Delta-rs delivered a focused March 2026 release with improvements across metadata accuracy, incremental processing, vacuum behavior, schema evolution, and query planning. The changes strengthen data integrity, performance, and API hygiene while expanding test coverage and regression safety.

February 2026: Enhanced data processing configurability and robustness in delta-rs. Implemented session-aware scan configuration, tightened file-id filter semantics in scan planning, added guard rails in the file_id helper with tests, and introduced robust writer batch channel size parsing with a default fallback to handle misconfigurations. These changes improve production reliability, correctness of data scans, and resilience of writer configurations, directly advancing data processing performance and safety.

January 2026 (2026-01) monthly summary for langchain-ai/delta-rs focused on stabilizing and upgrading DataFusion integration with Delta Lake. Delivered broad DF52 migration, improved compatibility, and reinforced testing and bindings to support DF 52.x deployments and reliable Delta Lake workloads.

December 2025: Delivered DML support in DataFusion by adding new TableProvider hooks for DELETE/UPDATE, wiring through the physical planner to MemTable, and establishing robust test coverage. This work unlocks SQL-based row-level mutations for downstream engines (e.g., iceberg-rust, delta-rs) without requiring bespoke planners, while preserving backward compatibility with default NotImplemented behavior for engines that don’t yet support DML.

Month: 2025-11 — Delivery focused on automating storage backend registration in the deltalake Rust crate, enabling seamless GCS URI support and preparatory cross-backend auto-registration. This reduces setup friction, improves reliability when using cloud URIs, and aligns Rust meta-crate behavior with Python bindings.