Bug #10967: Is "warning: private attribute?" wrong?

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.3.0dev (2015-03-01 trunk 49796) [x86_64-darwin14]
• Backport: 2.0.0: WONTFIX, 2.1: WONTFIX, 2.2: WONTFIX

[spastorino (Santiago Pastorino)]

The following code ...

class Y
  def initialize
    @x = "ZOMG"
  end

  def print_x
    puts x
  end

  private

  attr_reader :x
end

Y.new.print_x

outputs ...

test.rb:12: warning: private attribute?

I tend to think this warning is wrong, I was surprised by https://github.com/rack/rack/pull/811 and I think this is a completely valid use case.

Also this code ...

class Y
  def initialize
    @x = "ZOMG"
  end

  def print_x
    puts x
  end

  def assign_x
    self.x = "ZOMG ZOMG"
  end

  private

  attr_accessor :x
end

y = Y.new
y.assign_x
y.print_x

Works fine with warnings also. So a private writer works ok when the receiver is self because Ruby has a special case for it, this make me think that private writers were thought to be used.

So ... am I wrong thinking that the warning should be removed or the self special case shouldn't work and be removed from Ruby code?. It doesn't make sense to me to have both things.

[spastorino (Santiago Pastorino)]

One possible fix (the one removing the warning) ... https://github.com/ruby/ruby/pull/849

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

It's valid in your case, but may not in others.
It doesn't sound enough reason to remove that warning to me.

If you want to suppress the warning, you can explicitly do it after the definition.

class Y
  attr_writer :x
  private :x=
end

[marcandre (Marc-Andre Lafortune)]

FWIW, I feel that warning should be removed. There are too many false positives, and I suspect very very few cases where that warning is of any help. Note that without the warning, it will be obvious anyways when calling that reader/writer that it fails because it is private.

[spastorino (Santiago Pastorino)]

There's also a discussion going on in another PR, https://github.com/ruby/ruby/pull/889

@nobu (Nobuyoshi Nakada), not sure what are the invalid use cases you refer to, and also what if I need to use let's say 2 private accessors? should I do ...

  attr_accessor :a, :b

  private :a, :a=, :b, :b=

Doesn't sound great to me.

[matz (Yukihiro Matsumoto)]

We haven't thought of self as a receiver. Agreed to remove warnings.

Matz.

[zzak (zzak _)]

• Status changed from Open to Closed

Applied in changeset r50585.

---

• vm_method.c: Remove private attribute warning [Bug #10967]
  Patch by @spastorino (Santiago Pastorino) [Fixes GH-849]
  https://github.com/ruby/ruby/pull/849

• test/ruby/test_module.rb: Update test for changes

[usa (Usaku NAKAMURA)]

• Backport changed from 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN to 2.0.0: WONTFIX, 2.1: REQUIRED, 2.2: REQUIRED

[usa (Usaku NAKAMURA)]

Is this a spec change or a bug?

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.0.0: WONTFIX, 2.1: REQUIRED, 2.2: REQUIRED to 2.0.0: WONTFIX, 2.1: REQUIRED, 2.2: WONTFIX

Hmm, I don't know it's a bug fix or a spec change, but I think it's a trivial issue. I decided to not to backport to 2.2. Please notice if any objections.

[usa (Usaku NAKAMURA)]

• Backport changed from 2.0.0: WONTFIX, 2.1: REQUIRED, 2.2: WONTFIX to 2.0.0: WONTFIX, 2.1: WONTFIX, 2.2: WONTFIX

Ok, I follow you, chikanaga-san!