[ https://issues.apache.org/jira/browse/KYLIN-3284?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Billy Liu resolved KYLIN-3284. ------------------------------ Resolution: Fixed Fix Version/s: v2.4.0 > Refactor all OLAPRel computeSelfCost > ------------------------------------ > > Key: KYLIN-3284 > URL: https://issues.apache.org/jira/browse/KYLIN-3284 > Project: Kylin > Issue Type: Improvement > Components: Query Engine > Affects Versions: v2.3.0 > Reporter: yiming.xu > Assignee: yiming.xu > Priority: Major > Fix For: v2.4.0 > > > for queries like: > select lstg_format_name from test_kylin_fact order by case > when 1=1 then > cal_dt > ELSE > seller_id > end > the expected logical plan after volcano is: > {code:java} > OLAPToEnumerableConverter > OLAPLimitRel(ctx=[], fetch=[50000]) > OLAPSortRel(sort0=[$1], dir0=[ASC], ctx=[]) > OLAPProjectRel(LSTG_FORMAT_NAME=[$3], EXPR$1=[CASE(=(1, 1), $2, > CAST($7):DATE)], 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} > EnumerableSort(sort0=[$1], dir0=[ASC]) > OLAPToEnumerableConverter > OLAPSortRel(sort0=[$1], dir0=[ASC], ctx=[]) > OLAPProjectRel(LSTG_FORMAT_NAME=[$3], EXPR$1=[CASE(=(1, 1), $2, > CAST($7):DATE)], 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} > This is the RelOptPlanner: > {code:java} > Sets: > Set#0, type: RecordType(BIGINT TRANS_ID, BIGINT ORDER_ID, DATE CAL_DT, > VARCHAR(256) LSTG_FORMAT_NAME, BIGINT LEAF_CATEG_ID, INTEGER LSTG_SITE_ID, > SMALLINT SLR_SEGMENT_CD, INTEGER SELLER_ID, DECIMAL(19, 4) PRICE, INTEGER > ITEM_COUNT, VARCHAR(256) TEST_COUNT_DISTINCT_BITMAP, DECIMAL(19, 4) > DEAL_AMOUNT, INTEGER DEAL_YEAR, VARCHAR(256) > LEFTJOIN_BUYER_ID_AND_COUNTRY_NAME, VARCHAR(256) > LEFTJOIN_SELLER_ID_AND_COUNTRY_NAME, VARCHAR(256) > LEFTJOIN_BUYER_COUNTRY_ABBR, VARCHAR(256) LEFTJOIN_SELLER_COUNTRY_ABBR, > BIGINT _KY_COUNT__, INTEGER _KY_SUM_TEST_KYLIN_FACT_ITEM_COUNT_, DECIMAL(19, > 4) _KY_SUM_TEST_KYLIN_FACT_PRICE_, DECIMAL(19, 4) > _KY_MIN_TEST_KYLIN_FACT_PRICE_, DECIMAL(19, 4) > _KY_MAX_TEST_KYLIN_FACT_PRICE_, ANY > _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_SELLER_ID_, ANY > _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_LSTG_FORMAT_NAME_TEST_KYLIN_FACT_SELLER_ID_, > ANY _KY_COUNT_DISTINCT_TEST_KYLIN_FACT_TEST_COUNT_DISTINCT_BITMAP_, > DECIMAL(19, 4) _KY_SUM_TEST_KYLIN_FACT_DEAL_AMOUNT_, ANY > _KY_PERCENTILE_APPROX_TEST_KYLIN_FACT_PRICE_) > rel#140:Subset#0.OLAP.[], best=rel#131, importance=0.7290000000000001 > rel#131:OLAPTableScan.OLAP.[](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]), rowcount=100.0, > cumulative cost={5.0 rows, 5.050000000000001 cpu, 0.0 io} > > rel#179:OLAPSortRel.OLAP.[2](input=rel#140:Subset#0.OLAP.[],sort0=$2,dir0=ASC,ctx=), > rowcount=100.0, cumulative cost={2491.7919004335695 rows, 10.05 cpu, 0.0 io} > rel#169:Subset#0.NONE.[2], best=null, importance=0.7290000000000001 > > rel#167:LogicalSort.NONE.[2](input=rel#140:Subset#0.OLAP.