itertools comments [Was: Re: RELEASED: Python 2.3a2]

[Alex Martelli]

Alexander Schmolck wrote:

> Alex Martelli <aleax at aleax.it> writes:
>> 
>> > I think xrange, etc. are already a convention for the closest thing
>> > Python has traditionally had to iterators.  Maybe those functions
>> > should just be updated to make ordinary iterators.
>> 
>> Not without breaking, e.g., xrange(23, 17923, 12)[15] -- which now
>> has a perfectly good result of 203 and would break if xrange
>> returned an iterator (or any other non-indexable object).
> 
> It could always return and indexable iterator.

Yes, it's not forbidden for an iterator object to be also
indexable -- weird (but then some would say xrange objects
already ARE weird, I guess) but not forbidden.  Similarly,
len(xrange(7)) would have to keep returning 7, etc.

So what would be the added value of forcing:

x = xrange(n)
assert x is iter(x)

to be true?  It would only seem to break such usage as:

N_times = xrange(N)

for i in N_times:
    for j in N_times:
        process(i, j)


What advantage would you think breaking this usage would
confer?  Or, if you agree that breaking this usage is a
DIS-advantage (albeit not a huge one), what would in your
opinion be the compensating advantages?  I just cannot see
any -- but it is, of course, quite possible that I'm just
having a blind spot, so, please explain, thanks!


Alex