group_by with polars concatenating values

Question

I have a polars dataframe that I want to group by and concatenate the unique values in as a single entry.

in pandas, I go:

def unique_colun_values(x):
    return('|'.join(set(x)))

dd=pd.DataFrame({'col1':[1,1,2,3,4],'col2':['a','a','a','b','b'],'col3':['qwe','rty','asd','fgh','zxc']})
dd.groupby('col1').agg(lambda x: unique_colun_values(x))

This works fine

when I try to implement in polars:

pl.from_pandas(dd).group_by('col1').agg(lambda x: unique_colun_values(x), allow_object=True)

I get the following error:

TypeError: cannot create expression literal for value of type function.

Hint: Pass `allow_object=True` to accept any value and create a literal of type Object.

Am I missing something?

Rodalm · Accepted Answer · 2025-09-16 10:26:46Z

According to the documentation, unlike pandas .agg, polars .agg method does not accept a function as its first parameter, it only accepts polars expressions. This is what the error tells you, it fails when trying to convert the function into a polars expression.

The solution is to write unique_colun_values as an expression.

import pandas as pd
import polars as pl

dd = pd.DataFrame(
    {
        "col1": [1, 1, 2, 3, 4],
        "col2": ["a", "a", "a", "b", "b"],
        "col3": ["qwe", "rty", "asd", "fgh", "zxc"],
    }
)

res = pl.from_pandas(dd).group_by('col1').agg(pl.all().unique().str.join("|"))

print(res)

Output:

┌──────┬──────┬─────────┐
│ col1 ┆ col2 ┆ col3    │
│ ---  ┆ ---  ┆ ---     │
│ i64  ┆ str  ┆ str     │
╞══════╪══════╪═════════╡
│ 3    ┆ b    ┆ fgh     │
│ 2    ┆ a    ┆ asd     │
│ 4    ┆ b    ┆ zxc     │
│ 1    ┆ a    ┆ qwe|rty │
└──────┴──────┴─────────┘

If you prefer, you can still write the operation as a function that returns an expression.

def unique_colun_values(col: pl.Expr) -> pl.Expr:
    return col.unique().str.join("|")

# You can call the function using
res = pl.from_pandas(dd).group_by("col1").agg(unique_colun_values(pl.all()))

# Or, using pl.Expr.pipe 
res = pl.from_pandas(dd).group_by("col1").agg(pl.all().pipe(unique_colun_values))

Collectives™ on Stack Overflow

group_by with polars concatenating values

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related