Delta Lake 4.4.0 brought Apache Spark 4.2 support, SQL identity columns, and expanded Flink integrations.

Delta Lake 4.4.0 landed with support for Apache Spark 4.2, identity columns, and generated-column definitions in SQL DDL [5]. The release also expanded Unity Catalog Delta API integration to Delta Kernel and the experimental Flink connector, while introducing Flink upsert mode with merge-on-read updates [5]. Earlier in the month, maintenance release Delta Lake 3.3.3 resolved transaction log retention defects that broke time travel and Change Data Feed reads, fixed a Delta Sharing deletion vector cache bug, and added an opt-in prefix randomization setting for high-throughput S3 workloads [8].

Multimodal workloads gained native table support with the introduction of the FILE column type in beta, establishing a pattern for storing documents, audio, and video directly within Delta Lake tables under catalog governance [9]. In parallel, Unity Catalog 0.6.0 delivered Spark 4.2 support and introduced governed SQL and metric views [4]. These architectural additions fed into active community evaluations of table formats, where practitioners weighed Delta Lake with UniForm against Apache Iceberg for greenfield catalog-coordinated deployments [1, 2].

For local testing, the community introduced Minilake, an emulator enabling developers to test Databricks SDK and Terraform code against real Delta Lake storage without spinning up cloud clusters [6]. Engineers also weighed streaming patterns, analyzing Databricks Zerobus against Microsoft Fabric open mirroring [3].