[Openmcl-devel] type specifier '(simple-vector n) in defmethod

[James Anderson]

good morning;

pascal costanza published on the topic of alternative dispatching mechanisms fifteen years ago:

    http://www.p-cos.net/documents/filtered-dispatch.pdf


> On 5. Jan 2024, at 15:46, Tim McNerney <mc at media.mit.edu> wrote:
> 
> Thanks for the support, Phoe.  
> 
> From the technological context of early CL history, I do sort-of "get" some of the thinking behind method dispatch limitations. (Too bad Gregor isn't here to speak for himself). At the time, Common Lisp proponents were deathly afraid of Lisp's reputation as a "slow language," so it is unclear that CL being originally designed in the absence of CLOS is to blame. I suppose it is interesting that, at least in my circles, performance declarations never caught on (e.g. speed 3, safety 0). In my professional work, I have been known to seek out and destroy such monsters in every library we use.
> 
> --Tim
> 
> On 1/5/24 8:44 AM, Michał "phoe" Herda wrote:
>> I'm not arguing that it shouldn't be in CL, I'm arguing that it's not there because no one has done it. I guess you're free to extend CCL with the functionality you're proposing, i.e. by introducing a new method specializer in addition to the current class and EQL specialization.
>> Once a single working implementation is there, it's much easier to convince other implementations to add it and end up turning it into a de-facto standard (like with package-local nicknames).
>> On 5.01.2024 14:41, Tim McNerney wrote:
>>> To which, (at the risk of a civilized flame war), I say, "too bad."
>>> IMHO, this was a short-sighted and poor design decision by the X3J13 committee, et al.
>>> The same debate came up within the greater Java community, 
>>>    and it is one place where Java and C# diverged.
>>> Java's designers made similar choices to Common Lisp's: 
>>>    two kinds of datatypes, dispatchable objects and everything else.
>>> C# is arguably "Java done right," where everything is a dispatchable "object."
>>> 
>>> There are clear benefits: it would also result in a cleaner, more appealing, more powerful language.
>>> It's not too late to fix this flaw. What's the harm? 
>>> The "non-object" dispatch code would only be present in generic functions when necessary.
>>> 
>>> --Tim
>>> 
>>> On 1/2/24 6:17 AM, Michał "phoe" Herda wrote:
>>>> No, standard CLOS dispatch happens only on classes.
>>>> Have you tried https://github.com/digikar99/polymorphic-functions ?
>>>> W dniu 2024-01-02 12:14, Taoufik Dachraoui napisał(a):
>>>>> Hi 
>>>>>   I am trying to use '(simple-vector n) as type specifier in defmethod but it is not possible, is there another way?
>>>>> 
>>>>> (typep #(1 2 3 4) '(simple-vector 4))
>>>>> 
>>>>> (defmethod mul ((n fixnum) (m1 (simple-vector n)) (m2 (simple-vector n)))
>>>>>   (let ((r (make-array (list n))))
>>>>>     (dotimes (i n)
>>>>>       (setf (svref r i) (* (svref m1 i) (svref m2 i))))
>>>>>     r))
>>>>>   Kind regards
>>>>> Taoufik Dachraoui 
>>>>>  
>>>> 
>>> 
> 

---
james anderson | james at dydra.com | https://dydra.com