[ruby-talk:6722] Re: Why is WeakRef#__getobj__ missing from the p

Aleksi Niemel ([e-mail]) wrote:
# Cullen J O'Neill proposed:
#
# > My question is, isn't it a better idea to just get a real reference
# > to the object using __getobj__?
# > 
# > ----
# > if ref.weakref_alive?
# >   real_ref = ref.__getobj__
#
# I don't know much about weak references, but I'd guess you risk losing the
# real object with above code. I feel I would not rely on the fact there
# happens no GC between alive? and __getobj__.

Yeah, I was aware of the race condition, just ignored it for simplicity.

# Nevertheless I guess some combination of your's and book's versions would be
# ideal, while pretty wordy. I don't like the idea GC is disabled, at least
# for longer periods.

That's the problem I had.

# Maybe there's demand for 
#
#   weakref.if_alive do |ref|
#     # stuff
#   end
#
# And hide the ugly semantics of GC disabling, and allow immediate GC
# enabling.

This seems useful.

But I think there is also a need for for

   ref = WeakRef.makeNotWeakAnymore()

ie: I was caching it, and then I made it discardable via WeakRef since
it was LRU but now it got used again and is MRU so I'd like to pull it 
back in to being a real reference so it won't get GC'ed.  There needs
to be some way of doing this.

It's obviously quite simple to do:

class WeakRef
  def getRef
    GC.disable
    obj = weakref_alive? ? __getobj__ : nil
    GC.enable
    obj
  end

  def if_alive
    if block_given?
      obj = getRef
      yield(obj)
    end
  end
end
  

Cullen J O'Neill
--
[e-mail]