Python function behaves unexpectedly

Question

I am doing a very simple transformation on a pandas dataframe by using a function, but I didn't expect the function will change the input dataframe (but it did). I was wondering why...

Here's my code:

x = pd.DataFrame({'a': [1,2,3], 'b': [3,4,5]})

def transform(df, increment):
    new_df = df
    new_df.a = new_df.a + increment
    return new_df

new_x = transform(x, 1)

new_x # output shows new_x.a is [2,3,4], which is expected.
x # output shows x.a is also [2,3,4]. I thought it should be [1,2,3]

Why is this the case? I think, in the function, all the operations are executed on the new_df, so the input x should stay exactly the same before and after I ran this transform function, isn't it?

`new_df = df` does *not* copies the dataframe. You simply make another reference to the *same* object. — Willem Van Onsem, Nov 20 '17 at 19:57

score 1 · Accepted Answer · answered Nov 20 '17 at 19:59

This is because it does not create a copy, but another "reference" of the x object

x = pd.DataFrame({'a': [1,2,3], 'b': [3,4,5]})


def transform(df, increment):
    new_df = df.copy() # <--- piece to change
    new_df.a = new_df.a + increment
    return new_df

new_x = transform(x, 1)

new_x # output shows new_x.a is [2,3,4], which is expected.
x # output shows x.a is now [1,2,3].

When you add .copy() this will give you the expected behavior

Python function behaves unexpectedly

1 Answers1