np.array loses dtype after integer division

[Konstantin Lopukhin]

CPython:

>>> import numpy as np
>>> np.__version__
'1.9.0'
>>> np.array(np.zeros(2), dtype=np.float32) / 2
array([ 0.,  0.], dtype=float32)

PyPy 2.6.0:

>>>> np.array(np.zeros(2), dtype=np.float32) / 2
array([ 0.,  0.])
>>>> _.dtype
dtype('float64')
>>>> 

This does not happen with float division.

[Ronan Lamy]

This has already been fixed in PyPy's default branch. With any recent nightly, you should get:

>>>> import numpy as np
>>>> np.array(np.zeros(2), dtype=np.float32) / 2
array([ 0.,  0.], dtype=float32)

[Konstantin Lopukhin]

Great, thank you!