Release 0.149#
General#
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.
Hive#
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.