Feature #7019: allow `private` and `protected` keywords to take blocks

• Status: Rejected
• Assignee: matz (Yukihiro Matsumoto)
• Target version: 3.0

[alexeymuranov (Alexey Muranov)]

I like to indent my private methods one level deeper, but this indentation is inconsistent with the syntax:

class C
  def f
    #
  end
  private
    def g
    #
    end
end

I think it would be nice if the private keyword could take a block, then i would write:

class C
  def f
    #
  end
  private do
    def g
    #
    end
  end
end

[trans (Thomas Sawyer)]

I've often wondered about being able to use:

private def f
  #
end

I think it's true that the public/private/protected "modifiers" have always felt a bit awkward --which accounts for the variety of indention and comment "highlights" that have been used with them.

[shyouhei (Shyouhei Urabe)]

I know you like blocks but this proposal doesn't make sense to me because public/private/protected are declarations, not something done. So it's odd to say "private DO something". I'd rather +1 to private def, which is natural (to me at least).

[alexeymuranov (Alexey Muranov)]

What about

privately do
end

? :)
(I understand that a good reason is needed to add a new keyword).

[alexeymuranov (Alexey Muranov)]

Here is a usecase i have made up. The following does not work now:

class C
  def f
    def g
      #
    end
    private
    def h
      #
    end
  end
end

c = C.new
c.f  # => NameError: undefined local variable or method `private' for ...

So a block syntax could make sense here:

class C
  def f
    def g
      #
    end
    privately do
      def h
        #
      end
    end
  end
end

[drbrain (Eric Hodel)]

I don't know why you would expect this to work, private, public, etc. are methods on Module and your context is not a subclass of Module.

Why should we pollute every object with a new methods privately, publicly and protectedly (which isn't a word)?

[headius (Charles Nutter)]

+1 for "private def foo". I have never seen a good justification for why visibility is better as a value on the current frame, nor for why it's better to set a method's visibility after declaration rather than along with the declaration.

The fact that visibility lives on the frame also means implementations that attempt to optimize frames away (like JRuby does and eventually MRI will want to do) have a harder time of it.

[luislavena (Luis Lavena)]

• Category set to core
• Target version set to 2.0.0

headius (Charles Nutter) wrote:

+1 for "private def foo". I have never seen a good justification for why visibility is better as a value on the current frame, nor for why it's better to set a method's visibility after declaration rather than along with the declaration.

So two changes:

• def will require to return the method defined instead of nil
• private will require to accept a Method and not just a Symbol

I believe these two changes where already requested (can't find the bug reports yet).

This approach helps also to describe methods beyond documentation markers, making it clear when a method is private inline with the method definition and not after the method has been defined.

I like this.

[trans (Thomas Sawyer)]

Would need to accept array of those too (whether its a method or symbol) for things like:

private attr_accessor :foo

[alexeymuranov (Alexey Muranov)]

Excuse me Thomas, what's the point in a "private attribute accessor"?

[trans (Thomas Sawyer)]

Probably not much, but it is possible. Could be "protected" instead as well. Also, other helpers can be created besides attr. My point is only that it needs to accept an array should the helper return multiple methods/symbols.

Or are you thinking that private def would be some kind of keyword thing?

[ko1 (Koichi Sasada)]

• Assignee set to mame (Yusuke Endoh)

mame-san, could you judge this ticket?

IMO, it is too slow to introduce such a big feature.

[mame (Yusuke Endoh)]

• Status changed from Open to Assigned
• Assignee changed from mame (Yusuke Endoh) to matz (Yukihiro Matsumoto)
• Target version changed from 2.0.0 to 3.0

IMO, it is too slow to introduce such a big feature.

Completely agreed.

--
Yusuke Endoh [e-mail]

[alexeymuranov (Alexey Muranov)]

Please do not forget also module_function and such. (Or are they different in some sense?).

Alexey.

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

[kou (Kouhei Sutou)]

• Status changed from Assigned to Rejected

#3753 adds private def ... support. If you still want this feature, please reopen this issue.

[shyouhei (Shyouhei Urabe)]

• Has duplicate Feature #16276: For consideration: "private do...end" / "protected do...end" added