BUG: TypeError: '<' not supported between instances of 'float' and 'str' when using `.combine_first` with categorical index containing nan

[ssche]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Code Sample, a copy-pastable example

In [79]: x = ['b', 'b', 'c', 'a', 'b', np.nan]
    ...: y = ['a', 'b', 'c', 'a', 'b', 'd']
    ...: mi1 = pd.MultiIndex.from_arrays(
    ...:     [x, [1, 2, 3, 4, 5, 6]],
    ...:     names=['a', 'b']
    ...: )
    ...: df = pd.DataFrame({'c': [1, 1, 1, 1, 1, 1]}, index=mi1)
    ...: mi2 = pd.MultiIndex.from_arrays(
    ...:     [y, [1, 1, 1, 1, 1, 1]],
    ...:     names=['a', 'b']
    ...: )
    ...: s = pd.Series([1, 2, 3, 4, 5, 6], index=mi2)
    ...: df.combine_first(pd.DataFrame({'some_col': s}))
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/algorithms.py in safe_sort(values, codes, na_sentinel, assume_unique, verify)
   2060         try:
-> 2061             sorter = values.argsort()
   2062             ordered = values.take(sorter)

TypeError: '<' not supported between instances of 'float' and 'str'

During handling of the above exception, another exception occurred:

TypeError                                 Traceback (most recent call last)
<ipython-input-79-de018ddfae29> in <module>
     11 )
     12 s = pd.Series([1, 2, 3, 4, 5, 6], index=mi2)
---> 13 df.combine_first(pd.DataFrame({'some_col': s}))

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/frame.py in combine_first(self, other)
   6239             return expressions.where(mask, y_values, x_values)
   6240 
-> 6241         return self.combine(other, combiner, overwrite=False)
   6242 
   6243     def update(

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/frame.py in combine(self, other, func, fill_value, overwrite)
   6104         other_idxlen = len(other.index)  # save for compare
   6105 
-> 6106         this, other = self.align(other, copy=False)
   6107         new_index = this.index
   6108 

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/frame.py in align(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis, broadcast_axis)
   3955         broadcast_axis=None,
   3956     ) -> "DataFrame":
-> 3957         return super().align(
   3958             other,
   3959             join=join,

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/generic.py in align(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis, broadcast_axis)
   8542             axis = self._get_axis_number(axis)
   8543         if isinstance(other, ABCDataFrame):
-> 8544             return self._align_frame(
   8545                 other,
   8546                 join=join,

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/generic.py in _align_frame(self, other, join, axis, level, copy, fill_value, method, limit, fill_axis)
   8589         if axis is None or axis == 0:
   8590             if not self.index.equals(other.index):
-> 8591                 join_index, ilidx, iridx = self.index.join(
   8592                     other.index, how=join, level=level, return_indexers=True
   8593                 )

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/indexes/base.py in join(self, other, how, level, return_indexers, sort)
   3491                 )
   3492             else:
-> 3493                 return self._join_non_unique(
   3494                     other, how=how, return_indexers=return_indexers
   3495                 )

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/indexes/base.py in _join_non_unique(self, other, how, return_indexers)
   3618         rvalues = other._get_engine_target()
   3619 
-> 3620         left_idx, right_idx = _get_join_indexers(
   3621             [lvalues], [rvalues], how=how, sort=True
   3622         )

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/reshape/merge.py in _get_join_indexers(left_keys, right_keys, sort, how, **kwargs)
   1326         for n in range(len(left_keys))
   1327     )
-> 1328     zipped = zip(*mapped)
   1329     llab, rlab, shape = [list(x) for x in zipped]
   1330 

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/reshape/merge.py in <genexpr>(.0)
   1323     # get left & right join labels and num. of levels at each location
   1324     mapped = (
-> 1325         _factorize_keys(left_keys[n], right_keys[n], sort=sort, how=how)
   1326         for n in range(len(left_keys))
   1327     )

