Delta Lake 4.3, released in June 2026, is built on Apache Spark 4.1.0 and 4.0.1. The update introduces replaceUsing and replaceOn APIs for selective data overwriting in DataFrames, replacing the previous replaceWhere method. These tools allow developers to specify columns or logical conditions for updates without managing complex predicates.
A major structural change involves the transition to the Unity Catalog Delta REST API for all table operations, including metadata updates, DML, and schema changes. This ensures consistent behavior across different engines like Flink, Trino, and DuckDB, while enhancing commit validation and reducing operational risks for multi-engine environments.
UniForm now supports atomic and incremental metadata conversion to Iceberg. The experimental IcebergCompatV3 feature enables the simultaneous use of deletion vectors and UniForm. Additionally, Delta 4.3 improves performance for large tables by parallelizing checkpoint writes and collecting statistics for Variant columns to optimize data skipping.
The release updates Structured Streaming and Change Data Feed (CDF) to work with catalog-managed tables. Delta Sharing now supports streaming reads from shared tables. Furthermore, Delta Spark 4.3 introduces implicit type casting for DataFrame writes by name, aligning with SQL INSERT BY NAME behavior to improve data consistency.
