"cannot reindex from duplicate axis" when adding cell of missing row to non-unique index

[toobaz]

Code Sample, a copy-pastable example if possible

In [2]: df = pd.DataFrame([[1,2,5,6], [3,4,7,8]], index=['a', 'a'], columns=pd.MultiIndex.from_product([[1,2], ['A', 'B']]))

In [3]: df.loc['c'] = -1

In [4]: df.loc['c', (1, 'A')] = 3

In [5]: df.loc['d', (1, 'A')] = 3
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-5-34c3deb1645d> in <module>()
----> 1 df.loc['d', (1, 'A')] = 3

/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in __setitem__(self, key, value)
    176             key = com._apply_if_callable(key, self.obj)
    177         indexer = self._get_setitem_indexer(key)
--> 178         self._setitem_with_indexer(indexer, value)
    179 
    180     def _has_valid_type(self, k, axis):

/home/pietro/nobackup/repo/pandas/pandas/core/indexing.py in _setitem_with_indexer(self, indexer, value)
    348                     index = self.obj._get_axis(i)
    349                     labels = index.insert(len(index), key)
--> 350                     self.obj._data = self.obj.reindex_axis(labels, i)._data
    351                     self.obj._maybe_update_cacher(clear=True)
    352                     self.obj.is_copy = None

/home/pietro/nobackup/repo/pandas/pandas/core/frame.py in reindex_axis(self, labels, axis, method, level, copy, limit, fill_value)
   2844                      self).reindex_axis(labels=labels, axis=axis,
   2845                                         method=method, level=level, copy=copy,
-> 2846                                         limit=limit, fill_value=fill_value)
   2847 
   2848     @Appender(_shared_docs['rename'] % _shared_doc_kwargs)

/home/pietro/nobackup/repo/pandas/pandas/core/generic.py in reindex_axis(self, labels, axis, method, level, copy, limit, fill_value)
   2492                                                  limit=limit)
   2493         return self._reindex_with_indexers({axis: [new_index, indexer]},
-> 2494                                            fill_value=fill_value, copy=copy)
   2495 
   2496     def _reindex_with_indexers(self, reindexers, fill_value=np.nan, copy=False,

/home/pietro/nobackup/repo/pandas/pandas/core/generic.py in _reindex_with_indexers(self, reindexers, fill_value, copy, allow_dups)
   2515                                                 fill_value=fill_value,
   2516                                                 allow_dups=allow_dups,
-> 2517                                                 copy=copy)
   2518 
   2519         if copy and new_data is self._data:

/home/pietro/nobackup/repo/pandas/pandas/core/internals.py in reindex_indexer(self, new_axis, indexer, axis, fill_value, allow_dups, copy)
   3879         # some axes don't allow reindexing with dups
   3880         if not allow_dups:
-> 3881             self.axes[axis]._can_reindex(indexer)
   3882 
   3883         if axis >= self.ndim:

/home/pietro/nobackup/repo/pandas/pandas/indexes/base.py in _can_reindex(self, indexer)
   2733         # trying to reindex on an axis with duplicates
   2734         if not self.is_unique and len(indexer):
-> 2735             raise ValueError("cannot reindex from a duplicate axis")
   2736 
   2737     def reindex(self, target, method=None, level=None, limit=None,

ValueError: cannot reindex from a duplicate axis

Problem description

I think the reindexing code path shouldn't be taken at all (in this case - I'm not referring to the general design choice): anyway, the operation should work, coherently with the case where the index is unique and with the case where the row exists (In [4]).

Expected Output

Like In [4].

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+783.gcd35d22a0
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]

yep looks like a bug.

[mroeschke]

Looks to work in master now. Could use a test now

In [15]: In [2]: df = pd.DataFrame([[1,2,5,6], [3,4,7,8]], index=['a', 'a'], columns=pd.MultiIndex.from_product([
    ...: [1,2], ['A', 'B']]))
    ...:
    ...: In [3]: df.loc['c'] = -1
    ...:
    ...: In [4]: df.loc['c', (1, 'A')] = 3
    ...:
    ...: In [5]: df.loc['d', (1, 'A')] = 3

In [16]: df
Out[16]:
     1         2
     A    B    A    B
a  1.0  2.0  5.0  6.0
a  3.0  4.0  7.0  8.0
c  3.0 -1.0 -1.0 -1.0
d  3.0  NaN  NaN  NaN