You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@drill.apache.org by "Paul Rogers (JIRA)" <ji...@apache.org> on 2018/04/01 00:59:00 UTC
[jira] [Created] (DRILL-6304) In hyper-batch row set reader, handle
disjoint unions
Paul Rogers created DRILL-6304:
----------------------------------
Summary: In hyper-batch row set reader, handle disjoint unions
Key: DRILL-6304
URL: https://issues.apache.org/jira/browse/DRILL-6304
Project: Apache Drill
Issue Type: Improvement
Affects Versions: 1.14.0
Reporter: Paul Rogers
See {{exec/java-exec/src/test/java/org/apache/drill/test/rowSet/HyperRowSetImpl.java}}, in PR #1161 (or after that PR is committed), line 101.
The code in question implements a mechanism for managing hyper vectors within the row set reader mechanism.
A very special case occurs in a hyper vector over a union vector, and that union has disjoint types. (Say an INT in batch 1, and a VARCHAR in batch 2.) The desired behavior is that the resulting column metadata holds the union of all types in the component vectors, and that the accessor reports values as null if the union reader asks for a value of a type that is no present.
This is an obscure case, and unions are not supposed within Drill completely, so this case was not implemented or tested.
This JIRA serves as a reminder that this case should be handled if Drill decides to fully support the union type.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)