I am using anaconda python 2.7, pandas 0.12.0, with from __future__ import unicode_literals

The bug

When I call dataframe.to_csv(fname, sep=sep, index=index, float_format=float_format, **csvkwargs) and pass in sep=',', pandas throws an exception:

 File "C:\Anaconda\lib\site-packages\pandas\core\frame.py", line 1410, in to_csv formatter.save() File "C:\Anaconda\lib\site-packages\pandas\core\format.py", line 959, in save quoting=self.quoting) File "C:\Anaconda\lib\site-packages\pandas\core\common.py", line 1770, in __init__ self.writer = csv.writer(self.queue, dialect=dialect, **kwds) TypeError: "delimiter" must be an 1-character string 

The workaround

If I change this to:
dataframe.to_csv(fname, sep=str(sep), index=index, float_format=float_format, **csvkwargs) everything works as you might expect. (I have observerved this error both when I pass in no encoding and when I pass in encoding='Windows 1252')

What I would expect

When passed unicode, pandas should:

• If encoding is set, and the unicode sep is one character long i.e. len(sep)==1 (is it really necessary to have this one character restriction?), use the sep character passed in (encoded appropriately)
• If encoding is not set, then check that the default encoding (I don't know whether this is ascii or platform dependent) of sep is only one character long