polars-distance

0.5.3 verified Mon Apr 27 auth: no python

Polars plugin for pairwise distance functions. Version 0.5.3 supports f32 and uint64 types. Release cadence is irregular, with minor/patch releases every few months.

pip install polars-distance

Common errors

error AttributeError: module 'polars_distance' has no attribute 'pairwise_euclidean' ↓

cause Attempting to call the distance function as a top-level module attribute instead of using the expression API.

fix

Use pld.col('a').pairwise_euclidean(pld.col('b')) inside df.with_columns() or similar.

error polars.exceptions.ComputeError: ... dtype is not supported ↓

cause Passing a column with an unsupported dtype (e.g., categorical) to a distance function.

fix

Cast the column to a supported dtype: df.with_columns(pl.col('col').cast(pl.Utf8)).

Warnings

breaking Version 0.3.2 removed categorical dtype support. If you rely on categorical columns in list/set distances, upgrade with caution. ↓

fix Convert categorical columns to string or numeric types before applying distance functions.

deprecated The package previously allowed direct import of distance functions from polars_distance.polars. As of 0.5.0, the recommended import is the root module. ↓

fix Change imports to `import polars_distance as pld` and use `pld.col(...).pairwise_*`.

gotcha polars-distance requires Polars >= 0.20.0 and uses Polars plugins API. Incompatible with older Polars versions. ↓

fix Ensure Polars is updated: `pip install polars>=0.20.0`.

Imports

col
wrong
```
import polars_distance.polars as pld
```
correct
```
import polars_distance as pld
```
The correct import is just the root module; no subpackage.

Quickstart

Compute pairwise Euclidean distance between list columns 'a' and 'b'.

import polars as pl
import polars_distance as pld

df = pl.DataFrame({
    "a": [[1.0, 2.0], [3.0, 4.0]],
    "b": [[1.5, 2.5], [3.5, 4.5]]
})
df.with_columns(
    dist=pld.col("a").pairwise_euclidean(pld.col("b"))
)