21.22. Release 323

General Changes

  • Fix query failure when referencing columns from a table that contains hidden columns. (#1796)
  • Fix a rare issue in which the server produces an extra row containing the boolean value true as the last row in the result set. For most queries, this will result in a client error, since this row does not match the result schema, but is a correctness issue when the result schema is a single boolean column. (#1732)
  • Allow using .* on expressions of type ROW in the SELECT clause to convert the fields of a row into multiple columns. (#1017)

JDBC Driver Changes

  • Fix a compatibility issue when connecting to pre-321 servers. (#1785)
  • Fix reporting of views in DatabaseMetaData.getTables(). (#1488)

CLI Changes

  • Fix a compatibility issue when connecting to pre-321 servers. (#1785)

Hive Changes

  • Fix the ORC writer to correctly write the file footers. Previously written files were sometimes unreadable in Hive 3.1 when querying the table for a second (or subsequent) time. (#456)
  • Prevent writing to materialized views. (#1725)
  • Reduce metastore load when inserting data or analyzing tables. (#1783, #1793, #1794)
  • Allow using multiple Hive catalogs that use different Kerberos or other authentication configurations. (#760, #978, #1820)

PostgreSQL Changes

  • Support for PostgreSQL arrays is no longer considered experimental, therefore the configuration property postgresql.experimental.array-mapping is now named to postgresql.array-mapping. (#1740)

SPI Changes

  • Add support for unnesting dictionary blocks duration compaction. (#1761)
  • Change LazyBlockLoader to directly return the loaded block. (#1744)

Note

This is a backwards incompatible changes with the previous SPI. If you have written a plugin that instantiates LazyBlock, you will need to update your code before deploying this release.