[
https://issues.apache.org/jira/browse/DRILL-388?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13916264#comment-13916264
]
Julian Hyde commented on DRILL-388:
-----------------------------------
It might be that some code is not handling [RelCollationImpl.PRESERVE]
correctly. It only occurs rarely (and always very near the root of the tree).
It means 'this relational expression is sorted by a column that has been
projected away already. since you can't re-create the order by sorting, every
operation you do must preserve the ordering'.
> Doing an ORDER-BY on a column not in the SELECT list gives CannotPlanException
> ------------------------------------------------------------------------------
>
> Key: DRILL-388
> URL: https://issues.apache.org/jira/browse/DRILL-388
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Aman Sinha
>
> Queries of the following type where order-by column is not in SELECT list
> give a CannotPlanException. A partial stack trace is shown below.
> 0: jdbc:drill:schema=parquet-local> select _MAP['N_NATIONKEY'] as n_nationkey
> from "/Users/asinha/parquet/nation.parquet" order by _MAP['N_REGIONKEY'];
> org.eigenbase.relopt.RelOptPlanner$CannotPlanException: Node
> [rel#10788:Subset#3.ENUMERABLE.PRESERVE] could not be implemented; planner
> state:
> Root: rel#10788:Subset#3.ENUMERABLE.PRESERVE
> Original rel:
> AbstractConverter(subset=[rel#10788:Subset#3.ENUMERABLE.PRESERVE],
> convention=[ENUMERABLE], sort=[PRESERVE])
> ProjectRel(subset=[rel#10784:Subset#3.NONE.PRESERVE], N_NATIONKEY=[$0])
> SortRel(subset=[rel#10782:Subset#2.NONE.[1 Ascending]], sort0=[$1],
> dir0=[Ascending])
> ProjectRel(subset=[rel#10780:Subset#1.NONE.[]], N_NATIONKEY=[ITEM($0,
> 'N_NATIONKEY')], EXPR$1=[ITEM($0, 'N_REGIONKEY')])
> DrillScan(subset=[rel#10778:Subset#0.DRILL.[]],
> table=[[parquet-local, /Users/asinha/parquet/nation.parquet]])
> Sets:
> Set#0, type: RecordType((VARCHAR, ANY) MAP _MAP)
> rel#10778:Subset#0.DRILL.[], best=rel#10774, importance=0.6561
> rel#10774:DrillScan.DRILL.[](table=[parquet-local,
> /Users/asinha/parquet/nation.parquet]), rowcount=100.0, cumulative
> cost={100.0 rows, 101.0 cpu, 0.0 io}
>
> rel#12031:AbstractConverter.DRILL.[](child=rel#12030:Subset#0.ENUMERABLE.[],convention=DRILL,sort=[]),
> rowcount=100.0, cumulative cost={inf}
> rel#12030:Subset#0.ENUMERABLE.[], best=rel#12029,
> importance=0.5904900000000001
>
> rel#12032:AbstractConverter.ENUMERABLE.[](child=rel#10778:Subset#0.DRILL.[],convention=ENUMERABLE,sort=[]),
> rowcount=100.0, cumulative cost={inf}
>
> rel#12029:EnumerableDrillRel.ENUMERABLE.[](child=rel#10778:Subset#0.DRILL.[]),
> rowcount=100.0, cumulative cost={110.0 rows, 111.0 cpu, 0.0 io}
> Set#1, type: RecordType(ANY N_NATIONKEY, ANY EXPR$1)
> rel#10780:Subset#1.NONE.[], best=null, importance=0.7290000000000001
>
> rel#10779:ProjectRel.NONE.[](child=rel#10778:Subset#0.DRILL.[],N_NATIONKEY=ITEM($0,
> 'N_NATIONKEY'),EXPR$1=ITEM($0, 'N_REGIONKEY')), rowcount=100.0, cumulative
> cost={inf}
>
> rel#10791:CalcRel.NONE.[](child=rel#10778:Subset#0.DRILL.[],expr#0={inputs},expr#1='N_NATIONKEY',expr#2=ITEM($t0,
> $t1),expr#3='N_REGIONKEY',expr#4=ITEM($t0, $t3),N_NATIONKEY=$t2,EXPR$1=$t4),
> rowcount=100.0, cumulative cost={inf}
> rel#11466:SortRel.NONE.[](child=rel#10833:Subset#1.DRILL.[1
> Ascending]), rowcount=100.0, cumulative cost={inf}
>
> rel#11597:SortRel.NONE.[](child=rel#10828:Subset#1.DRILL.PRESERVE),
> rowcount=1.7976931348623157E308, cumulative cost={inf}
>
> rel#11701:SortRel.NONE.[](child=rel#11115:Subset#1.ENUMERABLE.PRESERVE),
> rowcount=1.7976931348623157E308, cumulative cost={inf}
>
> rel#11803:SortRel.NONE.[](child=rel#11082:Subset#1.ENUMERABLE.[1 Ascending]),
> rowcount=100.0, cumulative cost={inf}
> rel#10804:Subset#1.DRILL.[], best=rel#10821,
> importance=0.5904900000000001
>
> rel#10805:AbstractConverter.DRILL.[](child=rel#10780:Subset#1.NONE.[],convention=DRILL,sort=[]),
> rowcount=1.7976931348623157E308, cumulative cost={inf}
>
> rel#10821:DrillProjectRel.DRILL.[](child=rel#10778:Subset#0.DRILL.[],N_NATIONKEY=ITEM($0,
> 'N_NATIONKEY'),EXPR$1=ITEM($0, 'N_REGIONKEY')), rowcount=100.0, cumulative
> cost={110.0 rows, 121.0 cpu, 0.0 io}
>
> rel#10825:AbstractConverter.DRILL.[](child=rel#10824:Subset#1.NONE.[1
> Ascending],convention=DRILL,sort=[]), rowcount=1.7976931348623157E308,
> cumulative cost={inf}
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
