[Pytables-users] Numpy views stored as attributes

[Ask F. J. <as...@li...>]

Hey all,

When I store a view of a numpy array as an attribute it appears to be stored as the array that owns the data. Is this a bug? I find it confusing that the user has to check if the numpy array owns the data or always remember to do a copy() before storing a numpy array as an attribute.

Below is some sample code that highlights the problem.

Best regards, Ask

import numpy as np
import tables

with tables.openFile("test.h5", "w") as f:

    x=f.createArray("/", "test", [0])

    A=np.array([[0,1],[2,3]])

    x.attrs['a']=A
    x.attrs['b']=A.T.copy()
    x.attrs['c']=A.T

    assert np.all(x.attrs['a']==A)
    assert np.all(x.attrs['b']==A.T)
    assert np.all(x.attrs['c']==A)
    assert np.all(x.attrs['c']==A.T) # AssertionError!