
or we leave it as it is, and have it visible with reflection APIs since anyway it behaves like there is a "superclass".

FWIW currently in TruffleRuby we do a slightly different lookup when looking for the super method of a refined method, and look at the active refinements to find the super method,
and then use an inline cache to avoid repeated lookups.

[hsbt (Hiroshi SHIBATA)]

• Tags set to refine

[matz (Yukihiro Matsumoto)]

I am so sorry for being late.
I agree with @shugo's opinion in #note-2. We should hide Array in the example.

Matz.

[shugo (Shugo Maeda)]

• Status changed from Open to Assigned
• Assignee changed from matz (Yukihiro Matsumoto) to shugo (Shugo Maeda)

[Eregon (Benoit Daloze)]

@matz (Yukihiro Matsumoto) Should instance_method and other methods magically find the method on the refined class, even though self is a Module though, like in https://bugs.ruby-lang.org/issues/14744#note-3 ?

[shugo (Shugo Maeda)]

• Status changed from Assigned to Closed

Applied in changeset [ruby-core:86949] [Bug #14744] R..." href="/projects/ruby-master/repository/git/revisions/754adbee91c2d4a4e84e9271724ca33f630d1916">git|754adbee91c2d4a4e84e9271724ca33f630d1916.

---

Module#ancestors should not return superclasses of refinements

[ruby-core:86949] [Bug #14744]

Reported by Eregon (Benoit Daloze). Thanks!