Materialize v0.99

v0.99

Sources and sinks

  • Private preview. Support exporting objects and query results to Amazon s3 using the COPY TO command and AWS connections. Both CSV and Parquet are supported as file formats.

    Syntax

    CREATE CONNECTION s3_conn
     TO AWS (ASSUME ROLE ARN = 'arn:aws:iam::000000000000:role/Materializes3Exporter');
    
    COPY mv TO 's3://mz-to-s3/'
    WITH (
      AWS CONNECTION = aws_role_assumption,
      FORMAT = 'parquet'
    );
    

    It’s important to note that this command isn’t supported in the SQL Shell yet, but will be in the next release (#27114).

  • Support ingesting datetime columns as text via the TEXT COLUMNS option in the MySQL source to work around MySQL’s zero value for datetime types (0000-00-00, 0000-00-00 00:00:00), as well as other differences in the range of supported values between MySQL and PostgreSQL.

SQL

  • Private preview. Support setting a history retention period for sources, tables, materialized views, and indexes via the new RETAIN HISTORY option. This is useful to implement durable subscriptions.

    Syntax

    ALTER MATERIALIZED VIEW winning_bids SET (RETAIN HISTORY FOR '2hr');
    
    ALTER MATERIALIZED VIEW winning_bids RESET (RETAIN HISTORY);
    
  • Add mz_internal.mz_history_retention_strategies to the system catalog. This table describes the history retention strategies for tables, sources, indexes, and materialized views that are configured with a history retention period.

  • Add mz_internal.mz_materialized_view_refreshes to the system catalog. This table shows the time of the last successfully completed refresh and the time of the next scheduled refresh for each materialized view with a refresh strategy other than on-commit.

Bug fixes and other improvements

  • Allow interval types to be cast to mz_timestamp (#26970).

  • Move the mz_cluster_replica_sizes system catalog table from the mz_internal schema to mz_catalog, making the table definition stable. Any queries referencing the mz_internal.mz_cluster_replica_sizes catalog table must be adjusted to use mz_catalog.mz_cluster_replica_sizes instead.

Back to top ↑