Naming inconsistency in import of maybe_convert_indices in pandas.core.index.py

[djgagne]

In lines 930-931 of pandas.core.index.py, the following code states

from pandas.core.indexing import _maybe_convert_indices
return _maybe_convert_indices(indexer, len(self))

In pandas.core.indexing.py, the function is called maybe_convert_indices. In the 3 other places where the function is called, it is called using the correct name. The leading underscore should be removed from the function name in index.py.

[jreback]

I think code is unreachable actually.

Want to put a halt there and see if its every hit in the test suite?

if not can manufacture a test to see what gets's u there and fix.

[djgagne]

I got to that location by passing an array of integers and a list of integers as a column index.

[jreback]

oh ok

sounds like we need s test case and fix then!

[jreback]

can you post the example which hits this code?

[djgagne]

>>> import pandas as pd
>>> import numpy as np
>>> d = pd.DataFrame(np.random.random((10, 5)), columns=["a"] + np.arange(20,24).tolist())
>>> print(d[[23,26]])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/djgagne/anaconda/lib/python2.7/site-packages/pandas/core/frame.py", line 1791, in __getitem__
    return self._getitem_array(key)
  File "/Users/djgagne/anaconda/lib/python2.7/site-packages/pandas/core/frame.py", line 1835, in _getitem_array
    indexer = self.ix._convert_to_indexer(key, axis=1)
  File "/Users/djgagne/anaconda/lib/python2.7/site-packages/pandas/core/indexing.py", line 1082, in _convert_to_indexer
    indexer = labels._convert_list_indexer(objarr, kind=self.name)
  File "/Users/djgagne/anaconda/lib/python2.7/site-packages/pandas/core/index.py", line 930, in _convert_list_indexer
    from pandas.core.indexing import _maybe_convert_indices
ImportError: cannot import name _maybe_convert_indices

To trigger this section of code, one has to create a DataFrame with a column index that contains a mix of strings and integers as the labels. Then when accessing the columns, one has to pass a list of integers that contain at least one value outside the range of indices or values. Then the code is hit. It is admittedly a rare path to take, and having mixed integers and strings as column labels is probably not the best idea, but it can be done. A KeyError should at least be thrown for 26, which does happen when the column labels are all integers.

[jreback]

i c. ok, want to add that as a test and see if it fixes (the import fix).

[christopher-llop]

Just hit this error today. For anyone who gets here via google and is newer to Python - you can fix this in your code by just updating your column index to contain only strings.

You can print your columns, then copy/paste and update whichever need to be changed

print df.columns
df.columns = ['1', '2', '3', 'A', 'B', 'C']

[jreback]

@djgagne @christopher-llop want to do a pull-request to fix?

[christopher-llop]

@jreback would love to but dealing with finals. If this is still an issue in a couple weeks I'll circle back - been looking for a project for the week between semesters. @djgagne if you have time, by all means go ahead

[djgagne]

I should be able to make a pull-request to fix the problem in the next week.

[jreback]

@djgagne or @christopher-llop pull-request?

[djgagne]

I submitted a pull request here: #10872

[jreback]

closed by #10872