'sort_index"does not work with 'pd.set_option('mode.use_inf_as_na',True)'

[asmcat666]

In [12]: pd.set_option('mode.use_inf_as_na',False)

In [13]: df
Out[13]:
                            B
2018-09-25 14:47:31.500   0.0
2018-04-03 22:55:13.000   1.0
2018-09-04 09:10:07.500   2.0
2018-11-30 10:32:08.000   NaN
2018-03-27 14:21:12.500   4.0
2018-09-28 14:30:36.500   6.0
2018-07-16 21:56:48.000   7.0
2018-12-05 23:53:59.500   8.0
2018-09-04 09:46:58.500   9.0
2018-11-09 09:14:08.500  10.0

In [14]: df.sort_index()
Out[14]:
                            B
2018-03-27 14:21:12.500   4.0
2018-04-03 22:55:13.000   1.0
2018-07-16 21:56:48.000   7.0
2018-09-04 09:10:07.500   2.0
2018-09-04 09:46:58.500   9.0
2018-09-25 14:47:31.500   0.0
2018-09-28 14:30:36.500   6.0
2018-11-09 09:14:08.500  10.0
2018-11-30 10:32:08.000   NaN
2018-12-05 23:53:59.500   8.0

In [15]: pd.set_option('mode.use_inf_as_na',True)

In [16]: df.sort_index()
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-16-02a473b4132d> in <module>
----> 1 df.sort_index()

g:\py372\lib\site-packages\pandas\core\frame.py in sort_index(self, axis, lev
 ascending, inplace, kind, na_position, sort_remaining, by)
   5088
   5089             indexer = nargsort(
-> 5090                 labels, kind=kind, ascending=ascending, na_position=n
osition
   5091             )
   5092

g:\py372\lib\site-packages\pandas\core\sorting.py in nargsort(items, kind, as
ding, na_position)
    244
    245     items = extract_array(items)
--> 246     mask = np.asarray(isna(items))
    247
    248     if is_extension_array_dtype(items):

g:\py372\lib\site-packages\pandas\core\dtypes\missing.py in isna(obj)
    120     Name: 1, dtype: bool
    121     """
--> 122     return _isna(obj)
    123
    124

g:\py372\lib\site-packages\pandas\core\dtypes\missing.py in _isna_old(obj)
    177         return _isna_ndarraylike_old(obj)
    178     elif isinstance(obj, ABCGeneric):
--> 179         return obj._constructor(obj._data.isna(func=_isna_old))
    180     elif isinstance(obj, list):
    181         return _isna_ndarraylike_old(np.asarray(obj, dtype=object))

AttributeError: 'DatetimeArray' object has no attribute '_constructor'

[TomAugspurger]

Can you provide a reproducible example, including the code to create the dataframe?: http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports

[simonjayhawkins]

closing as can't reproduce

[annhak]

Bumped into the same problem @simonjayhawkins
Minimal reproducible example:

import pandas as pd

d = {'col1': [1, 2, 3], 'col2': [3, 4, 5]}
df = pd.DataFrame(data=d)
now = pd.Timestamp.now()
df.index = [now, now - pd.Timedelta('1m'), now + pd.Timedelta('2m')]

try:
df1 = df.sort_index()
print('test1, success')
except Exception as ex:
print('test 1 failed: ', ex)

pd.set_option('use_inf_as_na', True)

try:
df2 = df.sort_index()
print('test2, success')
except Exception as ex:
print('test 2 failed: ', ex)

Produces:
test1, success
test 2 failed: 'DatetimeArray' object has no attribute '_constructor'

[simonjayhawkins]

Thanks @annhak . You code sample raises on 0.25.3, but appears to be working on 1.0.5 and master.

will reopen for now with needs tests label but will also add closing candidate if it can be determined which commit fixed this and whether appropriate tests were added.

[ankushduacodes]

Hi @simonjayhawkins, I would love to contribute to this issue, Please let me know if I may :)

[weikhor]

Hi Guys I will start doing this.

[weikhor]

@TomAugspurger @simonjayhawkins @annhak @asmcat666
Guy have created test cases for this issues.