[
https://issues.apache.org/jira/browse/KYLIN-2777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16542436#comment-16542436
]
翟娜 commented on KYLIN-2777:
---------------------------
SQL like:
select sum(v_revenue)
from ssb.p_lineorder left join ssb.dates on lo_orderdate = d_datekey
and d_year = 1993
with the sql statements above, I got the following error message.
*version 2.2.X:* Error while applying rule OLAPJoinRule, args
[rel#85:LogicalJoin.NONE.[](left=rel#83:Subset#0.OLAP.[],right=rel#84:Subset#1.OLAP.[],condition=AND(=($2,
$5), =($6, 1993)),joinType=left)] while executing SQL: "select sum(v_revenue)
from ssb.p_lineorder left join ssb.dates on lo_orderdate = d_datekey and d_year
= 1993 LIMIT 50000"
*version 2.3.X:* 2 while executing SQL: "select sum(v_revenue) from
ssb.p_lineorder left join ssb.dates on lo_orderdate = d_datekey and d_year =
1993"
*version 2.4.X:* Node [rel#418:Subset#4.ENUMERABLE.[]] could not be
implemented; planner state: Root: rel#418:Subset#4.ENUMERABLE.[] Original rel:
LogicalAggregate(subset=[rel#418:Subset#4.ENUMERABLE.[]], group=[{}],
EXPR$0=[SUM($0)]): rowcount = 22.5, cumulative cost = \{25.593751072883606
rows, 0.0 cpu, 0.0 io}, id = 413
LogicalProject(subset=[rel#412:Subset#3.NONE.[]], V_REVENUE=[$3]): rowcount =
225.0, cumulative cost = \{225.0 rows, 225.0 cpu, 0.0 io}, id = 411
LogicalJoin(subset=[rel#410:Subset#2.NONE.[]], condition=[AND(=($2, $6), =($7,
1993))], joinType=[left]): rowcount = 225.0, cumulative cost = \{225.0 rows,
0.0 cpu, 0.0 io}, id = 409 OLAPTableScan(subset=[rel#407:Subset#0.OLAP.[]],
table=[[SSB, P_LINEORDER]], ctx=[], fields=[[0, 1, 2, 3, 4, 5]]): rowcount =
100.0, cumulative cost = \{5.0 rows, 5.050000000000001 cpu, 0.0 io}, id = 391
OLAPTableScan(subset=[rel#408:Subset#1.OLAP.[]], table=[[SSB, DATES]], ctx=[],
fields=[[0, 1]]): rowcount = 100.0, cumulative cost = \{5.0 rows,
5.050000000000001 cpu, 0.0 io}, id = 392 Sets: Set#0, type: RecordType(BIGINT
LO_ORDERKEY, INTEGER LO_CUSTKEY, INTEGER LO_ORDERDATE, BIGINT V_REVENUE, BIGINT
_KY_COUNT__, BIGINT _KY_SUM_P_LINEORDER_V_REVENUE_) rel#407:Subset#0.OLAP.[],
best=rel#391, importance=0.6561 rel#391:OLAPTableScan.OLAP.[](table=[SSB,
P_LINEORDER],ctx=,fields=[0, 1, 2, 3, 4, 5]), rowcount=100.0, cumulative
cost=\{5.0 rows, 5.050000000000001 cpu, 0.0 io} rel#446:Subset#0.ENUMERABLE.[],
best=rel#445, importance=0.32805
rel#445:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#407:Subset#0.OLAP.[]),
rowcount=100.0, cumulative cost=\{10.0 rows, 10.05 cpu, 0.0 io} Set#1, type:
RecordType(INTEGER D_DATEKEY, INTEGER D_YEAR) rel#408:Subset#1.OLAP.[],
best=rel#392, importance=0.6561 rel#392:OLAPTableScan.OLAP.[](table=[SSB,
DATES],ctx=,fields=[0, 1]), rowcount=100.0, cumulative cost=\{5.0 rows,
5.050000000000001 cpu, 0.0 io} rel#444:Subset#1.ENUMERABLE.[], best=rel#443,
importance=0.32805
rel#443:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#408:Subset#1.OLAP.[]),
rowcount=100.0, cumulative cost=\{10.0 rows, 10.05 cpu, 0.0 io} Set#2, type:
RecordType(BIGINT LO_ORDERKEY, INTEGER LO_CUSTKEY, INTEGER LO_ORDERDATE, BIGINT
V_REVENUE, BIGINT _KY_COUNT__, BIGINT _KY_SUM_P_LINEORDER_V_REVENUE_, INTEGER
D_DATEKEY, INTEGER D_YEAR) rel#410:Subset#2.NONE.[], best=null,
importance=0.7290000000000001
