20.104. Release 0.149

General Changes

  • 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() and rtrim() functions to produce more intuitive results when the argument contains invalid UTF-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 to query.min-expire-age.

  • optimizer.columnar-processing and optimizer.columnar-processing-dictionary properties were merged to optimizer.processing-optimization with possible values disabled, columnar and columnar_dictionary

  • columnar_processing and columnar_processing_dictionary session properties were merged to processing_optimization with possible values disabled, columnar and columnar_dictionary

  • Change %y (2-digit year) in date_parse() to evaluate to a year between 1970 and 2069 inclusive.

  • Add queued flag to StatementStats 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 and GROUP BY queries.

Hive Changes

  • 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.