Drill 1.6.0 - 1606 Release Notes

The following release notes apply to the 1.6.0 - 1606 version of the Apache Drill component included in the MapR Converged Data Platform.
Drill Version 1.6.0
Release Date June 30, 2016
MapR Version Interoperability See Interoperability Matrices and Drill Support Matrix.

Resolved Issues

The following table lists resolved issues in Drill 1.6.0 - 1606:
Issue Description
DRILL-4715 Queries with multiple, non-trivial expressions no longer cause compilation errors at runtime due to memory limits imposed by the JVM.
DRILL-4694 Drill no longer adds extraneous NULL fields to JSON output files when you issue the CTAS statement to create JSON output from an existing JSON file. A new option, store.json.writer.skip_null_fields, controls this behavior. By default, store.json.writer.skip_null_fields = true and skips NULL fields. Setting store.json.writer.skip_null_fields = false reverts the behavior.
DRILL-4693 Columns now appear in the correct order when you issue a SELECT list expression with the COVERT_FROM function and the format type is JSON.
DRILL-4679 Queries that contain the CONVERT_FROM function with the JSON format type no longer fail with the following error when the input data has 0 rows due to a filter condition:
Error: SYSTEM ERROR: IllegalStateException: next() returned NONE without first returning OK_NEW_SCHEMA [#16, ProjectRecordBatch]Fragment 0:0
DRILL-4676 Canceling a query during planning no longer blocks the drillbit from accepting new queries.
DRILL-4657 Running a query with the RANK function no longer produces incorrect results when a frame does not fit into two batches of data.
DRILL-4654 Drill now uses JMX (Java Management Extensions) to monitor queries at runtime and provide the following new metrics through the Drill Web Console and JConsole:
  • pending/running/completed queries

  • current memory usage (root allocator)

DRILL-4647 In instances when a drillbit is not running or query submission fails, the C++ client correctly propagates failed connection errors back to the application.
DRILL-4476 The UNION ALL operator now supports scenarios where the query expression on either side of the UNION ALL operator references an empty source.
DRILL-3714 Queries no longer remain in a CANCELLATION_REQUESTED state until the drillbits are restarted.
MD-850 Queries that contain two numbers CAST as decimals and added together no longer result in a data type mismatch.
MD-849 Queries on JSON files that contain two index levels no longer fail with an IndexOutOfBoundsException error.
MD-815 Queries on INFORMATION_SCHEMA (hive metadata) are significantly faster due to the use of a bulk load algorithm when the exec.enable_bulk_load_table_list system option is set to true:
ALTER SYSTEM SET `exec.enable_bulk_load_table_list` = true