Home Can I use pandas `groupby().apply()` to do stuff that doesn't return a value?

Questions

Can I use pandas `groupby().apply()` to do stuff that doesn't return a value?

November 10, 2023

Suppose I have a function that does something based on a DataFrame but doesn’t return a value. Something like, writing to a log file, or uploading to a database.

def some_func_no_return_value(df):
    # Do something based on df
    return # Doesn't return anything

I have some code that does this:

for key, df_group in df.groupby(some_column):
    some_func_no_return_value(df_group)

Should the below code work the same? For me, it doesn’t do anything, and I’m wondering if that’s expected or if I’ve made a mistake somehow:

df.groupby(some_column).apply(lambda df_group : some_func_no_return_value(df_group))

Thanks for your help.

>Solution :

Well, you can, the same way that you can use a list comprehension for side effects:

[print(x) for x in iterable]

But you shouldn’t.

This is frowned upon as you will create an useless object as output (here an empty DataFrame).

The loop is the correct way to proceed:

for key, df_group in df.groupby(some_column):
    some_func_no_return_value(df_group)

Example:

df = pd.DataFrame({'col1': list('ABABA'),
                   'col2': range(5)})

out = df.groupby('col1').apply(print)

print(type(out))

for _, g in df.groupby('col1'):
    print(g)

Output:

  col1  col2                           # groupby side-effect
0    A     0
2    A     2
4    A     4
  col1  col2
1    B     1
3    B     3
<class 'pandas.core.frame.DataFrame'>  # useless output
  col1  col2                           # loop
0    A     0
2    A     2
4    A     4
  col1  col2
1    B     1
3    B     3