Using loc[] or at[] to add rows doesn't work with an IntervalIndex doesn't work

[konstantinmiller]

It seems that a data frame with an IntervalIndex won't let you use loc[] or at[] to add rows.

I would expect the following to work:

df = pd.DataFrame(index=pd.IntervalIndex(closed='left', dtype='interval[int64]',
                                         data=[pd.Interval(0, 1, closed='left')]),
                  data={'a': [42]})
df.at[pd.Interval(2, 3, closed='left'), 'a'] = 43

However, it throws a confusing error

AttributeError: 'pandas._libs.interval.IntervalTree' object has no attribute 'set_value'

[simonjayhawkins]

@konstantinmiller Thanks for the report.

using loc appears to be appending a row. (on master and 0.25.3)

>>> pd.__version__
'1.1.0rc0+7.g04e9e0afd'
>>>
>>> df = pd.DataFrame(
...     index=pd.IntervalIndex(
...         closed="left", dtype="interval[int64]", data=[pd.Interval(0, 1, closed="left")]
...     ),
...     data={"a": [42]},
... )
>>> df
         a
[0, 1)  42
>>>
>>> df.loc[pd.Interval(2, 3, closed="left"), "a"] = 43
>>> df
           a
[0, 1)  42.0
[2, 3)  43.0
>>>

can you give an example or versions where loc is failing.

[simonjayhawkins]

The at issue is for getting values and not just setting values. This is covered by #27927 so closing as duplicate.