Resizing memmap arrays

I’m referring to this discussion on stackoverflow. I’m trying to resize memmap arrays just as it works with ordinary arrays.

a = np.memmap('bla.bin', mode='w+', dtype=int, shape=10)
>>> memmap([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])
a.resize(20, refcheck=False)
>>> ValueError: cannot resize this array: it does not own its data

I’m not sure about implications and side effects, but for me it looks like, if mmap supports resizing, np.memmap also should. I understand, that it might be hard to guarantee private access for a memmap. But if access can be excluded by the programmer, it should at least be possible to override any check by hand. And if resizing really shouldn’t be possible, at least the documentation could be clear about this point.

[ml31415]

Any news about that stuff? It’s a while ago, but the issue remains, I guess.

[maciejkula]

I have the same problem. If this is not going to be addressed, perhaps the docs for http://docs.scipy.org/doc/numpy/reference/generated/numpy.memmap.resize.html should be amended? At the moment they strongly imply resizing memmaped arrays is something numpy supports in a very straightforward way.