Release 0.90#

Warning

This release has a memory leak and should not be used.

General Changes#

  • Initial support for partition and placement awareness in the query planner. This can result in better plans for queries involving JOIN and GROUP BY over the same key columns.

  • Improve planning of UNION queries.

  • Add presto version to query creation and completion events.

  • Add property task.writer-count to configure the number of writers per task.

  • Fix a bug when optimizing constant expressions involving binary types.

  • Fix bug where a table writer commits partial results while cleaning up a failed query.

  • Fix a bug when unnesting an array of doubles containing NaN or Infinity.

  • Fix failure when accessing elements in an empty array.

  • Fix “Remote page is too large” errors.

  • Improve error message when attempting to cast a value to UNKNOWN.

  • Update the approx_distinct() documentation with correct standard error bounds.

  • Disable falling back to the interpreter when expressions fail to be compiled to bytecode. To enable this option, add compiler.interpreter-enabled=true to the coordinator and worker config properties. Enabling this option will allow certain queries to run slowly rather than failing.

  • Improve JDBC Driver conformance. In particular, all unimplemented methods now throw SQLException rather than UnsupportedOperationException.

Functions and Language Features#

Hive Changes#

  • Disable optimized metastore partition fetching for non-string partition keys. This fixes an issue were Presto might silently ignore data with non-canonical partition values. To enable this option, add hive.assume-canonical-partition-keys=true to the coordinator and worker config properties.

  • Don’t retry operations against S3 that fail due to lack of permissions.

SPI Changes#

  • Add getColumnTypes to RecordSink.

  • Use Slice for table writer fragments.

  • Add ConnectorPageSink which is a more efficient interface for column-oriented sources.

Note

This is a backwards incompatible change with the previous connector SPI. If you have written a connector, you will need to update your code before deploying this release.