From: "Hal E. Fulton" <hal9000@...>
Date: 2002-09-12T01:26:42+09:00
Subject: Re: Multimethods (was Re: Larry Wall's comments on Ruby)

----- Original Message -----
From: "Dan Sugalski" <dan@sidhe.org>
To: "ruby-talk ML" <ruby-talk@ruby-lang.org>
Sent: Wednesday, September 11, 2002 9:56 AM
Subject: Multimethods (was Re: Larry Wall's comments on Ruby)


> At 10:48 PM +0900 9/11/02, Gavin Sinclair wrote:
> >Despite the discussion and links, I haven't picked up on what
multimethods
> >actually are, so I can't comment on those.

(snip informative stuff}

> That's pretty much multimethods. They're terribly useful, as they cut
> out a lot of checks for the second type in methods. (Checks which
> tend to be slow, error prone, and missing stuff)

That's it? I thought it would be more esoteric and
interesting.

Do people really consider that a "multiple-receiver"
model??

I had envisioned some kind of bizarre framework
where you could do (obj1,obj2).method in an
order-independent fashion.

In any case, I wouldn't favor adding this to Ruby,
as variables have no types; and thus how do you
define a method "signature"?

What's more, a method with a single parameter can
be passed an object of ANY type. I'd hate to define
a method for each type I expected (as I would in
C++ or Java), especially since that would preclude
types I *didn't* expect.

Example: If a method does nothing to its parameter
but invoke the << operator, I might expect that it
would work on strings and arrays. But it would also
work on files, whether I planned that or not.

Hal