21.98. Release 0.163
- Fix data corruption when transporting dictionary-encoded data.
- Fix potential deadlock when resource groups are configured with memory limits.
- Improve performance for
OUTER JOIN queries.
- Improve exchange performance by reading from buffers in parallel.
- Improve performance when only a subset of the columns resulting from a
JOIN are referenced.
ANY non-reserved keywords.
xxhash64() return type from
- Change subscript operator for map types to fail if the key is not present in the map. The former
NULL) can be restored by setting the
EXPLAIN ANALYZE to render stats more accurately and to include input statistics.
- Improve tolerance to communication errors for long running queries. This can be adjusted
query.remote-task.max-error-duration config option.
- Fix issue that could cause incorrect results for large rows.
- Fix NullPointerException when a field contains a null.
- Add support for
REAL data types.
- Fix issue that would prevent predicates from being pushed into Parquet reader.
- Fix Hive metastore user permissions caching when tables are dropped or renamed.
- Add experimental file based metastore which stores information in HDFS or S3 instead of a database.