bneijt / polario

Polars IO utility library

Polars IO utility library

Helpers to make it easier to read and write Hive partitioned parquet dataset with Polars.

It is meant to be a library to deal with datasets easily, but also contains a commandline interface which allows you to inspect parquet files and datasets more easily.

Dataset

Example of use of polario.hive_dataset.HiveDataset

from polario.hive_dataset import HiveDataset
import polars as pl
df = pl.from_dicts(
        [
            {"p1": 1, "v": 1},
            {"p1": 2, "v": 1},
        ]
    )

ds = HiveDataset("file:///tmp/", partition_columns=["p1"])

ds.write(df)

for partition_df in ds.read_partitions():
    print(partition_df)

To model data storage, we use three layers: dataset, partition, fragment.

• Each dataset is a lexical ordered set of partitions
• Each partition is a lexical ordered set of fragments
• Each fragment is a file on disk with rows in any order