Parquet fits well into a system like Apache Arrow for good query performance, when considering partitioning (via directories), push-down predicates, and read-on

Parquet fits well into a system like Apache Arrow for good query performance, when considering partitioning (via directories), push-down predicates, and read-only-the-columns-you-need: https://arrow.apache.org/docs/python/dataset.html https://arrow.apache.org/docs/python/dataset.html "Tabular Datasets"