[IMPALA-10890] Show pushed down predicates in query plan

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Frontend
• Labels: observability
• Epic Color: ghx-label-10
• Assignee: Unassigned
• Reporter: Quanlong Huang
• Created: 27/Aug/21 01:25
• Updated: 27/Aug/21 01:26

Description

We push down predicates that can leverage parquet statistics. Similar works for ORC are on-going in IMPALA-6505 and IMPALA-10873. It'd be useful to show the final pushed down form of these predicates. E.g.

• predicate 'x = 5' is transformed into two predicates 'x <= 5' and 'x >= 5' for parquet.
• predicate 'x in [0, 50, 100]' is transformed into two predicates 'x >= 0' and 'x <= 100' for parquet.

The query plan only shows the original predicates in the line of "parquet statistics predicates:". It helps to understand the performance if the final predicates are also shown.

For ORC columns that have bloom filter enabled, we'd like to push down the IN-list predicate directly instead of transformed it into range predicates. Showing the final predicates make this visible.

Issue Links

• relates to: IMPALA-6505 Min-Max predicate push down in ORC scanner
• relates to: IMPALA-10873 Push down EQUALS, IS NULL and IN-list predicate to ORC reader