For installation instructions check out the getting started guide.
Added
- iceberg: Iceberg connector now auto-creates correctly-typed columns for temporal and UUID fields from Avro schemas and includes schema-aware numeric scaling for time-typed columns. (@Jeffail, #4427)
- iceberg: Added require_schema_metadata strict temporal mode option to reject numeric inputs into time-typed columns when schema metadata is absent, preventing silent corruption on schema drift. (@Jeffail, #4427)
- oracledb_cdc: Added optional external cache-backed transaction buffer to reduce in-process memory pressure for large long-running LogMiner transactions. (@josephwoodward, #4409)
Fixed
- confluent: Fixed decimal value serialization in Avro to match spec-compliant form when using default mode. (@Jeffail, #4427)
- general: Bumped golang.org/x/net dependency to address CVE-2024-79571. (@josephwoodward, #4452)
- avro: resolve named-type references in nullable unions (@Jeffail, #4429)
Changed
- confluent: Schema registry Avro decoder now honors all Avro logical types (timestamp, date, time, UUID) end-to-end and maps them correctly in Iceberg columns, with configurable preservation via preserve_logical_types flag. (@Jeffail, #4427)
- confluent: Schema registry decoder now honors Kafka Connect and Debezium temporal annotations in schema metadata, enabling correct column type selection in downstream sinks like Iceberg. (@Jeffail, #4427)
- general: Updated README example to showcase CDC to Iceberg-on-S3 lakehouse pipeline with AWS Glue catalog and dynamic routing instead of generic pubsub example. (@prakhargarg105, #4445)
- parquet: Parquet encoder now handles Date, TimeOfDay, and UUID types from schema metadata and coerces temporal values to numeric columns for rolling upgrades against existing tables. (@Jeffail, #4427)
The full change log can be found here.