[DRILL-7735] Query against empty parquet file fails with: IndexOutOfBoundsException: Index: 0, Size: 0

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.17.0
• Component/s: Server, Storage - Parquet

Environment

64Gb machine running on AWS.

• Assignee: Unassigned
• Reporter: Dave Challis
• Created: 06/May/20 08:40
• Updated: 06/May/20 08:40

Description

Running a `SELECT *` query against an empty Parquet file (i.e. one with correct column metadata written, but no rows) triggers an `IndexOutOfBoundsException`.

I've got an empty parquet file with the following schema:

$ parquet-tools schema dispute.parquet
message parquet_go_root {
  required int32 dispute_id (INT_32) = 0;
  required binary title (UTF8) = 0;
  optional int32 start_date (DATE) = 0;
  optional int32 end_date (DATE) = 0;
  optional binary docket_number (UTF8) = 0;
  required binary route (UTF8) = 0;
  required binary jurisdiction (UTF8) = 0;
}

If I then run the following query via the Drill web UI:

SELECT * FROM dfs.`/data/dispute.parquet`

then I get the following error from Drill:

org.apache.drill.common.exceptions.UserRemoteException: SYSTEM ERROR: IndexOutOfBoundsException: Index: 0, Size: 0 Please, refer to logs for more information. 

[Error Id: a93e1aa1-a7e6-4bc9-9f11-c42b9f6fe108 on e531a6492cf4:31010]

Expected result was just to get an empty result set (i.e. 0 rows).

 

I've attached the parquet file in question, and the relevant entries from the drillbit.log.