Release of Polars 2.0
4 hours ago
- Polars 2.0 ships with initial out-of-core (spill-to-disk) support enabled by default, improving memory resilience.
- SQL is now a first-class citizen with expanded coverage and performance optimizations including join reordering and bloom filters.
- Benchmarks show Polars leading DuckDB and DataFusion in TPC-H and TPC-DS queries, though with overhead at high thread counts.
- The new streaming engine is default in `collect`, breaking row-order guarantees for certain operations unless `maintain_order=True`.
- Polars now supports the Arrow `Map` dtype directly with dedicated expressions like key lookups and dictionary methods.
- Stricter dtype and explicitness defaults provide faster feedback by catching schema mismatches early.
- Future roadmap includes better out-of-core support for joins and group-by, improved scaling, and the start of GeoPolars.