`groupby` 列未传递给 `apply` 函数。 Pandas 中可能存在的错误？

Question

Possible bug in GroupBy.apply (pandas 0.25.3): The following code makes a groupby by the class column, and the applied function tries to access the class column. The code produces a KeyError: 'class' exception:

import pandas as pd

df = pd.DataFrame(
    [
        ("bird", "Falconiformes", 389.0),
        ("bird", "Psittaciformes", 24.0),
        ("mammal", "Carnivora", 80.2),
        ("mammal", "Primates", 20),
        ("mammal", "Carnivora", 58),
    ],
    index=["falcon", "parrot", "lion", "monkey", "leopard"],
    columns=("class", "order", "max_speed"),
)

class_to_features = {"bird": ["wings", "feathers", "beak"], "mammal": ["udder"]}


def exec_groupby(df, _temp, c_2_f=None):
    def _helper(df):
        if c_2_f is not None:
            return c_2_f[df["class"].iloc[0]] + _temp # KeyError "class"
        else:
            return "goo" + _temp

    return df.groupby(["class"]).apply(lambda df: _helper(df))

print(exec_groupby(df, "foo"))
print(exec_groupby(df, "foo", class_to_features))

However, if I remove the + _temp from the return statement it works fine!

def exec_groupby(df, _temp, c_2_f=None):
    def _helper(df):
        if c_2_f is not None:
            return c_2_f[df["class"].iloc[0]] # Works fine! no errors
        else:
            return "goo" + _temp

    return df.groupby(["class"]).apply(lambda df: _helper(df))

It seems like a combination of c_2_f not being None and access the column that causes the exception. Am I missing something?

Answer 1

The error log reads

TypeError: can only concatenate list (not "str") to list
...
During handling of the above exception, another exception occurred:
...
KeyError: 'class'

c_2_f has keys which are lists, so you can't concatenate it with str

def exec_groupby(df, _temp, c_2_f=None):
    def _helper(df):
        if c_2_f is not None:
            return c_2_f[df["class"].iloc[0]] + [_temp] # Works fine! no errors
        else:
            return "goo" + _temp

    return df.groupby(["class"]).apply(lambda df: _helper(df))

This works fine and gives

class
bird      goofoo
mammal    goofoo
dtype: object
class
bird      [wings, feathers, beak, foo]
mammal                    [udder, foo]
dtype: object