Use custom functions as dataframe methods in polars

pipe

Chaining operations in polars is encouraged to take advantage of the query engine in the backend. However, you may want to apply a function that is too specific to your needs and not provided by the built-in functions. Fortunately, there is a way to chain operations using custom-made functions in polars. Below is a dataframe showing Apple stock.

shape: (10, 2)

Date	Adj Close
date	f64
2019-08-16	49.954624
2019-08-19	50.885971
2019-08-20	50.88839
2019-08-21	51.43996
2019-08-22	51.39642
2019-08-23	49.020844
2019-08-26	49.952202
2019-08-27	49.388546
2019-08-28	49.719963
2019-08-29	50.561813

Create custom function

Suppose you want to calculate the delta change based on the Adj Close price. You can create a custom function as follows:

def delta_change(df, column):
    return df.with_columns(
        (pl.col(column) - pl.col(column).shift(-1)).alias('Delta Change')
    )

Apply custom function

To apply the custom function as a chained operation, use the pipe method in polars. This method accepts the custom function and its parameters. The custom function created above has parameters df and column. Let’s apply the function.

(df
 .pipe(delta_change, column='Adj Close')
 )

shape: (10, 3)

Date	Adj Close	Delta Change
date	f64	f64
2019-08-16	49.954624	-0.931347
2019-08-19	50.885971	-0.002419
2019-08-20	50.88839	-0.55157
2019-08-21	51.43996	0.04354
2019-08-22	51.39642	2.375576
2019-08-23	49.020844	-0.931358
2019-08-26	49.952202	0.563656
2019-08-27	49.388546	-0.331417
2019-08-28	49.719963	-0.84185
2019-08-29	50.561813	null

I updated my Polars course. Be sure to enroll.