20.134. Release 0.126
Add error location information (line and column number) for semantic errors.
Fix a CLI crash during tab-completion when no schema is currently selected.
Fix reset of session properties in CLI when running USE.
Fix occasional query planning failure due to a bug in the projection push down optimizer.
Fix a parsing issue when expressions contain the form
POSITION(x in (y)).
Add a new version of
approx_percentile()that takes an
Allow specifying columns names in INSERT queries.
field_lengthtable property to blackhole connector to control the size of generated
Bundle Teradata functions plugin in server package.
Improve handling of physical properties which can increase performance for queries involving window functions.
Add ability to control whether index join lookups and caching are shared within a task. This allows us to optimize for index cache hits or for more CPU parallelism. This option is toggled by the
task.share-index-loadingconfig property or the
Improve performance of queries that use an
INexpression with a large list of constant values.
Enable connector predicate push down for all comparable and equatable types.
Fix query planning failure when using certain operations such as
DISTINCT, etc. on the output columns of
maxResponseSizeto be slightly smaller than the configured value. This reduces the possibility of encountering
Fix memory leak in coordinator.
Add validation for names of table properties.
Fix reading structural types containing nulls in Parquet.
Fix writing DATE type when timezone offset is negative. Previous versions would write the wrong date (off by one day).
Fix an issue where
VARCHARcolumns added to an existing table could not be queried.
Fix over-creation of initial splits.
hive.immutable-partitionsconfig property to also apply to unpartitioned tables.
DATEcolumns as partition columns in parquet tables.
Improve error message for cases where partition columns are also table columns.