np.in1d doesn't work on memmaps

[Al Macmillan]

Hi

I have code that intersects an array b with a memmap a via np.in1d. On cpython numpy this behaves as expected but on pypy numpy it only works if the memmap is converted/is viewed as an ndarray.

In [49]: a[i:j][:,4]
Out[49]: 
memmap([   242,    507,    255,    505,    315,    316,    308,    506,
          309,    255,    211,    505,    315,    316,    308,    506,
          309,    255,    255,    711,    194,    232,    711,    711,
          709,    710,    709,    710,    882,    897,    711,    245,
          711,    711,    168,    245], dtype=uint32)

In [50]: b
Out[50]: array([245, 255, 293], dtype=uint32)

In [51]: in1d(a[i:j][:,4], b)
Out[51]: 
array([False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False,
       False, False, False, False, False, False, False, False, False], dtype=bool)

In [52]: c = a[i:j][:,4].view(type=np.ndarray)

In [53]: in1d(c, b)
Out[53]: 
array([False, False,  True, False, False, False, False, False, False,
        True, False, False, False, False, False, False, False,  True,
        True, False, False, False, False, False, False, False, False,
       False, False, False, False,  True, False, False, False,  True], dtype=bool)

[mattip]

I'll try to reproduce, but a complete script that gives samples of a, b would be most helpful

[mattip]

I didn't get too far, so an example script would be the first step. Here is what I tried, following the numpy documentation, which of course failed but differently::

>>>> import numpy as np
>>>> data = np.empty((10,10,36,5), dtype='float32')
>>>> data[5,5][:,4] = [   242,    507,    255,    505,    315,    316,    308,    506,   
               309,    255,    211,    505,    315,    316,    308,    506,
               309,    255,    255,    711,    194,    232,    711,    711,
               709,    710,    709,    710,    882,    897,    711,    245,
               711,    711,    168,    245]
>>>> data[5,5][:,4]
array([ 242.,  507.,  255.,  505.,  315.,  316.,  308.,  506.,  309.,
        255.,  211.,  505.,  315.,  316.,  308.,  506.,  309.,  255.,
        255.,  711.,  194.,  232.,  711.,  711.,  709.,  710.,  709.,
        710.,  882.,  897.,  711.,  245.,  711.,  711.,  168.,  245.], dtype=float32)
>>>> from tempfile import mkdtemp
>>>> import os.path as path
>>>> filename = path.join(mkdtemp(), 'newfile.dat')
>>>> fp = np.memmap(filename, dtype='float32', mode='w+', shape=data.shape)
>>>> fp[:] = data[:]
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
ValueError: assignment destination is read-only

[mattip]

Fixed in c5c4df0df240