BUG: df.shift(axis=1) complains about index out of range

[alan-wong]

I just tried to pass axis=1 on a shift call to a dataframe and it threw an error, I couldn't find an issue here and this is the first time I tried this function so not sure if this was something that used to work and then got broken.

I'm using python 64-bit 3.3.4, pandas 0.14.1 and numpy 1.8.1

In [57]:

df = pd.DataFrame({'a':randn(10), 'b':randn(10)})
df.shift(axis=1)
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-57-160107eee7e4> in <module>()
      1 df = pd.DataFrame({'a':randn(10), 'b':randn(10)})
----> 2 df.shift(axis=1)

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\generic.py in shift(self, periods, freq, axis, **kwds)
   3173             block_axis = self._get_block_manager_axis(axis)
   3174             indexer = com._shift_indexer(len(self), periods)
-> 3175             new_data = self._data.shift(indexer, periods, axis=block_axis)
   3176         else:
   3177             return self.tshift(periods, freq, **kwds)

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in shift(self, *args, **kwargs)
   2405 
   2406     def shift(self, *args, **kwargs):
-> 2407         return self.apply('shift', *args, **kwargs)
   2408 
   2409     def fillna(self, *args, **kwargs):

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in apply(self, f, *args, **kwargs)
   2373 
   2374             else:
-> 2375                 applied = getattr(blk, f)(*args, **kwargs)
   2376 
   2377             if isinstance(applied, list):

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in shift(self, indexer, periods, axis)
    916         """ shift the block by periods, possibly upcast """
    917 
--> 918         new_values = self.values.take(indexer, axis=axis)
    919         # convert integer to float if necessary. need to do a lot more than
    920         # that, handle boolean etc also

IndexError: index 2 is out of bounds for size 2
In [57]:

df = pd.DataFrame({'a':randn(10), 'b':randn(10)})
df.shift(axis=1)
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-57-160107eee7e4> in <module>()
      1 df = pd.DataFrame({'a':randn(10), 'b':randn(10)})
----> 2 df.shift(axis=1)

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\generic.py in shift(self, periods, freq, axis, **kwds)
   3173             block_axis = self._get_block_manager_axis(axis)
   3174             indexer = com._shift_indexer(len(self), periods)
-> 3175             new_data = self._data.shift(indexer, periods, axis=block_axis)
   3176         else:
   3177             return self.tshift(periods, freq, **kwds)

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in shift(self, *args, **kwargs)
   2405 
   2406     def shift(self, *args, **kwargs):
-> 2407         return self.apply('shift', *args, **kwargs)
   2408 
   2409     def fillna(self, *args, **kwargs):

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in apply(self, f, *args, **kwargs)
   2373 
   2374             else:
-> 2375                 applied = getattr(blk, f)(*args, **kwargs)
   2376 
   2377             if isinstance(applied, list):

C:\WinPython-64bit-3.3.5.0\python-3.3.5.amd64\lib\site-packages\pandas\core\internals.py in shift(self, indexer, periods, axis)
    916         """ shift the block by periods, possibly upcast """
    917 
--> 918         new_values = self.values.take(indexer, axis=axis)
    919         # convert integer to float if necessary. need to do a lot more than
    920         # that, handle boolean etc also

IndexError: index 2 is out of bounds for size 2

[jreback]

this works in 0.14.1 (on 3.3,numpy 1.8, on windows). Looking at the code that you are showing in the error, that looks like some prior code (this was changed in 0.14.0), see #6672. Pls verify that you are actually importing 0.14.1.

[alan-wong]

Argh.. so sorry Jeff, I'm actually running 0.13.1, I thought my laptop was running the same version as my desktop which is running 0.14.1. My apologies, thanks for the quick response, how do I close this?

[jreback]

@alan-wong np...always appreciate the reports