rel#409:LogicalJoin.NONE.[](left=rel#407:Subset#0.OLAP.[],right=rel#408:Subset#1.OLAP.[],condition=AND(=($2,
$6), =($7, 1993)),joinType=left), rowcount=225.0, cumulative cost=\{inf}
rel#434:Subset#2.OLAP.[], best=null, importance=0.7290000000000001 Set#3, type:
RecordType(BIGINT V_REVENUE) rel#412:Subset#3.NONE.[], best=null,
importance=0.81
rel#411:LogicalProject.NONE.[](input=rel#410:Subset#2.NONE.[],V_REVENUE=$3),
rowcount=225.0, cumulative cost=\{inf} rel#420:Subset#3.OLAP.[], best=null,
importance=0.81
rel#435:OLAPProjectRel.OLAP.[](input=rel#434:Subset#2.OLAP.[],V_REVENUE=$3,ctx=),
rowcount=225.0, cumulative cost=\{inf} rel#439:Subset#3.ENUMERABLE.[],
best=null, importance=0.405
rel#438:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#420:Subset#3.OLAP.[]),
rowcount=225.0, cumulative cost=\{inf} Set#4, type: RecordType(BIGINT EXPR$0)
rel#414:Subset#4.NONE.[], best=null, importance=0.9
rel#413:LogicalAggregate.NONE.[](input=rel#412:Subset#3.NONE.[],group={},EXPR$0=SUM($0)),
rowcount=22.5, cumulative cost=\{inf}
rel#429:LogicalProject.NONE.[](input=rel#428:Subset#5.NONE.[],EXPR$0=CASE(=($1,
0), null, $0)), rowcount=22.5, cumulative cost=\{inf}
rel#418:Subset#4.ENUMERABLE.[], best=null, importance=1.0
rel#419:AbstractConverter.ENUMERABLE.[](input=rel#414:Subset#4.NONE.[],convention=ENUMERABLE,sort=[]),
rowcount=22.5, cumulative cost=\{inf}
rel#423:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#422:Subset#4.OLAP.[]),
rowcount=22.5, cumulative cost=\{inf} rel#422:Subset#4.OLAP.[], best=null,
importance=0.9
rel#421:OLAPAggregateRel.OLAP.[](input=rel#420:Subset#3.OLAP.[],group={},EXPR$0=SUM($0),ctx=),
rowcount=22.5, cumulative cost=\{inf}
rel#431:OLAPProjectRel.OLAP.[](input=rel#430:Subset#5.OLAP.[],EXPR$0=CASE(=($1,
0), null, $0),ctx=), rowcount=22.5, cumulative cost=\{inf} Set#5, type:
RecordType(BIGINT EXPR$0, BIGINT $f1) rel#428:Subset#5.NONE.[], best=null,
importance=0.81
rel#426:LogicalAggregate.NONE.[](input=rel#412:Subset#3.NONE.[],group={},EXPR$0=$SUM0($0),agg#1=COUNT($0)),
rowcount=22.5, cumulative cost=\{inf} rel#430:Subset#5.OLAP.[], best=null,
importance=0.81
rel#436:OLAPAggregateRel.OLAP.[](input=rel#420:Subset#3.OLAP.[],group={},EXPR$0=$SUM0($0),agg#1=COUNT($0),ctx=),
rowcount=22.5, cumulative cost=\{inf} rel#442:Subset#5.ENUMERABLE.[],
best=null, importance=0.405
rel#441:OLAPToEnumerableConverter.ENUMERABLE.[](input=rel#430:Subset#5.OLAP.[]),
rowcount=22.5, cumulative cost=\{inf} while executing SQL: "select
sum(v_revenue) from ssb.p_lineorder left join ssb.dates on lo_orderdate =
d_datekey and d_year = 1993"
> Should throw error when sql contains non-equi left join
> -------------------------------------------------------
>
> Key: KYLIN-2777
> URL: https://issues.apache.org/jira/browse/KYLIN-2777
> Project: Kylin
> Issue Type: Bug
> Reporter: Roger Shi
> Priority: Major
>
> SQL like:
> select fact_table.A, lookup_table.B
> from fact_table fa left join lookup_table lo
> on fa.C = lo.D and fa.E = 'asdf'
> The join is non-equi left join. It's not support in Kylin data model, should
> throw error instead of ignore it. Current implementation leads to wrong
> result instead of clear error message.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
