ENH: DataFrame.replace does not work with dictionary subclasses

[timgeb]

Is your feature request related to a problem?

I want to use DataFrame.replace with a dictionary that has a custom __getitem__ logic. The current behavior is easy to demonstrate with a defaultdict (see code below).

Describe the solution you'd like

I would like DataFrame.replace to honor the __getitem__ method of dict-like arguments.

API breaking implications

I don't know.

Code

>>> from collections import defaultdict
>>> import pandas as pd
>>> df = pd.DataFrame([1, 2, 3])
>>> d = defaultdict(int, {1: -1, 2: -2})
>>> df.replace(d)
   0
0 -1
1 -2
2  3
>>> d[3]
0

Expected behavior: the 3 inside the DataFrame should be replaced with 0, because d[3] returns 0.

[mroeschke]

• replace doesn't work inplace. For the replacement to carry over, you must reassign it to a new variable df = df.replace(d)
• Your example d[3] is a getitem on the defaultdict not the dataframe

[mroeschke]

In [3]: >>> from collections import defaultdict
   ...: >>> import pandas as pd
   ...: >>> df = pd.DataFrame([1, 2, 3])
   ...: >>> d = defaultdict(int, {1: -1, 2: -2})
   ...: >>> df = df.replace(d)
   ...:
   ...:

In [4]: df
Out[4]:
   0
0 -1
1 -2
2  3

[timgeb]

You misunderstand. The issue is that 3 inside the DataFrame is not replaced with 0.

[mroeschke]

Ah, sorry for the misunderstanding. I see your point now.

I suppose this could be supported. Investigation and PR's welcome!

[timgeb]

No problem, I clarified the initial post.