[DRILL-6528] Planner setting the wrong number of records to read (Parquet Reader)

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.17.0
• Component/s: Query Planning & Optimization
• Assignee: Arina Ielchiieva
• Reporter: Salim Achouche
• Created: 22/Jun/18 18:09
• Updated: 13/Aug/19 14:02
• Resolved: 13/Aug/19 14:02

Description

• Recently fixed the Flat Parquet reader to honor the number of records to read
• Though few tests failed:
  TestUnionDistinct.testUnionDistinctEmptySides:356 Different number of records returned expected:<5> but was:<1>
  TestUnionAll.testUnionAllEmptySides:355 Different number of records returned expected:<5> but was:<1>

• I debugged one of them and realized the Planner was setting the wrong number of rows to read (in this case, one)
• You can put a break point and see this happening:
  Class: ParquetGroupScan
  Method: updateRowGroupInfo(long maxRecords)

Issue Links

• is caused by: DRILL-5846 Improve Parquet Reader Performance for Flat Data types
• is part of: DRILL-4517 Reading emtpy Parquet file failes with java.lang.IllegalArgumentException