polars-list-zip-struct

polars-list-zip-struct is a native expression plugin for the accepted Polars list.zip proposal from pola-rs/polars#22719.

It adds:

• pl.col("a").list.zip(pl.col("b"), pad=False)
• pl.zip_list("a", "b", "c", pad=False)

Both return a list[struct] expression. By default, lists are truncated to the shortest input, like Python's zip. With pad=True, shorter lists are padded with nulls to match the longest input. Struct fields default to the input column/expression names, matching the upstream Polars pull request.

If a future Polars release ships native list.zip or pl.zip_list, this package leaves the native implementation untouched.

Installation

pip install polars-list-zip-struct

The package supports Polars releases from roughly the last three months onward and does not cap future Polars versions. For this release that means polars>=1.39.3.

Usage

Import the package once to register the Polars helpers:

import polars as pl
import polars_list_zip_struct  # noqa: F401

df = pl.DataFrame(
    {
        "a": [[1, 2], [3], None, [None]],
        "b": [[10, 20], [30, 35], [40], [35]],
    }
)

out = df.with_columns(
    pl.col("a").list.zip(pl.col("b")).alias("zipped")
)

shape: (4, 3)
┌───────────┬───────────┬──────────────────┐
│ a         ┆ b         ┆ zipped           │
│ ---       ┆ ---       ┆ ---              │
│ list[i64] ┆ list[i64] ┆ list[struct[2]]  │
╞═══════════╪═══════════╪══════════════════╡
│ [1, 2]    ┆ [10, 20]  ┆ [{1,10}, {2,20}] │
│ [3]       ┆ [30, 35]  ┆ [{3,30}]         │
│ null      ┆ [40]      ┆ null             │
│ [null]    ┆ [35]      ┆ [{null,35}]      │
└───────────┴───────────┴──────────────────┘

Zip more than two list expressions with pl.zip_list:

df.with_columns(
    pl.zip_list("a", "b", "c", fields=["a", "b", "c"]).alias("all_zipped")
)

Use pad=True for zip_longest behavior:

df.with_columns(
    pl.col("a").list.zip(pl.col("b"), pad=True).alias("zipped")
)

Notes

The core implementation is a Rust Polars expression plugin built with pyo3-polars. It follows the original upstream implementation idea: build a LargeListArray of StructArray values with Arrow builders, offsets, and row validity, rather than running a Python row loop.

Development

python -m pip install -e ".[dev]"
maturin develop
python -m pytest

Build and upload:

python -m pip install ".[publish]"
maturin build --release
python -m twine upload target/wheels/*