Fix named queries where its output does not match its column list (#14333).
Fix issue where queries with large filters in the where clause would cause excessive memory usage. Now such queries will fail quickly with
Improve performance of
CREATE FUNCTIONto allow parameter type list and return type to have a length up to 30k characters.
Add optimization for filters and projections to extract common subexpressions and evaluate them only once. This optimization can be turned off by the session property
Add support for
SHOW CREATE FUNCTION.
Add soft memory limit configuration properties. Soft memory limits are default memory limits given to each query that can be overridden using session properties up to the hard limit set by the existing configuration properties. Available soft memory limit configuration properties are
Add check to disallow invoking SQL functions in SQL function body.
Add support for limiting the total number of buffers per optimized repartitioning operator. The limit can be set by the configuration property
Add peak task total memory to query stats.
scale_qdigest()function to scale a
qdigestto a new weight.
Add support for
ST_AsText()to accept Spherical Geographies.
Add support for
ST_Centroid()to accept Spherical Geography Points and MultiPoints.
Pinot Connector Changes#
Add support for mapping Pinot
BYTESdata type to Presto
Add support for mapping Pinot time fields with days since epoch value to Presto
DATEtype via the system property
Add support for mapping Pinot time fields with milliseconds since epoch value to Presto
TIMESTAMPtype via the system prpoerty
Add Pinot Field type in to column comment field shown as
DATETIME, to provide more information.
Add support for pushing down distinct count query to Pinot on a best-effort basis.
Add support for new Pinot Routing Table APIs.
Hive Connector Changes#
Fix a bug in Hive split calculation which affects Parquet reader in few corner cases.
Fix ZSTD compression issue with zero row file for missing buckets.
Fix AWS client metric reporting when using S3 select.
Add AWS client retry pause time metrics to
Add table property
preferred_ordering_columnsto support writing sorted files for unbucketed table.
Add native Parquet Writer for Presto.
Add support for impersonation access by using HMS delegation token.
Add support for multi-HMS instances load balancing and breakdown metrics by HMS hosts.
Add support in file status cache to cache all tables. This could be enabled by setting the configuration property
Add configuration property
hive.compression-codecfor ORC and DWRF formats. If specified, ORC and DWRF files are compressed using this codec. RC, Parquet, and other files are compressed using hive.compression-codec.
Druid Connector Changes#
Fix druid connector segment scan.
Fix an issue where distinct is not respected in count aggregation.
Add support for query processing pushdown via the
Fix an issue where resubmitted queries always fail.
Add support for verifying
SELECTqueries that produce columns of
TIMESTAMP WITH TIME ZONE, or
DECIMALtypes, or columns of structured types with those types.
Add support for specifying table properties override for temporary Verifier tables, through configuration property
Add support to output verification results for failures due to Verifier internal errors.
Add support to skip teardown queries in case control and test queries succeeds but verification fails. This can be enabled by configuration property
smart-teardown, which replaces