[jira] [Created] (CALCITE-5506) RelToSqlConverter get error result because RelFieldTrimmer lost aggregate function

["Jiajun Xie (Jira)" <ji...@apache.org>]

Jiajun Xie created CALCITE-5506:
-----------------------------------

             Summary: RelToSqlConverter get error result because RelFieldTrimmer lost aggregate function
                 Key: CALCITE-5506
                 URL: https://issues.apache.org/jira/browse/CALCITE-5506
             Project: Calcite
          Issue Type: Improvement
          Components: core
            Reporter: Jiajun Xie


Here is a sql

 
{code:java}
select  l.v as l_cost
from    (
            select  0 as v,
                    1 as k
        ) l
join    (
    select  sum("cost") as v,
                    1 as k
    from    (
        select 1 as "cost"
        union
        select 2 as "cost"
        )
) r
on l.k = r.k {code}
Before trimming, the RelNode is 

 

 
{code:java}
LogicalProject(L_COST=[$0])
  LogicalJoin(condition=[=($1, $3)], joinType=[inner])
    LogicalValues(tuples=[[{ 0, 1 }]])
    LogicalProject(V=[$0], K=[1])
      LogicalAggregate(group=[{}], V=[SUM($0)]) <!-- Here is SUM() -->
         LogicalUnion(all=[false])
          LogicalValues(tuples=[[{ 1 }]])
          LogicalValues(tuples=[[{ 2 }]]) {code}
After trimming, the RelNode is

 

 
{code:java}
LogicalProject(L_COST=[$0])
  LogicalJoin(condition=[=($1, $2)], joinType=[inner])
    LogicalValues(tuples=[[{ 0, 1 }]])
    LogicalProject(K=[1]) 
      LogicalAggregate(group=[{}], DUMMY=[COUNT()]) <!-- Missing SUM() -->
        LogicalUnion(all=[false])
          LogicalValues(tuples=[[{ 1 }]])
          LogicalValues(tuples=[[{ 2 }]]){code}
 

If we convert trimmed RelNode to sql, the sql will be
{code:java}
SELECT *
FROM 
  (VALUES (0, 1)) AS "t" ("V", "K")
INNER JOIN 
  (SELECT 1 AS "K" -- Missing SUM()
   FROM (SELECT *
         FROM (VALUES (1)) AS "t" ("cost")
         UNION
         SELECT *
         FROM (VALUES (2)) AS "t" ("cost")) AS "t2"
  ) AS "t4" 
ON "t"."K" = "t4"."K" {code}
The origin sql only has one row result, but the new sql that be trimmed has two row result.

 



--
This message was sent by Atlassian Jira
(v8.20.10#820010)