[Gcl-devel] Re: subtypep

[Camm Maguire]

Greetings!


Robert Boyer <address@hidden> writes:

> Fixed? Hmmmm?
> 
> Consider these predicates:
> 
>   (defun foo (x) nil)
>   (defun bar (x) t)
> 
> (subtypep '(satisfies foo) '(satisfies bar)) was NIL, T and is now NIL, NIL.
> That is a better answer and an acceptable answer.  Good.  But T, T would be
> even better, even best!
> 
> (subtypep '(satisfies bar) '(satisfies foo)) was NIL, T and that was a better
> answer than the current NIL, NIL, which is ok but merely because it is an
> always acceptable, mealy-mouthed cop-out: "I dunno!"
> 
> Note: I am not asking for any further work here, just thinking out loud.
> Maybe in some future life we'll open up SATISFIES predicates.  Probably a
> giant swamp.

I think swamp is accurate.  I was quite proud that we could determine
that satisfies was some unknown blob and still infer the subtypep
relationship where possible, e.g. `(and cons (satisfies foo)) 'list,
until I see that cmucl and clisp do the same.  As they do on your
examples above.  Maybe we have merely caught up.  Sigh.

Take care,

> 
> Bob
> 
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah