Andreas Eder wrote:
> Ray wrote:
>
>
>> What lisp did you use? Declaring memq inline for cmucl and sbcl (and
>> maybe others?) would have been just as effective and way less
>> work.
>>
>
> I use sbcl, clisp, gcl and openmcl (and soon I will also have
> access to cmucl).
>
Did you profile with these lisps? If so, it would be very interesting
to see what the bottlenecks are for the various lisps.
> Well, an inline declaration might or might not do what we want;
> the standard says that an implementations is not obliged to inline
> anything.
>
Make it a macro then. If speed were the goal, then every caller would
win, not just the ones you've manually changed. And a lot less work too.
> And besides, it wasn't much work and I am still convinced that
> there are also other reasons to remove memq and use member (see
> what Robert Dodier wrote in another mail).
> I really don't get it why some people are so attached to memq.
>
It's not the attachment to memq. That's just one particular example.
What about delq? assq? What's next? Was it worth it? Could the time
have been spent doing something else?
I fail to see how changing memq to member will get more people to become
developers. If you commented the code and the algorithms, and made them
understandable, that will do far more to get developers. People will be
able to understand and improve them. For me, it's all just a big black
box, but memq wasn't one of those black boxes.
Ray
> Andreas
>
>
>