Series.replace and DataFrame.replace have same docstring?

[kyleabeauchamp]

In [4]: pd.__version__
Out[4]: u'0.18.1'

The docstring for Series.replace refers to looking up column names, which AFAIK doesn't make sense for the Series version of replace. It seems like the Series function is just blindly inheriting the DataFrame docstring?

http://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.replace.html?highlight=replace#pandas.Series.replace

http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.replace.html?highlight=replace#pandas.DataFrame.replace

For example, the following trivial example of Series.replace() does not agree with the docstring:

import pandas as pd

x = pd.DataFrame([dict(A=1, B=2, C=3), dict(A=2, B=3, C=4), dict(A=3, B=4, C=5)])

print(x.A.replace({1:0}))
print(x.A)

0    0
1    2
2    3
Name: A, dtype: int64
0    1
1    2
2    3
Name: A, dtype: int64

[kyleabeauchamp]

One other minor nit. Should the following edge case be supported? It currently raises a ValueError if you pass an empty dict.

x.A.replace({})

~/miniconda2/lib/python2.7/site-packages/pandas/core/generic.pyc in replace(self, to_replace, value, inplace, limit, regex, method, axis)
   3334 
   3335             items = list(compat.iteritems(to_replace))
-> 3336             keys, values = zip(*items)
   3337 
   3338             are_mappings = [is_dictlike(v) for v in values]

ValueError: need more than 0 values to unpack

I'm not 100% sure this should even be allowed because of the docstring confusion, though.

[sinhrks]

Thanks for the report. Yeah doc improvement is appreciated. It should be re-defined in Series/DateFrame to show correct signature as docstring. You can find an example here.

-https://github.com/pydata/pandas/blob/master/pandas/core/generic.py#L2219

• https://github.com/pydata/pandas/blob/master/pandas/core/series.py#L2315

I think the empty case should be supported to do nothing like other methods.

x.rename({})
#    A  B  C
# 0  1  2  3
# 1  2  3  4
# 2  3  4  5

[RaghavBatra]

Hey, I am interested! How should I go about solving this?

[gfyoung]

@RaghavBatra : I think the comment by @sinhrks above nicely describes what needs to be done.

If you can write a new docstring for Series.replace that properly reflects the behavior of the method, that will be great!

[RaghavBatra]

@gfyoung, I looked at the counter example @kyleabeauchamp gave above and I don't see any problem with it:
{1: 0} implies you replace the 1 with 0, which the code does!

0 1
1 2
2 3
Name: A, dtype: int64
0 0
1 2
2 3
Name: A, dtype: int64

[gfyoung]

@RaghavBatra : Can you confirm that passing in both {1: 0} and {} don't break anything?

[RaghavBatra]

@gfyoung, yes it works for the DataFrame counterexample above.

I also tried the Series.replace function. That worked as well!

ser = pd.Series([1, 2, 3, 4])
ser
0 1
1 2
2 3
3 4
dtype: int64
ser.replace({1:0})
0 0
1 2
2 3
3 4
dtype: int64
ser.replace({})
0 1
1 2
2 3
3 4
dtype: int64

[gfyoung]

@sinhrks @jreback : It seems like the behavior matches the docs now. Closing in light of this development.

[gfyoung]

@RaghavBatra : Thanks for looking into this!

[RaghavBatra]

@gfyoung glad to be of help!
Let me know if there are other beginner level issues to work on!

[gfyoung]

@RaghavBatra : https://github.com/pandas-dev/pandas/issues lists all open issues that you are more than welcome to tackle!