polars-lance 0.5.0

Polars plugin for reading and writing Lance datasets

polars-lance

Polars plugin for reading Lance datasets into Polars dataframes and writing Polars dataframes to Lance datasets.

Installation

pip install polars-lance

Read

from polars_lance import scan_lance

lf = scan_lance("example.lance")
df = lf.collect()

Write

import polars as pl

from polars_lance import write_lance

df = pl.DataFrame({"id": [1, 2], "val": ["a", "b"]})
write_lance(df, "example.lance")

Pass mode to control behavior in case the target dataset already exists. Valid values are error (default), append, and overwrite.

write_lance(df, "example.lance", mode="append")

Cloud storage

Pass storage_options to work with Lance datasets stored in AWS S3, Azure Blob Storage, or Google Cloud Storage.

from polars_lance import scan_lance, write_lance

uri = "s3://my-bucket/example.lance"
storage_options = {
    "aws_access_key_id": "AKIAIOSFODNN7EXAMPLE",
    "aws_secret_access_key": "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY",
    "aws_region": "us-east-1",
}

write_lance(df, uri, storage_options=storage_options)

lf = scan_lance(uri, storage_options=storage_options)
df = lf.collect()

Cloud storage IO is powered by the object_store Rust crate. These are the supported schemes and options:

• s3://: AWS options
• az://: Azure options
• gs://: GCP options