Polars is an analytical query engine for DataFrames built from the ground up in Rust for maximum performance. With bindings for Python, Node.js, R, and native Rust, it provides a powerful alternative to pandas that handles datasets far larger than available RAM through its streaming execution engine.

This skill teaches agents how to use the Polars library for data transformation workflows. Agents learn to construct lazy evaluation pipelines that optimize query execution automatically, apply columnar operations using Polars’ expressive API, and process data in parallel using multi-threaded execution. The lazy execution model means Polars can reorder, eliminate, and fuse operations before any data is touched, resulting in query plans that run orders of magnitude faster than naive row-by-row processing.

Key capabilities include reading and writing Parquet, CSV, JSON, and Arrow IPC files; performing joins, aggregations, window functions, and group-by operations; handling missing data with built-in null-aware operations; and leveraging SIMD vectorization for compute-heavy transformations. The streaming engine allows processing of datasets that exceed available memory by executing queries in configurable batches.

Agents using this skill produce clean, transformed datasets ready for downstream analytics, machine learning pipelines, or database loading. Integration points include reading directly from cloud storage (S3, GCS, Azure), interoperating with Apache Arrow for zero-copy data exchange with other tools, and outputting to any supported format. Polars is MIT-licensed, actively maintained with weekly releases, and has over 37,000 GitHub stars, making it one of the most popular data processing libraries in the open-source ecosystem.