Marcel Kornacker has posted comments on this change. Change subject: IMPALA-3084: Cache the sequence of table ref and materialized tuple ids during analysis. ......................................................................
Patch Set 3: (5 comments) http://gerrit.cloudera.org:8080/#/c/2367/3/fe/src/main/java/com/cloudera/impala/analysis/BinaryPredicate.java File fe/src/main/java/com/cloudera/impala/analysis/BinaryPredicate.java: Line 319: public void swapChildren() { what's that for? http://gerrit.cloudera.org:8080/#/c/2367/3/fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java File fe/src/main/java/com/cloudera/impala/planner/SingleNodePlanner.java: Line 1392: if (analyzer.hasValueTransfer(lhsSid, rhsSid) && we (mostly) use a\n && b in this file, let's stick to that Line 1393: analyzer.hasValueTransfer(rhsSid, lhsSid)) { isn't the lhs->rhs value transfer sufficient? Line 1396: // equalities are redundant this comments seems to contradict the fact that you're doing this for all lhsSlotIds. also, why is this done for all lhsSlotIds? http://gerrit.cloudera.org:8080/#/c/2367/3/testdata/workloads/functional-planner/queries/PlannerTest/join-order.test File testdata/workloads/functional-planner/queries/PlannerTest/join-order.test: Line 799: 09:HASH JOIN [INNER JOIN] > This is a plan improvement because joining with alltypestiny first is bette where does the improvement come from? it doesn't look like the predicates changed (so the card. estimates shouldn't have changed either). -- To view, visit http://gerrit.cloudera.org:8080/2367 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I298b8695c9f26644a395ca9f0e86040e3f5f3846 Gerrit-PatchSet: 3 Gerrit-Project: Impala Gerrit-Branch: cdh5-2.5.0_5.7.0 Gerrit-Owner: Alex Behm <[email protected]> Gerrit-Reviewer: Alex Behm <[email protected]> Gerrit-Reviewer: Marcel Kornacker <[email protected]> Gerrit-HasComments: Yes
