Table column with maked data drops mask

[EDIT]: See https://github.com/astropy/astropy/issues/8816#issuecomment-500598908 for a more direct example of the bug.

---

import numpy as np
from astropy.table import Table
data = np.ma.masked_greater(np.arange(5), 3)
tab = Table({'data': [data]})
mtab = Table({'data': [data]}, masked=True)
data, tab['data'][0], mtab['data'][0]

gives:

(masked_array(data=[0, 1, 2, 3, --],
              mask=[False, False, False, False,  True],
        fill_value=999999),
 array([0, 1, 2, 3, 4]),
 masked_BaseColumn(data=[0, 1, 2, 3, 4],
                   mask=[False, False, False, False, False],
             fill_value=999999))

So neither the table, nor the masked table respects the mask of the input data and silently converts the array to use the “data behind the mask”.

CC:@taldcroft

[hamogu]

@pllim Don’t worry. I’ve been programming with @taldcroft for almost one decade now. I’m just joking. I know he takes me serious.

[pllim]

For now, workaround seems to be:

>>> mtab['data'][0].mask = data.mask
>>> mtab['data'][0]
masked_BaseColumn(data=[0, 1, 2, 3, --],
                  mask=[False, False, False, False,  True],
            fill_value=999999)

FWIW, it works as expected without the square brackets:

>>> mtab = Table({'data': data}, masked=True)
>>> mtab 
<Table masked=True length=5>
 data
int64
-----
    0
    1
    2
    3
   --