Question Details

No question body available.

Tags

python dataframe python-polars

Answers (4)

Accepted Answer Available
Accepted Answer
October 29, 2025 Score: 5 Rep: 359 Quality: High Completeness: 50%

I used a regular expression to solve this issue. In Polars you signal regular expressions with the start and end symbols ^ and $ and the needs to be escaped so the full solution looks like

python
import polars as pl
df = pl.DataFrame({"A": [1, 2], "B": [3, None], "": [4, 5]})
print(df.select(pl.col(r"^\$")))  # --> prints only the "" column
October 30, 2025 Score: 3 Rep: 15,304 Quality: Low Completeness: 60%

You could use pl.DataFrame.getitem as well:

df[[""]]

Output:

shape: (2, 1)
┌─────┐
│    │
│ --- │
│ i64 │
╞═════╡
│ 4   │
│ 5   │
└─────┘
October 30, 2025 Score: 0 Rep: 24,776 Quality: Medium Completeness: 80%

Update: Polars 1.38.1

cs.byname treating * as a wildcard was considered a bug and fixed by pull/26437

df.drop(cs.byname(""))

shape: (2, 2)


┌─────┬──────┐


│ A   ┆ B    │


│ --- ┆ ---  │


│ i64 ┆ i64  │


╞═════╪══════╡


│ 1   ┆ 3    │


│ 2   ┆ null │


└─────┴──────┘

Original answer

It seems only the col() docs mention being a wildcard, i.e. pl.col("") is the same as pl.all()

I believe df.drop("") ends up as df.drop(pl.col("")) so bare strings go through the wildcard / regex parsing.

df.drop(r"^A.$")
shape: (2, 2)


┌──────┬─────┐


│ B    ┆    │


│ ---  ┆ --- │


│ i64  ┆ i64 │


╞══════╪═════╡


│ 3    ┆ 4   │


│ null ┆ 5   │


└──────┴─────┘

There is a growing Selectors API, but it doesn't seem to contain a direct approach either.

I thought cs.by_name() would only take "literal names" but it seems to be parsed the same as col():

import polars.selectors as cs
df.select(cs.by_name(""))
shape: (2, 3)


┌─────┬──────┬─────┐


│ A   ┆ B    ┆    │


│ --- ┆ ---  ┆ --- │


│ i64 ┆ i64  ┆ i64 │


╞═════╪══════╪═════╡


│ 1   ┆ 3    ┆ 4   │


│ 2   ┆ null ┆ 5   │


└─────┴──────┴─────┘


>>> print(cs.by_name(""))
cs.all()

contains / startswith / endswith regex escape their inputs.

>>> print(cs.starts_with(""))
cs.matches("^(\).$")
>>> print(cs.contains(""))
cs.matches("^.(\).$")

df.drop(cs.contains(""))
shape: (2, 2)


┌─────┬──────┐


│ A   ┆ B    │


│ --- ┆ ---  │


│ i64 ┆ i64  │


╞═════╪══════╡


│ 1   ┆ 3    │


│ 2   ┆ null │


└─────┴──────┘

I would have expected matches to produce an invalid pattern error or such:

df.drop(cs.matches(""))
shape: (0, 0)


┌┐


╞╡


└┘

There are also some general selectors which could exclude (and others):

df.select(cs.alphanumeric())
shape: (2, 2)


┌─────┬──────┐


│ A   ┆ B    │


│ --- ┆ ---  │


│ i64 ┆ i64  │


╞═════╪══════╡


│ 1   ┆ 3    │


│ 2   ┆ null │


└─────┴──────┘
October 30, 2025 Score: 0 Rep: 99 Quality: Low Completeness: 30%

You could use

df[:, [index]]

if you know the column position.