itertools.count(-3)

[bearophileHUGS]

itertools.count docs say:
Does not currently support python long integers.
Note, count() does not check for overflow and will return negative
numbers after exceeding sys.maxint. This behavior may change in the
future.

But it seems it doesn't support negative numbers too:

from itertools import count, islice
list(islice(count(), 0, 5)) [0, 1, 2, 3, 4]
list(islice(count(10), 0, 5)) [10, 11, 12, 13, 14]
list(islice(count(-3), 0, 5)) [4294967293L, 4294967294L, 4294967295L, 0, 1]
def count2(n=0):

.... while True:
.... yield n
.... n += 1
....[-3, -2, -1, 0, 1]

If this isn't a bug, then maybe docs can tell about this too.

Bye,
bearophile

 

[Carl Banks]

itertools.count docs say:
Does not currently support python long integers.
Note, count() does not check for overflow and will return negative
numbers after exceeding sys.maxint. This behavior may change in the
future.

But it seems it doesn't support negative numbers too:

from itertools import count, islice
list(islice(count(), 0, 5)) [0, 1, 2, 3, 4]
list(islice(count(10), 0, 5)) [10, 11, 12, 13, 14]
list(islice(count(-3), 0, 5)) [4294967293L, 4294967294L, 4294967295L, 0, 1]
def count2(n=0):

... while True:
... yield n
... n += 1
...[-3, -2, -1, 0, 1]

If this isn't a bug, then maybe docs can tell about this too.

Seems like a regression to me. itertools was documented as taking a
sequence of integers, not necessarily positive integers. It worked on
Python 2.4.

Looking at the source (from 2.5 rc2), it looks like they accidentally
used PyInt_FromSize_t rather than PyInt_FromSSize_t in the count
iterator. size_t is unsigned, of course, hence the large negative
numbers.


Carl Banks

 

[Jack Diederich]

itertools.count docs say:
Does not currently support python long integers.
Note, count() does not check for overflow and will return negative
numbers after exceeding sys.maxint. This behavior may change in the
future.

But it seems it doesn't support negative numbers too:

from itertools import count, islice
list(islice(count(), 0, 5)) [0, 1, 2, 3, 4]
list(islice(count(10), 0, 5)) [10, 11, 12, 13, 14]
list(islice(count(-3), 0, 5))

[4294967293L, 4294967294L, 4294967295L, 0, 1]

def count2(n=0):

... while True:
... yield n
... n += 1
...

list(islice(count2(-3), 0, 5))

[-3, -2, -1, 0, 1]

If this isn't a bug, then maybe docs can tell about this too.

Seems like a regression to me. itertools was documented as taking a
sequence of integers, not necessarily positive integers. It worked on
Python 2.4.

Looking at the source (from 2.5 rc2), it looks like they accidentally
used PyInt_FromSize_t rather than PyInt_FromSSize_t in the count
iterator. size_t is unsigned, of course, hence the large negative
numbers.

Nuts, that was me. I'll fix and add some tests.

-Jack