Ndtyjky

My question is about the line

df.apply(lambda x: pd.to_numeric(x, errors='coerce'))

I do understand that this statement converts the dataframe columns to integer values, but was not able to understand the usage of the lambda function or the errors='coerce' part.

apply works on either row- or column-series by applying a function to it. lambda is just defining an anonymous function.

For readability, you could define a regular function with better variable names. Consider the following demo:

>>> df = pd.DataFrame([['1', '2'], ['3', 'foo']])

>>> df.dtypes

>>> 

0    object

1    object

dtype: object

We have a dataframe full of strings which we want to make numeric. Non-convertible values should be set to NaN (this is what errors='coerce' does).

>>> def make_numeric(series):

...:    return pd.to_numeric(series, errors='coerce')

>>> 

>>> new_df = df.apply(make_numeric)

>>> 

>>> new_df

>>> 

   0    1

0  1  2.0

1  3  NaN

>>> 

>>> new_df.dtypes

>>> 

0      int64

1    float64

dtype: object

As you can see, using the lambda is just a short way of defining a function. If you don't like it, you can always write a normal function that does the same and is probably more readable.

In this case, defining your own function is a little pointless, because you can just write:

>>> df.apply(pd.to_numeric, errors='coerce')

>>> 

   0    1

0  1  2.0

1  3  NaN