Re: Non-overridable and non-redefinable methods

On 8/20/05, Eric Mahurin <[e-mail]> wrote:
> --- Austin Ziegler <[e-mail]> wrote:
>> On 8/19/05, Eric Mahurin <[e-mail]> wrote:
>>> Fixnum#<most methods>
>>> String#<most methods>
>>> Object#<most methods>
>> I disagree on all three. In no way do I believe that such overriding
>> is problematic, and I would never trade this flexbility for an
>> attempt at speed. Some methods, perhaps, but only those that fit the
>> /\b__[\w\d]+__\b/ pattern and are defined in the core. That's a lot
>> fewer than you're suggesting.
> You pulled this list out of context.

I pulled the list, but I believe that I did not misrepresent as I didn't
say anything about inheritance.

> Do you seriously believe that it really makes sense to redefine
> something like Fixnum#+? When you do that, you are only affecting the
> ruby code using Fixnum's, because in typical C code they have already
> bypassed the method call altogether and effectively inlined the
> method.  You'll end up with something strangely inconsistent.

The question is whether that should be programmatically prevented -- and
if so, for what reason. It's important to note that Fixnum#+ may call
#coerce in some cases, so even a C method could end up calling a
non-optimized method.

-austin
-- 
Austin Ziegler * [e-mail]
               * Alternate: [e-mail]