Github user sureshsubbiah commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/66#discussion_r38109423
--- Diff: core/sql/generator/GenPreCode.cpp ---
@@ -154,15 +154,19 @@ static NABoolean processConstHBaseKeys(Generator *
generator,
// convert built-in search key to entries with constants, if possible
if (skey->areAllKeysConstants(TRUE))
{
- ValueIdSet exePreds;
ValueIdSet nonKeyColumnSet;
idesc->getNonKeyColumnSet(nonKeyColumnSet);
- // seed exePreds with all predicates, key and non-key
- // they will be reduced by the method below
- skey->getKeyPredicates(exePreds);
+ // seed keyPreds with only the full key predicate from skey
+ ValueIdSet keyPreds = skey->getFullKeyPredicates();
+
+ // include executorPreds and selection predicates
+ // but exclude the full key predicates.
+ ValueIdSet exePreds;
+
exePreds += executorPreds;
exePreds += relExpr->getSelectionPred();
+ exePreds.subtractSet(keyPreds);
--- End diff --
I am wondering if this is a good thing to do. There are instances where we
want a predicate to be both an executor predicate as well as a key predicate.
In certain MDAM disjuncts the key predicate guarantees that a minimal set of
rows is scanned but an exe pred has to be evaluated on the same pred again to
guarantee correct results. I realize that the comment is vague without an
example or two. I can try to find one if necessary. Also since this PreCodeGen
a predicate can be physically implemented differently based on what is
available. That may also require us to keep things in the exepred?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
