From: murphy <murphy@...>
Date: 2007-12-10T17:17:51+09:00
Subject: Re: redefineable not operator

David Flanagan wrote:
> I saw the following commit the other day:
>  Commit by matz  ::  r14162 /trunk/ (7 files in 2 dirs):
>     * parse.y (expr): redefinable not (!) operator.
>     * parse.y (arg): ditto.
>     * object.c (rb_obj_not): new method "!".
>     * object.c (rb_obj_not_equal): new method "!=".
>     * object.c (rb_obj_not_match): new method "!~".
why such heavy syntax changes, two weaks before the release of Ruby
1.9.1? I mean, I'm happy that changes are always possible here, but I
thought 1.9 was fixed some time ago. it's really confusing for me, since
I don't know any (English) website that's summarizing what's going on
recently...

so there is no 1.9-stable branch? no feature-fixed release candidates
before 1.9.1?

anyway, the change is interesting. I remember having a lot of arguments
for != not being a method when I tried to explain the exception in the
German Ruby forum. I thought it was a design decision with positive
side-effects on performance and simplicity.

the change raises some questions:

- it seems I can no longer rely on !x returning a boolean :(
  how do I make my !!x hacks 1.9-compatible?

- does it affect performance? reading the source, I assume that a != b
  is now slower than a == b, because it does another funcall.

- can someone give an example where it would be useful to redefine !,
  != or !~?

- when I tried it in irb, I got this. I have no idea what's happening.

irb(main):001:0> def !; 42; end
=> nil
irb(main):002:0> public :!
/usr/local/lib/ruby/1.9/irb/input-method.rb:100:in `gets': private
method `!' called for false:FalseClass (NoMethodError)
[ long callchain ]

I hope you make %...() and :: redefinable next ;)
[murphy]