Grouping with over in polars

over

Group by operations usually reduce the size of a dataframe, both in the number of columns and rows returned. However, sometimes you may want to maintain the original dataframe size.

Below is a dataframe showing different types of groceries bought on each day of the week.

shape: (9, 3)

Day	Product	Price
str	str	f64
"Mon"	"Bread"	3.89
"Tue"	"Milk"	4.0
"Wed"	"Tea"	3.89
"Mon"	"Cheese"	7.99
"Tue"	"Oats"	8.0
"Wed"	"Yogurt"	4.67
"Mon"	"Apple"	4.0
"Tue"	"Juice"	7.89
"Wed"	"Lettuce"	5.99

Group with over

To perform a group by operation while keeping the same dataframe length in Polars, you use the over expression like this:

(df
 .with_columns(Avg_Spend=pl.col('Price').mean().over('Day'))
 )

shape: (9, 4)

Day	Product	Price	Avg_Spend
str	str	f64	f64
"Mon"	"Bread"	3.89	5.293333
"Tue"	"Milk"	4.0	6.63
"Wed"	"Tea"	3.89	4.85
"Mon"	"Cheese"	7.99	5.293333
"Tue"	"Oats"	8.0	6.63
"Wed"	"Yogurt"	4.67	4.85
"Mon"	"Apple"	4.0	5.293333
"Tue"	"Juice"	7.89	6.63
"Wed"	"Lettuce"	5.99	4.85

Here, we’ve calculated the average spend for each day.

Note

The size of the dataframe remains the same in both dataframes, with 9 rows.

Enroll in Polars course to learn more.