numpy.linspace broken with Quantities

[mhvk]

Numpy's linspace used to work nicely with quantities:

Numpy 1.9.3

>>> np.linspace(-1 * u.m, 1 * u.m, 10)
<Quantity [-1.        ,-0.77777778,-0.55555556,-0.33333333,-0.11111111,
            0.11111111, 0.33333333, 0.55555556, 0.77777778, 1.        ] m>

Numpy 1.10

>>> np.linspace(-1 * u.m, 1 * u.m, 10)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/tom/miniconda3/envs/np110/lib/python2.7/site-packages/numpy/core/function_base.py", line 115, in linspace
    y[-1] = stop
ValueError: setting an array element with a sequence.

@mhvk - is this a bug in Astropy, or a bug introduced in Numpy?

A lot of user code might rely on this, so we should get a workaround out soon unless it has to be fixed in Numpy.

[mhvk]

It is a numpy regression: see numpy/numpy#7142

[eteq]

@mhvk - is this now fixed post- numpy/numpy#7156 ? If so, how should we proceed on this issue - should we mention this as a "known issue" in numpy 1.10.x<1.10.5 ? (do you know how far back it goes?)

[mhvk]

@eteq - yes, it would be a known issue in numpy 1.10, up to 1.10.5. If anywhere, it would probably be part of http://docs.astropy.org/en/latest/known_issues.html#quantity-issues

[mhvk]

@eteq - I went ahead and made this into a PR with a small documentation update. Only thing I'm not quite sure about is whether I set the labels right...

[astrofrog]

@mhvk - I would rephrase to say 'linspace does not work correctly with quantities when using Numpy 1.10.0 to 1.10.5 due to a bug in Numpy. The solution is to upgrade to Numpy 1.10.6, in which the bug was fixed'

[mhvk]

@astrofrog - agreed, might as well be clear about where the problem was. Now changed the text.

[mhvk]

I'm going ahead and merge this small documentation update -- there's a bit of a cost to leaving things like this open, if only that they keep popping up any time I look at my open issues/PRs.

[eteq]

Backported to v1.0.x in 542f8c3 and v1.1.x in 12cbe2f