impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Matthew Jacobs (Code Review)" <>
Subject [Impala-ASF-CR] IMPALA-5602: Fix kudu queries being incorrectly optimized as small query
Date Thu, 03 Aug 2017 00:30:10 GMT
Matthew Jacobs has posted comments on this change.

Change subject: IMPALA-5602: Fix kudu queries being incorrectly optimized as small query

Patch Set 2:

(1 comment)
File fe/src/main/java/org/apache/impala/planner/

PS2, Line 802:   /**
             :    * Returns true if this node does not contain any conjuncts.
             :    */
             :   public boolean isConjunctsEmpty() {
             :     // Can be overridden by inherited classes to consider all effective
             :     // conjuncts like those pushed to kudu.
             :     return getConjuncts().isEmpty();
             :   }
This could be confusing because it's not obvious why someone would know to use this vs getConjuncts().isEmpty().

I think it'll be better to further differentiate by instead exposing a separate term in the
interface, e.g. 'effective conjuncts' as we discussed. I think a clean interface will be

public List<Expr> getEffectiveScanConjuncts() ...

Which can be put in the ScanNode (there's no reason this needs to be on PlanNode).

Yes, the code in MaxRowsProcessedVisitor has to then call .isEmpty(), but that's fine and
I think exposing the conjuncts in this way makes sense.

So to summarize:
1) move to ScanNode
2) change to getEffectiveScanConjuncts()

To view, visit
To unsubscribe, visit

Gerrit-MessageType: comment
Gerrit-Change-Id: I93822d67ebda41d5d0456095c429e3915a3f40c4
Gerrit-PatchSet: 2
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Bikramjeet Vig <>
Gerrit-Reviewer: Bikramjeet Vig <>
Gerrit-Reviewer: Matthew Jacobs <>
Gerrit-Reviewer: Tim Armstrong <>
Gerrit-HasComments: Yes

View raw message