[ https://issues.apache.org/jira/browse/KYLIN-3281?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Billy Liu resolved KYLIN-3281. ------------------------------ Resolution: Fixed Fix Version/s: v2.4.0 > OLAPProjectRule can't normal working with projectRel[input=sortRel] > -------------------------------------------------------------------- > > Key: KYLIN-3281 > URL: https://issues.apache.org/jira/browse/KYLIN-3281 > Project: Kylin > Issue Type: Bug > Components: Query Engine > Affects Versions: v2.3.0 > Reporter: yiming.xu > Assignee: yiming.xu > Priority: Critical > Fix For: v2.4.0 > > > for queries like: > select sum("count") from (SELECT LSTG_FORMAT_NAME , COUNT(*) AS "count" FROM > TEST_KYLIN_FACT GROUP BY "LSTG_FORMAT_NAME" ORDER BY "count" DESC) as a > the expected logical plan after volcano is: > {code:java} > OLAPToEnumerableConverter > OLAPLimitRel(ctx=[], fetch=[50000]) > OLAPAggregateRel(group=[{}], EXPR$0=[SUM($0)], ctx=[]) > OLAPProjectRel(count=[$1], ctx=[]) > OLAPSortRel(sort0=[$1], dir0=[DESC], ctx=[]) > OLAPAggregateRel(group=[{0}], count=[COUNT()], ctx=[]) > OLAPProjectRel(LSTG_FORMAT_NAME=[$3], ctx=[]) > OLAPTableScan(table=[[DEFAULT, TEST_KYLIN_FACT]], ctx=[], > fields=[[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, > 19, 20, 21, 22, 23, 24, 25, 26]]) > {code} > however the actual is: > {code:java} > EnumerableLimit(fetch=[50000]) > EnumerableAggregate(group=[{}], EXPR$0=[SUM($0)]) > EnumerableCalc(expr#0..1=[{inputs}], count=[$t1]) > EnumerableSort(sort0=[$1], dir0=[DESC]) > EnumerableAggregate(group=[{0}], count=[COUNT()]) > OLAPToEnumerableConverter > OLAPProjectRel(LSTG_FORMAT_NAME=[$3], ctx=[]) > OLAPTableScan(table=[[DEFAULT, TEST_KYLIN_FACT]], ctx=[], > fields=[[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, > 19, 20, 21, 22, 23, 24, 25, 26]]) > {code} > VolcanoPlanner apply OLAPProjectRule to logicalProject,convert child node > should use child traitset not current proejct traitset. > If we use current project trait set, the child node trait set is [olap][0 > DESC], it will create a new subset but the subset doesn't have any rel. the > cost is inf, > so volcanoPlanner choose a the EnumerableAggregate. -- This message was sent by Atlassian JIRA (v7.6.3#76005)