[],sort0=$2,dir0=ASC), > rowcount=100.0, cumulative cost={inf} > rel#171:Subset#0.ENUMERABLE.[2], best=rel#182, importance=0.81 > > rel#181:EnumerableSort.ENUMERABLE.[2](input=rel#177:Subset#0.ENUMERABLE.[],sort0=$2,dir0=ASC), > rowcount=100.0, cumulative cost={1.0E100 rows, 110.05 cpu, 0.0 io} > > rel#182:OLAPToEnumerableConverter.ENUMERABLE.[2](input=rel#180:Subset#0.OLAP.[2]), > rowcount=100.0, cumulative cost={1.0E100 rows, 10.05 cpu, 0.0 io} > rel#173:Subset#0.OLAP.[1], best=null, importance=0.6561 > rel#177:Subset#0.ENUMERABLE.[], best=rel#182, > importance=0.7290000000000001 > > rel#181:EnumerableSort.ENUMERABLE.[2](input=rel#177:Subset#0.ENUMERABLE.[],sort0=$2,dir0=ASC), > rowcount=100.0, cumulative cost={1.0E100 rows, 110.05 cpu, 0.0 io} > > rel#182:OLAPToEnumerableConverter.ENUMERABLE.[2](input=rel#180:Subset#0.OLAP.[2]), > rowcount=100.0, cumulative cost={1.0E100 rows, 10.05 cpu, 0.0 io} > > rel#185:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#140:Subset#0.OLAP.[]), > rowcount=100.0, cumulative cost={1.0E100 rows, 5.050000000000001 cpu, 0.0 io} > rel#180:Subset#0.OLAP.[2], best=rel#179, importance=0.6561 > > rel#179:OLAPSortRel.OLAP.[2](input=rel#140:Subset#0.OLAP.[],sort0=$2,dir0=ASC,ctx=), > rowcount=100.0, cumulative cost={2491.7919004335695 rows, 10.05 cpu, 0.0 io} > Set#1, type: RecordType(VARCHAR(256) LSTG_FORMAT_NAME, DATE EXPR$1) > rel#142:Subset#1.NONE.[], best=null, importance=0.81 > > rel#141:LogicalProject.NONE.[](input=rel#140:Subset#0.OLAP.[],LSTG_FORMAT_NAME=$3,EXPR$1=CASE(=(1, > 1), $2, CAST($7):DATE)), importance=0.0, rowcount=100.0, cumulative > cost={inf} > > rel#143:LogicalSort.NONE.[1](input=rel#142:Subset#1.NONE.[],sort0=$1,dir0=ASC), > rowcount=100.0, cumulative cost={inf} > > rel#163:LogicalProject.NONE.[](input=rel#140:Subset#0.OLAP.[],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={inf} > > rel#170:LogicalProject.NONE.[1](input=rel#169:Subset#0.NONE.[2],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={inf} > rel#149:Subset#1.ENUMERABLE.[], best=rel#157, importance=0.9 > > rel#148:AbstractConverter.ENUMERABLE.[1](input=rel#152:Subset#1.NONE.[1],convention=ENUMERABLE,sort=[1]), > rowcount=100.0, cumulative cost={inf} > > rel#150:EnumerableSort.ENUMERABLE.[1](input=rel#149:Subset#1.ENUMERABLE.[],sort0=$1,dir0=ASC), > rowcount=100.0, cumulative cost={1.0E100 rows, 130.05 cpu, 0.0 io} > > rel#153:AbstractConverter.ENUMERABLE.[1](input=rel#149:Subset#1.ENUMERABLE.[],convention=ENUMERABLE,sort=[1]), > rowcount=100.0, cumulative cost={inf} > > rel#157:OLAPToEnumerableConverter.ENUMERABLE.[1](input=rel#156:Subset#1.OLAP.[1]), > rowcount=100.0, cumulative cost={1.0E100 rows, 30.05 cpu, 0.0 io} > > rel#162:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#154:Subset#1.OLAP.[]), > rowcount=100.0, cumulative cost={1.0E100 rows, 25.05 cpu, 0.0 io} > > rel#172:EnumerableProject.ENUMERABLE.[[1]](input=rel#171:Subset#0.ENUMERABLE.