[Python-Dev] Int FutureWarnings and other 2.4 TODOs

[Aahz]

On Fri, Dec 05, 2003, Guido van Rossum wrote:
>Aahz:
>> 
>> Hmmmm...  How important is the "less memory" angle versus the "faster"
>> angle?  Why not just add a few bytes to the long type to store an int?
>> If you're using the int, the pointer is NULL, and you're saving that
>> memory, at least.  The only harm is some extra memory in the type
>> object.
> 
> I'd say that the "less memory" angle is pretty important.  The int
> implementation has been using every trick in the book to save memory
> from very early days on: it has the most sophisticated special-purpose
> allocator, *and* it uses a cache for popular values, *and* you really
> can't squeeze any more bits out of the structure.

Right, that's what's historically true.  Question is, by the time 3.0
comes out, *should* it still be true?  As pointed out downthread, we've
already got a problem with the int free list; earlier discussions on
python-dev have talked about ways of dealing with that (the simplest of
which will be to make range() an iterator, since that's the most common
problem source).  Why not use some of that memory savings to make the
int object easier to handle?  "Simple is better than complex"; I suspect
that we might have a win-win situation.

Perhaps we could push some of the memory-saving techniques into the long
object?
-- 
Aahz (aahz at pythoncraft.com)           <*>         http://www.pythoncraft.com/

Weinberg's Second Law: If builders built buildings the way programmers wrote 
programs, then the first woodpecker that came along would destroy civilization.