Bug: Incorrect handling of NULL in results returning astropy Table

[rickyegeland]

For example:

result = db.fetch_data_table('Baliunas1995', ['S_mean', 'activity_class', 'Pcyc', 'Pcyc2'], meta={'Pcyc':'FAPGrade'}, nulls=True

has a dtype:

[('star', 'S12'), ('S_mean', '<f8'), ('activity_class', 'S12'), ('Pcyc', 'O'), ('Pcyc2', 'O'), ('FAPGrade', 'O')]

Which causes all sorts of problems when trying to plot, for example. The ugly workaround is to cast the column post-fetch:

   print result.dtype
   for i, col in enumerate(result.colnames):
        if col in ('Pcyc', 'Pcyc2'):
            dtype[i] = 'float64'
            result[col].fill_value = np.nan
            result[col] = result[col].astype('float64')
        if col in ('FAPGrade'):
            dtype[i] = 'S'
            result[col].mask = [x is None for x in result[col]]
    result = astropy.table.Table(result, dtype=dtype)
    for col in ('Pcyc', 'Pcyc2'):
        result[col].mask = np.isnan(result[col])
    print "Final:", result.dtype

This needs to be handled in fetchall_astropy()