BUG: Series with non-unique index: "Index length did not match values" error upon assignment

[jgehrcke]

Create a Series with non-unique index:

>>> import pandas as pd
>>> pd.__version__
'0.12.0'
>>> s1 = pd.Series(range(3))
>>> s2 = pd.Series(range(3))
>>> comb = pd.concat([s1,s2])
>>> comb
0    0
1    1
2    2
0    0
1    1
2    2
dtype: int64

Assign value by boolean mask:

>>> comb[comb<1] = 5
>>> comb
0    5
1    1
2    2
0    5
1    1
2    2
dtype: int64

This has worked. Now add a value by boolean mask:

>>> comb[comb<2] += 10
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "***/pandas/core/series.py", line 852, in __setitem__
    self.where(~key,value,inplace=True)
  File "***/pandas/core/series.py", line 749, in where
    other = other.reindex(ser.index)
  File "***/pandas/core/series.py", line 2646, in reindex
    return self._reindex_with_indexers(new_index, indexer, copy=copy, fill_value=fill_value)
  File "***/pandas/core/series.py", line 2650, in _reindex_with_indexers
    return Series(new_values, index=index, name=self.name)
  File "***/pandas/core/series.py", line 492, in __new__
    subarr.index = index
  File "properties.pyx", line 74, in pandas.lib.SeriesIndex.__set__ (pandas/lib.c:29541)
AssertionError: Index length did not match values

Is this expected behavior? If it is, I am sorry, because this was not clear to me from the docs and I am just wondering why simple assignment via = works and special assignment via += does not...

[jreback]

its a bug

in general using non_unique indicies is not a good idea; try using multiindexes: http://pandas.pydata.org/pandas-docs/dev/indexing.html#hierarchical-indexing-multiindex

[jgehrcke]

Thanks for confirming. Side note: I don't really need the index so pd.concat([s1,s2], ignore_index=True) also is a good approach to circumvent the problem.

[jreback]

yep....thanks for the case though, this already fixed in #3482; basically since Series is a sub-class of ndarray things are pretty tricky; in 0.13 via that PR its going to be a sub-class of NDFrame (which is what DataFrame subclasses); then things like this are much easier

[jreback]

closed by #4779