print df.ix[..,..] changes dataframe?

[jankatins]

These two codeblocks result in different dataframes:

import pandas as pd
import numpy as np
cont = ['one', 'two','three', 'four', 'five', 'six', 'seven']
dfb = pd.DataFrame({'a' : cont, "b":cont[3:]+cont[:3] ,'c' : np.arange(7)})
#print dfb.ix[0,"c"]
dfb.ix[7,'c'] = 1
print dfb
print dfb.ix[7,"c"]
print dfb.ix[0,"c"]

       a      b  c
0    one   four  0
1    two   five  1
2  three    six  2
3   four  seven  3
4   five    one  4
5    six    two  5
6  seven  three  6
7    NaN    NaN  1
1.0
0.0

and

import pandas as pd
import numpy as np
cont = ['one', 'two','three', 'four', 'five', 'six', 'seven']
dfb = pd.DataFrame({'a' : cont, "b":cont[3:]+cont[:3] ,'c' : np.arange(7)})
print dfb.ix[0,"c"]
dfb.ix[7,'c'] = 1
print dfb
print dfb.ix[7,"c"]
print dfb.ix[0,"c"]

0
       a      b   c
0    one   four   0
1    two   five   1
2  three    six   2
3   four  seven   3
4   five    one   4
5    six    two   5
6  seven  three   6
7    NaN    NaN NaN
nan
0.0

pd.version = '0.12.0-922-gac1609e'

In the first case I also don't understand why the last column is changed to a float when I set an int.

[jtratner]

I don't have time to look at this today, but I'm guessing it has something to do with ix maybe caching something on the Series. The change to a float is a nice pointer, because it suggests that the column was astyped to float when it gained a missing value and never changed back. Or @jreback will come in and fix it in 30 seconds with his hands tied behind his back blind-folded. :)

[jreback]

was about 10 minutes.....:) (didn't start for a while)

[jreback]

The astyping is in theory possible but I don't do it for efficiency because otherwise you have a do a semi-expensive check anytime you do pretty much anything (its in there, just not the float -> int conversions).