- Fix runtime failure for queries that use grouping sets over unions.
- Do not ignore null values in
array_agg
. - Fix failure when casting row values that contain null fields.
- Fix failure when using complex types as map keys.
- Fix potential memory tracking leak when queries are cancelled.
- Fix rejection of queries that do not match any queue/resource group rules. Previously, a 500 error was returned to the client.
- Fix
trim
andrtrim
functions to produce more intuitive results when the argument contains invalidUTF-8
sequences. - Add a new web interface with cluster overview, realtime stats, and improved sorting and filtering of queries.
- Add support for
FLOAT
type. - Rename
query.max-age
toquery.min-expire-age
. optimizer.columnar-processing
andoptimizer.columnar-processing-dictionary
properties were merged tooptimizer.processing-optimization
with possible valuesdisabled
,columnar
andcolumnar_dictionary
columnar_processing
andcolumnar_processing_dictionary
session properties were merged toprocessing_optimization
with possible valuesdisabled
,columnar
andcolumnar_dictionary
- Change
%y
(2-digit year) indate_parse
to evaluate to a year between 1970 and 2069 inclusive. - Add
queued
flag toStatementStats
in REST API. - Improve error messages for math operations.
- Improve memory tracking in exchanges to avoid running out of Java heap space.
- Improve performance of subscript operator for the
MAP
type. - Improve performance of
JOIN
andGROUP BY
queries.
- Clean up empty staging directories after inserts.
- Add
hive.dfs.ipc-ping-interval
config for HDFS. - Change default value of
hive.dfs-timeout
to 60 seconds. - Fix ORC/DWRF reader to avoid repeatedly fetching the same data when stripes are skipped.
- Fix force local scheduling for S3 or other non-HDFS file systems.