v4.0.0·Delta Lake 4.0.0
Summary
Delta Lake 4.0.0 introduces Delta Connect for Spark Connect connectivity, catalog-managed tables in preview for catalog-controlled commits, full Variant data type support for semi-structured storage, and graduates Type Widening to enable column type changes without rewriting data. The release requires Apache Spark 4.0 and discontinues active development of Delta Standalone and related connectors (Flink, Hive) in 4.x, shifting them to maintenance-only mode.
Summary from the captured archive. Follow the original source for full context.
Topics
More from delta-io/delta
Delta Lake 4.4.1
Delta Lake 4.4.1 adds safeguards for catalog-managed tables by blocking unsupported cleanups and skipping automatic maintenance so writes can succeed reliably. It also fixes snapshot construction errors caused by unratified staged commits, allowing safe re-enablement of S3A fast listing.
Delta Lake 4.4.0
Delta Lake 4.4.0 supports Apache Spark 4.2 and adds identity columns and generated-column support in SQL DDL. The release expands UC Delta API integration to Delta Kernel and the experimental Flink connector for catalog-managed tables, introduces Flink upsert mode with merge-on-read updates, and improves VOID-column and partition handling.
Delta Lake 3.3.3
Delta 3.3.3 fixes transaction log retention bugs that broke time travel and CDF reads, and a Delta Sharing deletion vector cache bug causing long-running queries to fail. It adds opt-in RANDOMIZE_FILE_PREFIXES to spread S3 object keys for high-throughput workloads and upgrades Delta Sharing client for improved OAuth and retry handling.
Delta Lake 4.3.1
Delta Lake 4.3.1 fixes OAuth authentication failures in the Delta REST Catalog caused by incorrect key lowercasing and enables S3A fast listing when using OSS UnityCatalog's CredScopedFileSystem wrapper. It also prevents the reserved is_managed_location property from persisting into managed table metadata.
Delta Lake 4.3.0
Delta 4.3.0 deepens Unity Catalog integration by making it the source of truth for managed table operations via the UC Delta REST API, introduces replaceOn/replaceUsing DataFrame APIs for selective row-level data replacement, and improves UniForm with atomic Iceberg conversion and incremental metadata updates. Delta Sharing gains streaming support, Change Data Feed capabilities, and Trigger.AvailableNow, plus performance improvements like better V2 checkpoint parallelization and variant column statistics for data skipping.
Delta Lake 4.2.0
Delta 4.2.0 enables atomic REPLACE TABLE, RTAS, and DPO for catalog-managed tables, enhances streaming capabilities, and adds a Kernel-based Flink connector. The release makes Variant generally available, adds geospatial and collation support, and includes comprehensive security hardening.