On Fri, Mar 19, 2021 at 8:59 PM Dennis Sweeney <sweeney.dennis650@gmail.com> wrote:
I think these don't generally fit with the "feel" of the itertools module though since almost everything there accepts iterables and returns iterators
itertools.product as written barely seems to fit. It might as well accept only sequences, since it immediately converts everything into a sequence anyway. It's also documented as "equivalent to a nested for-loop" which is plainly untrue for several different reasons. A simple improvement would be to keep the first argument as an uncached iterator if repeat=1. Not only would that improve memory usage in many cases, but it would allow for things like product(count(1), ...) which ought to work in my opinion. Another option is to accept iterables that repeatedly get iter() called on
them, but that's hard to make work with generators
That variant of itertools.product (one which is truly equivalent to a nested for loop) would have to be a new function, since it wouldn't be backward compatible. The lack of a convenient way to make it work with generators feels like a hole in the standard library that should be fixed: class ReiterableGenerator: # defined in itertools def __init__(self, f, *args, **kwargs): self._f, self._args, self._kwargs = f, args, kwargs def __iter__(self): return self._f(*self._args, **self._kwargs) primes = ReiterableGenerator(lambda: (n for n in count(2) if isprime(n)))