Scanning a hive-partitioned dataset

The team also stores cleaned-up Parquet checkouts in a hive-partitioned layout, with one directory per year (checkoutyear=2023/, checkoutyear=2024/). Scan the partitioned dataset and filter on the partition column so Polars only reads the years you actually need.

polars is loaded as pl, and the root directory is in HIVE_DIR. The partition directories are printed for you, so you can see the layout.

Cet exercice fait partie du cours

<cours>Scaling and Optimizing Data Pipelines with Polars</cours>

Voir le cours

Instructions de l’exercice

Scan HIVE_DIR using the right argument to enable hive partitioning.
Filter the result to checkouts from 2024 onward.

Exercice interactif pratique

Essayez cet exercice en complétant ce code d’exemple.

requests = pl.scan_parquet(
    HIVE_DIR,
    # Enable hive partitioning
    ____=True,
)

result = (
    requests
    # Filter to the 2024 partition
    .filter(pl.col("checkoutyear") >= ____)
    .group_by("format")
    .agg(pl.col("checkouts").sum().alias("total"))
    .sort("total", descending=True)
    .collect()
)
print(result)

Modifier et exécuter le code