special-casing np.array(range(...))

[anntzer]

Compare (on Python3 -- for Python2, read "xrange" instead of "range"):

In [2]: %timeit np.array(range(1000000), np.int64)
10 loops, best of 3: 156 ms per loop

In [3]: %timeit np.arange(1000000, dtype=np.int64)
1000 loops, best of 3: 853 µs per loop

Note that while iterating over a range is not very fast, it is still much better than the array creation:

In [4]: from collections import deque

In [5]: %timeit deque(range(1000000), 1)
10 loops, best of 3: 25.5 ms per loop

On one hand, special cases are awful. On the other hand, the range builtin is probably important enough to deserve a special case. Or not? (Feel free to close this issue if that's not going to happen.)
(Real issue which prompted this suggestion: I was building sparse matrices using scipy.sparse.csc_matrix with some indices specified using range, and that construction step turned out to take a significant portion of the time because of the calls to np.array).

Cross-posted to the mailing list: https://mail.scipy.org/pipermail/numpy-discussion/2016-February/074960.html

[jakirkham]

Seems like the sort of thing that should be put on the mailing list. Did you already put it on there? If so, could we please have a link?

[charris]

Wouldn't it be better to fix sparse matrices or use arange in the first place?

[anntzer]

Sure, this fix can go into scipy.sparse, but it's obviously more generally applicable than just for sparse matrices.

[charris]

But that is what arange is for. What is the problem with arange?

[rgommers]

I think I'd be for not special-casing this. We're also removing the special case for np.sum(generator). Close?

[eric-wieser]

np.sum(generator) was a special and inconsistent case of behavior, whereas this is just a suggested special case for optimization reasons. I don't think the reasoning is transferable here.

Having said that, I don't think this optimization is worthwhile - it seems like optimization efforts would be better invested elsewhere

[anntzer]

I guess I may try my hand at implementing this at some point but feel free to close it for now...

[rgommers]

Thanks @anntzer