Double-dispatch pattern

I do not have a link at hand. Try google. It is used in Smalltalk to 
define automatic coercion for number. I looked into Squeak but the code 
on Integer, Float is a bit more complex and does not show the essence 
of the pattern.

Look in my slides at: 
http://www.iam.unibe.ch/~ducasse/Web/Lectures/Lectures-OOP-0203/

Stef


On Saturday, June 14, 2003, at 08:25 PM, Joel Shellman wrote:

> Has someone put in something to squeak to allow multiple dispatch? I 
> thought
> I heard this was not supported in Smalltalk in general, but is there
> something in Squeak to do it?
>
> Thanks!
>
> -joel
>
>
>
> ----- Original Message -----
> From: "Brian T Rice" <water at tunes.org>
> To: "The general-purpose Squeak developers list"
> <squeak-dev at lists.squeakfoundation.org>
> Sent: Monday, June 09, 2003 2:04 AM
> Subject: Re: Object: Identity vs. Environment
>
>
>> I'm glad that someone eventually mentioned something related to 
>> multiple
>> dispatch. Slate's is: seems to work rather well for this case. ;)
>>
>> You can also override is: for ad hoc pretending on a pair of 
>> arguments,
>> and the original method is reallyIs: and is not overridden.
>>
>> On Mon, 9 Jun 2003, Joel Shellman wrote:
>>
>>> Ah, I think what you're referring to is double dispatch, isn't it?
>>>
>>> http://c2.com/cgi/wiki?DoubleDispatch
>>>
>>> Which so far as I understand it makes good sense.
>>>
>>> -joel
>>>
>>> ----- Original Message -----
>>> From: "Andres Valloud" <sqrmax at comcast.net>
>>> To: "The general-purpose Squeak developers list"
>>> <squeak-dev at lists.squeakfoundation.org>
>>> Sent: Monday, June 09, 2003 12:04 AM
>>> Subject: Re: Object: Identity vs. Environment
>>>
>>>
>>>>> But you might have a dozen methods with isFoo + ifTrue/ifFalse.
>>>>> Then you would need Object>>fooAction1, Object>>fooAction2, etc.
>>>>
>>>> The dozen actions will get written one way or the other, either 
>>>> inside
>>>> ifTrue/ifFalse blocks or in messages.  As far as the amount of code 
>>>> is
>>>> concerned, fooActions are at least as compact - if not more since 
>>>> you
>>>> don't need ifTrue/ifFalse.
>>>>
>>>> The key issue here is that the ifTrue/ifFalse actions should be in
>>>> messages the receivers of isXYZ understand.  Otherwise, the context 
>>>> in
>>>> which you write the ifTrue/ifFalse actions has to grow larger to
> provide
>>>> meaning to the actions.  You end up as if you had used a traditional
>>>> programming language: the receivers of isXYZ are more value holders
> than
>>>> behaviorful distinctions as far as the context of usage is 
>>>> concerned.
>>>>
>>>> Disclaimer: these arguments apply only to the "isKindOf:" kind of
> isXYZ
>>>> messages.
>>>>
>>>> Andres.
>>
>> -- 
>> Brian T. Rice
>> LOGOS Research and Development
>> http://tunes.org/~water/
>>
>
>