.loc with list of incomplete labels misbehaves or raises

[toobaz]

Code Sample, a copy-pastable example if possible

In [2]: df = pd.DataFrame(-1, index=pd.MultiIndex.from_product([[1,2], ['a', 'b'], ['i', 'ii']]), columns=['A', 'B'])

In [3]: df.loc[[1]] # works
Out[3]: 
        A  B
1 a i  -1 -1
    ii -1 -1
  b i  -1 -1
    ii -1 -1

In [4]: df.loc[[(1, 'a')]] # misbehaves (acts as if labels were missing)
Out[4]: 
      A   B
1 a NaN NaN

In [5]: df.loc[[(1,)]] # raises
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in _getitem_iterable(self, key, axis)
   1100                 try:
-> 1101                     return self.obj.reindex_axis(keyarr, axis=axis)
   1102                 except AttributeError:

/home/pietro/nobackup/repo/pandas/pandas/core/frame.py in reindex_axis(self, labels, axis, method, level, copy, limit, fill_value)
   2836                                         method=method, level=level, copy=copy,
-> 2837                                         limit=limit, fill_value=fill_value)
   2838 

/home/pietro/nobackup/repo/pandas/pandas/core/generic.py in reindex_axis(self, labels, axis, method, level, copy, limit, fill_value)
   2509         new_index, indexer = axis_values.reindex(labels, method, level,
-> 2510                                                  limit=limit)
   2511         return self._reindex_with_indexers({axis: [new_index, indexer]},

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/multi.py in reindex(self, target, method, level, limit, tolerance)
   1845                                                limit=limit,
-> 1846                                                tolerance=tolerance)
   1847                 else:

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/multi.py in get_indexer(self, target, method, limit, tolerance)
   1791             # don't have the same dtypes
-> 1792             if self._inferred_type_levels != target._inferred_type_levels:
   1793                 return Index(self.values).get_indexer(target.values)

AttributeError: 'Int64Index' object has no attribute '_inferred_type_levels'

During handling of the above exception, another exception occurred:

AttributeError                            Traceback (most recent call last)
<ipython-input-5-e5c902934a28> in <module>()
----> 1 df.loc[[(1,)]]

/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in __getitem__(self, key)
   1326         else:
   1327             key = com._apply_if_callable(key, self.obj)
-> 1328             return self._getitem_axis(key, axis=0)
   1329 
   1330     def _is_scalar_access(self, key):

/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in _getitem_axis(self, key, axis)
   1539                     raise ValueError('Cannot index with multidimensional key')
   1540 
-> 1541                 return self._getitem_iterable(key, axis=axis)
   1542 
   1543             # nested tuple slicing

/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in _getitem_iterable(self, key, axis)
   1105                     if axis != 0:
   1106                         raise AssertionError('axis must be 0')
-> 1107                     return self.obj.reindex(keyarr)
   1108 
   1109             # existing labels are non-unique

/home/pietro/nobackup/repo/pandas/pandas/core/frame.py in reindex(self, index, columns, **kwargs)
   2827     def reindex(self, index=None, columns=None, **kwargs):
   2828         return super(DataFrame, self).reindex(index=index, columns=columns,
-> 2829                                               **kwargs)
   2830 
   2831     @Appender(_shared_docs['reindex_axis'] % _shared_doc_kwargs)

/home/pietro/nobackup/repo/pandas/pandas/core/generic.py in reindex(self, *args, **kwargs)
   2421         # perform the reindex on the axes
   2422         return self._reindex_axes(axes, level, limit, tolerance, method,
-> 2423                                   fill_value, copy).__finalize__(self)
   2424 
   2425     def _reindex_axes(self, axes, level, limit, tolerance, method, fill_value,

/home/pietro/nobackup/repo/pandas/pandas/core/frame.py in _reindex_axes(self, axes, level, limit, tolerance, method, fill_value, copy)
   2773         if index is not None:
   2774             frame = frame._reindex_index(index, method, copy, level,
-> 2775                                          fill_value, limit, tolerance)
   2776 
   2777         return frame

/home/pietro/nobackup/repo/pandas/pandas/core/frame.py in _reindex_index(self, new_index, method, copy, level, fill_value, limit, tolerance)
   2781         new_index, indexer = self.index.reindex(new_index, method=method,
   2782                                                 level=level, limit=limit,
-> 2783                                                 tolerance=tolerance)
   2784         return self._reindex_with_indexers({0: [new_index, indexer]},
   2785                                            copy=copy, fill_value=fill_value,

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/multi.py in reindex(self, target, method, level, limit, tolerance)
   1844                     indexer = self.get_indexer(target, method=method,
   1845                                                limit=limit,
-> 1846                                                tolerance=tolerance)
   1847                 else:
   1848                     raise Exception("cannot handle a non-unique multi-index!")

/home/pietro/nobackup/repo/pandas/pandas/core/indexes/multi.py in get_indexer(self, target, method, limit, tolerance)
   1790             # we may not compare equally because of hashing if we
   1791             # don't have the same dtypes
-> 1792             if self._inferred_type_levels != target._inferred_type_levels:
   1793                 return Index(self.values).get_indexer(target.values)
   1794 

AttributeError: 'Int64Index' object has no attribute '_inferred_type_levels'

Problem description

This was already discussed here and here, opening a separate issue for clarity.

Expected Output

df.loc[[(1, 'a',)]] should return the same as df.loc[1, 'a', :] (or more explicitly, df.loc[(1, 'a', slice(None)), :]), while df.loc[[(1,)]] should return the same as df.loc[[1]].

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.3.final.0
python-bits: 64
OS: Linux
OS-release: 4.7.0-1-amd64
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: it_IT.utf8
LOCALE: it_IT.UTF-8

pandas: 0.19.0+834.g8c7b9731f
pytest: 3.0.6
pip: 9.0.1
setuptools: 33.1.1
Cython: 0.25.2
numpy: 1.12.0
scipy: 0.18.1
xarray: 0.9.1
IPython: 5.1.0.dev
sphinx: 1.4.9
patsy: 0.3.0-dev
dateutil: 2.5.3
pytz: 2016.7
blosc: None
bottleneck: 1.2.0
tables: 3.3.0
numexpr: 2.6.1
feather: 0.3.1
matplotlib: 2.0.0
openpyxl: 2.3.0
xlrd: 1.0.0
xlwt: 1.1.2
xlsxwriter: 0.9.6
lxml: 3.7.1
bs4: 4.5.3
html5lib: 0.999999999
sqlalchemy: 1.0.15
pymysql: None
psycopg2: None
jinja2: 2.8
s3fs: None
pandas_gbq: None
pandas_datareader: 0.2.1

[jreback]

df.loc[1, 'a', :] is not valid syntax

[toobaz]

df.loc[1, 'a', :] is not valid syntax

I used it because it works, but you're right that df.loc[(1, 'a', slice(None)), :] is clearer (updated decription)

[toobaz]

df.loc[1, 'a', :] is not valid syntax

By the way: I tend to think it is valid syntax, since the index has three levels - as valid as df.loc[1, 'a', 'i']', which is (officially, I guess) supported as a shortcut for df.loc[(1, 'a', 'i'), :]'.

Indeed, I would expect df.loc[1, 'a', 'i', :] to raise (and it does, but with a different error from what it should because of #14885).

[toobaz]

As noticed by @bfollinprm in #22151 , the behavior of In [4]: changed in 0.23.0, and an error is now raised.