Manipulate pandas dataframe with custom function-CodePudding

I am having trouble with applying a customer function to a dataframe. The function works fine and returns the correct dataframe. However, after having it applied my dataframe is still the old one.

My DataFrame is like this:

d = {'col 1' : ['a', 'a', 'a', 'b', 'b', 'b'],'col 2' : [1, 1, 2, 2, 1, 2]}
df = pd.DataFrame(data = d)
df.set_index('col 1')

I wrote a little test function to group by col 1 and find the highest value in col 2 for each group:

def tester(x):
x = x.groupby('col 1', group_keys=False).apply(lambda x: x.nlargest(1, 'col 2'))
return x

then using the function on my dataframe returns the correctly grouped result:

tester(df)

However outside the function the df is still the old one. Why does this happen? This behavior does not occur with lists or dictionaries for instance. How can I continue working with the manipulated df after the function?

Thanks!

CodePudding user response：

You just need to store the return in a new variable or store it in the same to replace

import pandas as pd
d = {'col 1' : ['a', 'a', 'a', 'b', 'b', 'b'],'col 2' : [1, 1, 2, 2, 1, 2]}
df = pd.DataFrame(data = d)
df.set_index('col 1')


def tester(x):
    x = x.groupby('col 1', group_keys=False).apply(lambda x: x.nlargest(1, 'col 2'))
    return x

nw_DF = tester(df) -> New dataframe 
print(str(nw_DF))

CodePudding user response：

Short answer: Once the assignment occurs inside the tester function, the left side of the assignment comes under the local scope of the function.

What this means: When x = ... is assigned, x is now local to the function, thus breaking the link to the DataFrame 'outside'. Sure, x is returned to the caller, yet the caller is not assigning the returned value to anything. So the DataFrame (df) remains unchanged.

The fix is simple. Change this:

tester(df)

To this:

df = tester(df)