From: Yukihiro Matsumoto <matz@...>
Date: 2010-09-10T00:49:13+09:00
Subject: [ruby-core:32237] Re: Ruby 2.0 Wiki/Wish-list?

Hi,

In message "Re: [ruby-core:32235] Re: Ruby 2.0 Wiki/Wish-list?"
    on Fri, 10 Sep 2010 00:37:45 +0900, Charles Oliver Nutter <headius@headius.com> writes:

|> |Module#prepend (but there is a ticket for that, if I remember correctly):
|>
|> I am positive about prepending a module. �It will be addressed on Ruby
|> 2.0 in whatever form.
|
|I'd like to be part of design and implementation discussions on this,
|if possible. I've prototyped it previously, but never committed that
|work anywhere. I believe it will be implemented mostly like the
|current "include wrapper" for modules, but positioned and dealt with a
|bit differently.

Modulre#prepend will be introduced in 2.0, so I am happy to discuss
with other implementers.  At RubyKaigi keynote, I said it would be
addressed by #mix, but I have realized I was wrong.  #prepend is
better solution for method wrapping.

|In JRuby, there are as many weak references as there are subclasses
|(including module wrappers) and inclusions of modules (modules keep a
|list of all inclusion points). After booting RubyGems, the total
|number of such elements is in the 1200-1300 range. Rails and other
|frameworks would obviously bump this number up a lot. It's not clear
|what impact it has on the JVM, though it's safe to say that it's not
|free and that it's probably cheaper than on MRI.

Maybe it wouldn't be that bad if we implemented "real" weak reference
map for our implementation.  We will keep investigation.

							matz.