[DRILL-5829] RecordBatchLoader schema comparison is not case insensitive

• Type: Bug
• Status: In Progress
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.11.0
• Assignee: Paul Rogers
• Reporter: Paul Rogers
• Created: 29/Sep/17 23:55
• Updated: 12/Jun/18 22:16

Description

The class RecordBatchLoader decodes batches received over the wire. It determines if the schema of the new batch matches that of the old one. To do that, it uses a map of existing columns.

In Drill, column names follow SQL rules: they are case insensitive. Yet, the implementation of RecordBatchLoader uses a case sensitive map:

    final Map<String, ValueVector> oldFields = Maps.newHashMap();

This should be:

    final Map<String, ValueVector> oldFields = CaseInsensitiveMap.newHashMap();

Without this change, the receivers will report schema changes if a column differs only in name case. However, Drill semantics say that names that differ in case are identical, and so no schema change should be issued.

Issue Links

• Is contained by: DRILL-5830 Resolve regressions to MapR DB from DRILL-5546