Closed
Description
If we have a structured/record/masked array with variable types:
In [1]: import numpy as np, numpy.lib.recfunctions as nrf
In [2]: me = nrf.merge_arrays((np.array([1, 2]), np.array([10., 20., 30.])), usemask=True)
In [3]: me
Out[3]:
masked_array(data = [(1, 10.0) (2, 20.0) (--, 30.0)],
mask = [(False, False) (False, False) (True, False)],
fill_value = (999999, 1e+20),
dtype = [('f0', '<i8'), ('f1', '<f8')])we cannot run reduce based functions on it:
In [4]: np.any(me)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
...
TypeError: cannot perform reduce with flexible typeOk, this is fine. But mask has a lot more predictable content despite the data type or the type of the array, yet we still get the same error:
In [5]: me.mask
Out[5]:
array([(False, False), (False, False), (True, False)],
dtype=[('f0', '?'), ('f1', '?')])
In [6]: np.any(me.mask)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
...
TypeError: cannot perform reduce with flexible typeCould numpy be changed to allow it to perform reduce on the masks of these arrays? Here's the above as a copypasteable snippet:
import numpy as np, numpy.lib.recfunctions as nrf
me = nrf.merge_arrays((np.array([1, 2]), np.array([10., 20., 30.])), usemask=True)
np.any(me) # should not work
np.any(me.mask) # maybe possibly could work?Metadata
Metadata
Assignees
Labels
No labels