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/reshape/merge.py in _factorize_keys(lk, rk, sort, how)
   1978     if sort:
   1979         uniques = rizer.uniques.to_array()
-> 1980         llab, rlab = _sort_labels(uniques, llab, rlab)
   1981 
   1982     # NA group

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/reshape/merge.py in _sort_labels(uniques, left, right)
   2003     labels = np.concatenate([left, right])
   2004 
-> 2005     _, new_labels = algos.safe_sort(uniques, labels, na_sentinel=-1)
   2006     new_labels = ensure_int64(new_labels)
   2007     new_left, new_right = new_labels[:llength], new_labels[llength:]

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/algorithms.py in safe_sort(values, codes, na_sentinel, assume_unique, verify)
   2063         except TypeError:
   2064             # try this anyway
-> 2065             ordered = sort_mixed(values)
   2066 
   2067     # codes:

~/envs/pandas-test/lib/python3.8/site-packages/pandas/core/algorithms.py in sort_mixed(values)
   2046         # order ints before strings, safe in py3
   2047         str_pos = np.array([isinstance(x, str) for x in values], dtype=bool)
-> 2048         nums = np.sort(values[~str_pos])
   2049         strs = np.sort(values[str_pos])
   2050         return np.concatenate([nums, np.asarray(strs, dtype=object)])

<__array_function__ internals> in sort(*args, **kwargs)

~/envs/pandas-test/lib/python3.8/site-packages/numpy/core/fromnumeric.py in sort(a, axis, kind, order)
    989     else:
    990         a = asanyarray(a).copy(order="K")
--> 991     a.sort(axis=axis, kind=kind, order=order)
    992     return a
    993 

TypeError: '<' not supported between instances of 'float' and 'str'

Problem description

I use df.combine_first(...) to add a column to a dataframe while extending the index in case an index value does not exist in the target dataframe. However, if the MultIindex of the dataframe/series contains mixed np.nan/str values in their index value, the above TypeError is raised. I originally noticed this for categorical types (x and y), but the problem can be simplified to plain string types.

The reproduction of this error also seems to depend on the length/order of x and y (I tried to reduce it to fewer elements and kept the np.nan, but that didn't reproduce the error).

Expected Output

No exception and the dataframe containing the new column of series.

Output of pd.show_versions()

In [84]: pd.show_versions()

INSTALLED VERSIONS

commit : 2a7d332
python : 3.8.5.final.0
python-bits : 64
OS : Darwin
OS-release : 19.6.0
Version : Darwin Kernel Version 19.6.0: Thu Jun 18 20:49:00 PDT 2020; root:xnu-6153.141.1~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_AU.UTF-8
LOCALE : en_AU.UTF-8

pandas : 1.1.2
numpy : 1.19.2
pytz : 2020.1
dateutil : 2.8.1
pip : 20.1.1
setuptools : 46.4.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 7.18.1
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[phofl]

Hi, thanks for your report.

x = ['b', np.nan]
y = ['a', 'a']
mi1 = pd.MultiIndex.from_arrays(
        [x, [1, 4]],
             names = ['a', 'b']
    )
df = pd.DataFrame({'c': [1, 1]}, index=mi1)
mi2 = pd.MultiIndex.from_arrays(
        [y, [1, 1]],
             names = ['a', 'b']
    )
s = pd.Series([1, 4], index=mi2)
df.combine_first(pd.DataFrame({'some_col': s}))

This raises too, while changing y to something like ["a", "b"] raises not. Seems to be related to a non-unique y.

[phofl]

The error is in get_join_indexers, which means that

df.merge(pd.DataFrame({'some_col': s}), left_index=True, right_index=True)

raises the same error.

[jreback]

the join should raise because of incompatible types

[phofl]

@jreback

Substituting y with something like ["a", "b"] does not raise in case of merge. As long as y has no duplicates everything is fine.