[DRILL-4505] Can't group by or sort across files with different schema

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.5.0
• Component/s: Storage - Parquet

Environment

Java 1.8

• Assignee: Unassigned
• Reporter: Tobias
• Created: 14/Mar/16 03:59
• Updated: 18/Mar/19 16:52

Description

We are currently trying out the support for querying across parquet files with different schemas.
Simple selects work well but when we wan't to do sort or group by Drill returns "UNSUPPORTED_OPERATION ERROR: Sort doesn't currently support sorts with changing schemas Fragment 0:0 [Error Id: ff490670-64c1-4fb8-990e-a02aa44ac010 on zookeeper-1:31010]"

This is despite not even including the new columns in the query.
Expected result would be to treat the non existing columns in certain files as either null or default value and allow them to be grouped and sorted

Example
SELECT APPLICATION_ID ,dir0 AS year_ FROM dfs.`/PRO/UTC/1` WHERE dir2 >='2016-01-01' AND dir2<'2016-04-02' work with changing schema

but SELECT max(APPLICATION_ID ),dir0 AS year_ FROM dfs.`/PRO/UTC/1` WHERE dir2 >='2016-01-01' AND dir2<'2016-04-02'  group by dir0 does not work

For us this hampers any possibility to have an evolving schema with moderatly complex queries

Issue Links

• relates to: DRILL-3577 Counting nested fields on CTAS-created-parquet file/s reports inaccurate results
• relates to: DRILL-4538 Turn on Union type by default