RE : Re: Re: Function memoization

Colin Paul Adams wrote:

> >>>>> "Dimitre" == Dimitre Novatchev <[email protected]> writes:

>     >>  What do you mean by the ID?

>     Dimitre> I realize that this is implementation-dependent, but it
>     Dimitre> is practical not to pass long sequences by
>     Dimitre> value. Instead, they can be passed by reference.  The
>     Dimitre> reference is what I was referring to as the "ID" of the
>     Dimitre> sequence.

  When I saw that, I understood by "ID" the "identity".  Not
an "identity number" or something else too technical, but
the concept of "identity".  Without saying how to implement
that, the argument is considered equal if it is "the same".
But...:

> This is no good. That means two functions calls that pass
> the value 3 will not (usually) be memoized.

  Maybe it would make sense to say:

    - if the argument is an atomic item (or a sequence of 1
      such item, as these are indistinguishable), compare by
      value (string() or data()?);

    - if the argument is a node, compare with the "is"
      operator;

    - else (so if it is a sequence of at least 2 items),
      compare by identity (the two sequences must be the
      same one).  I think such a "sequence identity" doesn't
      exist in the CRs.

> Now if long strings are being passed as arguments, then
> indeed the cost of recognizinf them may be less than the
> cost of running a small function.

  This is the /raison d'être/ of this extension.  The
implementation doesn't have to guess itself where to use
memoization, the developer will tell it that.

  Does it make sense?

  Regards,

--drkm






























	

	
		
___________________________________________________________________________ 
Yahoo! Mail réinvente le mail ! Découvrez le nouveau Yahoo! Mail et son interface révolutionnaire.
http://fr.mail.yahoo.com