Simplify concurrent.futures.process code by using itertools.batched()

[NewUserHa]

Advantages:

• use new standard library, to encourage users to also use stdlibs.

d=list(range(50000))

list(concurrent.futures.process._get_chunks(d, chunksize=1))
# 105 ms ± 4.43 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

list(_get_chunks(d, chunksize=1))
# 5.87 ms ± 378 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

• performance. the difference is significant when chunksize is small, and the chunksize is usually =1.

[AA-Turner]

This seems fine -- there's a check that chunksize is >=1 in ProcessPoolExecutor.map.

A

[serhiy-storchaka]

You can inline the it variable, and the inline _get_chunks() itself. This will remove some overhead for packing/unpacking *iterables.

[NewUserHa]

inline _get_chunks() itself means moving its content back to map()?

[serhiy-storchaka]

Yes, since it is a short one-liner.

[serhiy-storchaka]

LGTM.

[serhiy-storchaka]

Thank you for your contribution, @NewUserHa.

But note that usually all but the most trivial changes require opening an issue, and if it is an optimization, its effect should be demonstrated in benchmarks. Not only that some line in a tight loop becomes faster, but that the stdlib code that include this line becomes faster.

I accepted this change as exception, because it makes the code simpler and adds a demonstration of a new itertools function. But the bar may be higher for other changes.