[2],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={1.0E100 rows, 210.05 cpu, 0.0 io} > > rel#178:EnumerableProject.ENUMERABLE.[](input=rel#177:Subset#0.ENUMERABLE.[],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={1.0E100 rows, 210.05 cpu, 0.0 io} > rel#151:Subset#1.ENUMERABLE.[1], best=rel#150, importance=1.0 > > rel#148:AbstractConverter.ENUMERABLE.[1](input=rel#152:Subset#1.NONE.[1],convention=ENUMERABLE,sort=[1]), > rowcount=100.0, cumulative cost={inf} > > rel#150:EnumerableSort.ENUMERABLE.[1](input=rel#149:Subset#1.ENUMERABLE.[],sort0=$1,dir0=ASC), > rowcount=100.0, cumulative cost={1.0E100 rows, 130.05 cpu, 0.0 io} > > rel#153:AbstractConverter.ENUMERABLE.[1](input=rel#149:Subset#1.ENUMERABLE.[],convention=ENUMERABLE,sort=[1]), > rowcount=100.0, cumulative cost={inf} > > rel#157:OLAPToEnumerableConverter.ENUMERABLE.[1](input=rel#156:Subset#1.OLAP.[1]), > rowcount=100.0, cumulative cost={1.0E100 rows, 30.05 cpu, 0.0 io} > > rel#172:EnumerableProject.ENUMERABLE.[[1]](input=rel#171:Subset#0.ENUMERABLE.[2],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={1.0E100 rows, 210.05 cpu, 0.0 io} > rel#152:Subset#1.NONE.[1], best=null, importance=0.9 > > rel#143:LogicalSort.NONE.[1](input=rel#142:Subset#1.NONE.[],sort0=$1,dir0=ASC), > rowcount=100.0, cumulative cost={inf} > > rel#170:LogicalProject.NONE.[1](input=rel#169:Subset#0.NONE.[2],LSTG_FORMAT_NAME=$3,EXPR$1=$2), > rowcount=100.0, cumulative cost={inf} > rel#154:Subset#1.OLAP.[], best=rel#160, importance=0.7290000000000001 > > rel#155:OLAPSortRel.OLAP.[1](input=rel#154:Subset#1.OLAP.[],sort0=$1,dir0=ASC,ctx=), > rowcount=100.0, cumulative cost={199.2068074395237 rows, 30.05 cpu, 0.0 io} > > rel#160:OLAPProjectRel.OLAP.[](input=rel#140:Subset#0.OLAP.[],LSTG_FORMAT_NAME=$3,EXPR$1=CASE(=(1, > 1), $2, CAST($7):DATE),ctx=), rowcount=100.0, cumulative cost={15.0 rows, > 25.05 cpu, 0.0 io} > > rel#164:OLAPProjectRel.OLAP.[](input=rel#140:Subset#0.OLAP.[],LSTG_FORMAT_NAME=$3,EXPR$1=$2,ctx=), > rowcount=100.0, cumulative cost={15.0 rows, 25.05 cpu, 0.0 io} > > rel#174:OLAPProjectRel.OLAP.[1](input=rel#173:Subset#0.OLAP.[1],LSTG_FORMAT_NAME=$3,EXPR$1=$2,ctx=), > rowcount=100.0, cumulative cost={inf} > rel#156:Subset#1.OLAP.[1], best=rel#155, importance=0.81 > > rel#155:OLAPSortRel.OLAP.[1](input=rel#154:Subset#1.OLAP.[],sort0=$1,dir0=ASC,ctx=), > rowcount=100.0, cumulative cost={199.2068074395237 rows, 30.05 cpu, 0.0 io} > > rel#174:OLAPProjectRel.OLAP.[1](input=rel#173:Subset#0.OLAP.[1],LSTG_FORMAT_NAME=$3,EXPR$1=$2,ctx=), > rowcount=100.0, cumulative cost={inf} > {code} > The best root rel is rel#151:Subset#1.ENUMERABLE.[1] : > rel#150:EnumerableSort.ENUMERABLE ,because it cost equal > rel#157:OLAPToEnumerableConverter.ENUMERABLE. The OLAPToEnumerableConverter > cost is so big that the details are neglected.We should refactor the cost > system. -- This message was sent by Atlassian JIRA (v7.6.3#76005)