__repr__ wrong column alignment with non-ascii characters

[manuteleco]

Hi,

it seems that when DataFrame, Series and maybe other objects contain non-ascii characters inside non-unicode strings the __repr__ method is not able to give the correct column alignment to its values. However, we see that this issue does not affect unicode strings. I'm using pandas '0.8.1.dev-70c3deb' in a Linux box.

Sample code:

# -*- coding: utf-8 -*-

from pandas import Series, DataFrame

df1 = DataFrame([["aaaa", 1], ["bbbb", 2]])
df2 = DataFrame([["aaää", 1], ["bbbb", 2]])
df3 = DataFrame([[u"aaää", 1], ["bbbb", 2]])

# Comparison between "similar dataframes"
print df1
print
print df2
print
print df3
print

# Other cases:
s1 = Series(["ä", "bbbb", "ßß"])
print
print s1

This results in:

      0  1
0  aaaa  1
1  bbbb  2

        0  1
0  aaää  1
1    bbbb  2

      0  1
0  aaää  1
1  bbbb  2


0      ä
1    bbbb
2    ßß

Thanks and regards.

[jseabold]

I don't think this should've been closed. The original problem in the given example still exists AFAICT.

>>> from pandas import Series, DataFrame
>>> 
>>> df1 = DataFrame([["aaaa", 1], ["bbbb", 2]])                                    
>>> df2 = DataFrame([["aaää", 1], ["bbbb", 2]])                                    
>>> df3 = DataFrame([[u"aaää", 1], ["bbbb", 2]])                                   
>>>                                                                                
>>> # Comparison between "similar dataframes"                                      
>>> print df1                                                                      
      0  1                                                                         
0  aaaa  1                                                                         
1  bbbb  2                                                                         
>>> print                                                                          

>>> print df2                                                                      
      0    1                                                                       
0  aaää  1                                                                         
1  bbbb    2                                                                       
>>> print                                                                          

>>> print df3                                                                      
      0  1                                                                         
0  aaää  1                                                                         
1  bbbb  2  
>>> pandas.version.version
'0.8.2.dev-c99d9cd'

[jseabold]

Though maybe this is intentional for strings?

[jseabold]

Thinking about this a bit more, I'm thinking this maybe should be fixed, but does that imply always using unicode in to_string? Regardless, force_unicode=True fails for df2. I'll push a fix for this.