Back to Materialize

Materialize v0.29

doc/user/content/releases/v0.29.md

1231.4 KB
Original Source

v0.29.0

  • Fix a bug where implicit type casts prevented indexes from being used {{% gh 15476 %}}.

  • Improve Materialize's ability to use indexes when comparing column expressions to literal values, particularly in cases where e.g. col_a was of type VARCHAR:

    mzsql
    SELECT * FROM table_foo WHERE col_a = 'hello';
    
  • Fix a bug that prevented using pre-existing topics with multiple partitions in Kafka sinks {{% gh 15609 %}}. Previously, the sink would use the default Kafka cluster configuration also for pre-existing topics, instead of the user-configured number of partitions.

  • Improve ordering for joins that have filters applied to their inputs. This leads to an order of magnitude performance improvement in cases with highly selective filters {{% gh 15120 %}}.

  • Treat some errors as transient instead of fatal in the PostgreSQL source. Errors that would previously set the source into an error state will now retry {{% gh 15200 %}}.

  • Allow users to create indexes on system objects to optimize the performance of troubleshooting queries.

  • Include indexes created on system objects when running the SHOW INDEXES command if the IN CLUSTER clause is specified.

  • Add a TPCH load generator source, which implements the TPC-H benchmark specification.