Re: To CLOS or not to CLOS?

On 2007-02-20, at 15:31 EST, [email protected] wrote:

> - Are generic functions really the right or only way to go?  It  
> seems many
> of CLOS's advantages could still be had in a object-dispatched object
> system designed with a similar mentality, and perhaps there are  
> advantages
> to object-dispatch.

The only advantage I can think of is that it takes fewer cpu cycles.   
But as I have said in the past:  cpu cycles we have plenty of, it is  
good programmer cycles we need to optimize.  Craig Chambers and  
Jonathan Bachrach both have a lot to say about a) how to optimize  
generic dispatch and b) why some languages chose not to implement  
generic dispatch.

Generic functions are the evolution of object (or message) dispatch  
to solve the problem of which object a method belongs to when it  
operates on more than one object.  If you artificially attach it to  
one of its argument objects, you have to simulate generic dispatch on  
the remaining arguments.  (Some will argue that this is a rare enough  
occurrence that you don't need a special language mechanism to handle  
it.  But by the same token, who needs bignums or garbage collection?)

One thing that CLOS got wrong, which was improved in Dylan, is how to  
choose the most applicable method of a generic function.  The Dylan  
improvement was the fruit of research started by Ducournau, Habib, et  
al.  See http://en.wikipedia.org/wiki/C3_linearization for more.