bug in repeat() method with MultiIndex - fixed in 0.15

[glyg]

Hi,

I found a bug in version 0.14, which has since been fixed in 0.15. I thought I'd report it because I couldn't find a mention of it and it might have been fixed silently. Also I couldn't find a specific test for this bug (as far as I could see).

Here is a test that fails with the bug:

import pandas as pd
print(pd.version.version)

def test_multiindex_repeat():
    idx = pd.Index([1, 2, 3, 4])
    m_idx = pd.MultiIndex.from_tuples([(1, 2), (3, 4),
                                       (5, 6), (7, 8)])
    data = ['a', 'b', 'c', 'd']

    df = pd.Series(data, index=idx)
    m_df = pd.Series(data, index=m_idx)

    assert df.repeat(3).shape == (3 * len(data),)
    assert m_df.repeat(3).shape == (3 * len(data),)

test_multiindex_repeat()

With pandas 0.14, here is the backtrace:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-11-7fcd65f0e025> in <module>()
----> 1 assert m_df.repeat(3).shape == (3 * len(data),)

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/series.pyc in repeat(self, reps)
    705         new_values = self.values.repeat(reps)
    706         return self._constructor(new_values,
--> 707                                  index=new_index).__finalize__(self)
    708 
    709     def reshape(self, *args, **kwargs):

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/series.pyc in __init__(self, data, index, dtype, name, copy, fastpath)
    229                                        raise_cast_failure=True)
    230 
--> 231                 data = SingleBlockManager(data, index, fastpath=True)
    232 
    233         generic.NDFrame.__init__(self, data, fastpath=True)

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/internals.pyc in __init__(self, block, axis, do_integrity_check, fastpath)
   3028             block = make_block(block,
   3029                                placement=slice(0, len(axis)),
-> 3030                                ndim=1, fastpath=True)
   3031 
   3032         self.blocks = [block]

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/internals.pyc in make_block(values, placement, klass, ndim, dtype, fastpath)
   1835 
   1836     return klass(values, ndim=ndim, fastpath=fastpath,
-> 1837                  placement=placement)
   1838 
   1839 

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/internals.pyc in __init__(self, values, ndim, fastpath, placement)
   1231         super(ObjectBlock, self).__init__(values, ndim=ndim,
   1232                                           fastpath=fastpath,
-> 1233                                           placement=placement)
   1234 
   1235     @property

/home/guillaume/anaconda/envs/pelican/lib/python2.7/site-packages/pandas/core/internals.pyc in __init__(self, values, placement, ndim, fastpath)
     72             raise ValueError('Wrong number of items passed %d,'
     73                              ' placement implies %d' % (
---> 74                                  len(self.values), len(self.mgr_locs)))
     75 
     76     @property

ValueError: Wrong number of items passed 12, placement implies 4

Shall I make a PR adding this test to test_multiindex or is it not worth the (minimal) work?

[jreback]

yeah this was fixed with the index refactor
sure - send a PR to confirm the behavior (if u want to look up the index refactor pr and reference that issues would be gr8)

[glyg]

Test added in #9362 .
I guess you're referring to #7891 which looks kinda big?

[jreback]

yep just put it in the comment in the test as well

[jreback]

closed by #9362