DataFrame rename fails with TypeError when give a ChainMap instead of a dict

[dave-kirby]

Example code:

>>> from collections import ChainMap
>>> import pandas as pd
>>> df = pd.DataFrame({'A': range(1, 11), 'B': np.random.randn(10)})
>>> df.rename(ChainMap({'A': 'a'}, {'B':'b'}), axis='columns')
...
TypeError: 'ChainMap' object is not callable

# also fails if called with the columns parameter:
>>> df.rename(columns=ChainMap({'A': 'a'}, {'B':'b'}))

I am trying to rename columns in a DataFrame using mappings from two or more dicts combined using the collections.ChainMap class, but it fails with a TypeError.

The docs for DataFrame.rename says

mapper, index, columns : dict-like or function, optional
dict-like or functions transformations to apply to
that axis' values. Use either mapper and axis to
specify the axis to target with mapper, or index and
columns

the ChainMap class should definitely qualify as "dict-like" since it implements the Mapping protocol:

>>> from collections import Mapping
>>> issubclass(ChainMap, Mapping)
True

I am using Python 3.6.6 with Pandas 0.23.4.

[dave-kirby]

N.B. I had a quick look at the source code and the problem is in the rename method in generic.py:

        # renamer function if passed a dict
        def _get_rename_function(mapper):
            if isinstance(mapper, (dict, ABCSeries)):

                def f(x):
                    if x in mapper:
                        return mapper[x]
                    else:
                        return x
            else:
                f = mapper

            return f

The line if isinstance(mapper, (dict, ABCSeries)): could be replaced with if isinstance(mapper, (collections.Mapping, ABCSeries)): - this would be backward compatible to python 2.6. However if that is going to be done, IMHO it should be done everywhere a dict is allowed as a parameter, otherwise I would submit a one-line PR. Alternatively the docs should be updated to say that only dicts and their subclasses are accepted, not "dict-like" objects.

[jreback]

we have an is_dict_like function for this
welcome a PR to fix here (we use this almost everywhere)

[dave-kirby]

OK, I will create a PR when I have some free time.

[bourbaki]

@dave-kirby I could fix the issue. I have some free time tomorrow. Do you mind?

[dave-kirby]

@Ma3aXaKa sorry I am using my work GitHub account so did not see your message over the weekend. If you still have some free time then go ahead.

[mroeschke]

Looks to work on master. Could use a test.

In [136]: >>> from collections import ChainMap
     ...: >>> import pandas as pd
     ...: >>> df = pd.DataFrame({'A': range(1, 11), 'B': np.random.randn(10)})
     ...: >>> df.rename(ChainMap({'A': 'a'}, {'B':'b'}), axis='columns')
Out[136]:
    a         b
0   1 -0.527207
1   2 -0.261854
2   3 -0.247372
3   4  0.522214
4   5  0.415626
5   6 -0.361259
6   7 -1.349333
7   8  1.096590
8   9  0.342080
9  10  1.911814

In [137]: >>> df.rename(columns=ChainMap({'A': 'a'}, {'B':'b'}))
     ...:
Out[137]:
    a         b
0   1 -0.527207
1   2 -0.261854
2   3 -0.247372
3   4  0.522214
4   5  0.415626
5   6 -0.361259
6   7 -1.349333
7   8  1.096590
8   9  0.342080
9  10  1.911814

In [138]: pd.__version__
Out[138]: '0.26.0.dev0+684.g953757a3e'