Arranging columns in a specific order using index in polars

Stacking dataframes vertically only works when the column names are the same and arranged in the same order. For example, if you have df1 with columns Name and Age, and df2 with columns Age and Name, these two dataframes cannot be stacked into a single dataframe. Although they contain the same column names, the order of the columns does not align.

Inefficient ordering of columns

You can rearrange columns using select by typing out the column names one by one. The columns in the dataframe will appear in the order you specify. However, this method can be tedious, especially when working with many columns or columns with long names.

Efficient ordering of columns

If you’re too lazy to type out the columns names, you can reference them by index. Polars provides the pl.nth function that lets you to do this. Here’s how you can rearrange the columns in the dataframe below.

shape: (5, 5)

Number_of_People	Budget	Department_Name	Location	Manager
i64	i64	str	str	str
17	50000	"Hospitality"	"Thailand"	"Sritala"
25	120000	"Legal"	"New York"	"Harvey Specter"
15	80000	"Finance"	"London"	"Ken Leung"
12	60000	"Advertising"	"New York"	"Don Draper"
20	90000	"Sales"	"Scranton"	"Michael Scott"

Now let’s arrange the columns by referencing their index (position) instead of manually typing the column names.

(df
 .select(pl.nth(2,4,3,0,1))
 )

shape: (5, 5)

Department_Name	Manager	Location	Number_of_People	Budget
str	str	str	i64	i64
"Hospitality"	"Sritala"	"Thailand"	17	50000
"Legal"	"Harvey Specter"	"New York"	25	120000
"Finance"	"Ken Leung"	"London"	15	80000
"Advertising"	"Don Draper"	"New York"	12	60000
"Sales"	"Michael Scott"	"Scranton"	20	90000

See how much time this method can save you?

Check out my Polars course!