Fix planning performance regression that can affect queries over Hive tables with many partitions.
Fix deadlock in memory management logic introduced in the previous release.
n(number of result elements) to 10,000 or less for
min_by(col1, col2, n), and
max_by(col1, col2, n).
Improve error message when a session property references an invalid catalog.
Reduce memory usage of
Improve coordinator CPU efficiency when discovering splits.
Include minimum and maximum values for columns in
Web UI Changes#
Fix previously empty peak memory display in the query details page.
Fix regression in CLI that makes it always print “query aborted by user” when the result is displayed with a pager, even if the query completes successfully.
Return a non-zero exit status when an error occurs.
--client-infooption for specifying client info.
--ignore-errorsoption to continue processing in batch mode when an error occurs.
JDBC Driver Changes#
Allow configuring connection network timeout with
Allow setting client tags via the
ClientTagsclient info property.
Expose update type via
Consistently fail queries that attempt to read partitions that are offline. Previously, the query can have one of the following outcomes: fail as expected, skip those partitions and finish successfully, or hang indefinitely.
Allow setting username used to access Hive metastore via the
respect_table_formatsession properties, corresponding to the
Reduce ORC file reader memory consumption by allocating buffers lazily. Buffers are only allocated for columns that are actually accessed.
Fix failure when querying
information_schema.columnswhen there is no equality predicate on