Re: [Sbcl-devel] Re: Callbacks, details, the devil

[Brian M. <br...@ma...>]

On Jun 27, 2005, at 6:35 PM, William Harold Newman wrote:

> In my experience this DEFMETHODish syntax approach is strictly more of
> a headache than the DEFSTRUCTish
>   macro DEFINE-ALIEN-CALLBACK (name &key calling-convention) return- 
> type
>          typed-lambda-list &body forms
> with shorthand
>   macro DEFINE-ALIEN-CALLBACK name return-type
>          typed-lambda-list &body forms
> The DEFSTRUCTish way is not especially pretty, but I find it usually
> gets the job done. However, my experience does not include the
> possibility that NAME might want to be a list (generalized alien
> function names! What could be more necessary?) and probably other
> accidents as well, so take it as a suggestion, rather than a confident
> judgement.

If you do have a generalized alien function name :-) you can say  
(define-alien-callback ((setf foo)) ...) - simply using the "long  
form" is enough disambiguation here.

I think I'd like to see the defstructish form with all options in the  
&key part of the destructuring of the first argument; this will make  
define-alien-callback forms easier to parse, easier to extend and  
easier to define as a compatibility macro on some other implementation.
--
Brian Mastenbrook
br...@ma...
http://www.iscblog.info/