[HAWQ-1404] PXF to leverage file-level stats of ORC file and emit records for COUNT(*)

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.3.0.0-incubating
• Component/s: PXF
• Assignee: Oleksandr Diachenko
• Reporter: Oleksandr Diachenko
• Created: 24/Mar/17 02:10
• Updated: 04/Apr/17 19:06
• Resolved: 04/Apr/17 19:06

Description

For cases when user issues COUNT queries without WHERE clause PXF should be able to leverage file-level stats for a ORC file and emit given number of records back to HAWQ, avoiding reading actual tuples from disk. This should be a first step in enabling PXF to use ORC stats(file, stripe and row group levels) so we can improve a wider range of aggregate queries.

So whenever PXF receives "count" as AGG-TYPE parameters value - it should optimize it by emitting tuples using ORC file-level stats.

Issue Links

• links to: GitHub Pull Request #1209

Sub-Tasks

• Do not use StatsAccessor when column specified in SELECT clause: Resolved
• Do not use AggBridge when WHERE clause specified: Resolved
• HAWQ send additional header to PXF to indicate aggregate function type: Resolved