bug#18494: 24.4.50; defclass creates undocumented *-list-p function

[Eric Ludlam <ericludlam@gmail.com>]

  On 09/20/2014 06:41 AM, David Engster wrote:
> [add CC Eric]
>
> Tom Tromey writes:
>> Try:
>>
>> (defclass zzz-whatever () ((hi)))
>>
>> Now notice there is a new function:
>>
>> (symbol-function 'zzz-whatever-list-p)
>> (lambda (obj) "Test OBJ to see if it a list of objects which are a child of type zzz-whatever" (when (listp obj) (let ((ans t)) (while (and obj ans) (setq ans (and (eieio-object-p (car obj)) (object-of-class-p (car obj) zzz-whatever))) (setq obj (cdr obj))) ans)))
>>
>>
>> This new -list-p function is not documented in the EIEIO manual.
>>
>> It seems to me that it would be a bit better not to define it.
>> It doesn't seem generally useful enough, to me.
> I agree it's not generally useful. It was added as a helper function to
> validate a slot which should hold a list of objects (as part of the
> EIEIO/EDE security fix for Emacs 23.4).
>
> At least that's what I remember. I was under the impression that
> eieio-persistent-validate/fix-slot-value would use this, but I cannot
> see it; maybe it's just hidden very well. If it is not used, I'd agree
> that it should better be removed or at least renamed to X--list-p.

Hi,

I found that many of my programs that use EIEIO have slots that are of 
:type that is an object of some class, which is why there are base 
predicates.  Thus, if I have some object class foo, I can set to the 
:type of some other class to foo, instead of some other more complex 
type.  Having another predicate that also handles the list of objects 
that include subclasses is just handy.

David is right about the time when it was introduced, where only slots 
with a constrained :type for an object were allowed to be loaded in, 
allowing the loader to never 'eval anything, only read.

Technically, I could hand write a few of those predicates for the EDE 
loader, but I thought it would be generally useful, especially if other 
folk want to save their objects.  I'm not sure how widely used it is 
outside of EDE, semanticdb, and cogre which all have save files.  I 
certainly think it should be documented if consensus is that it stays.  
Perhaps an extra option for the defclass to add extra predicates along 
with :method-invocation-order.

Eric