[
https://issues.apache.org/jira/browse/IGNITE-24707?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Orlov updated IGNITE-24707:
--------------------------------------
Fix Version/s: 3.1
> Sql. Fix type coercion for quantify operators
> ---------------------------------------------
>
> Key: IGNITE-24707
> URL: https://issues.apache.org/jira/browse/IGNITE-24707
> Project: Ignite
> Issue Type: Bug
> Components: sql
> Reporter: Konstantin Orlov
> Assignee: Konstantin Orlov
> Priority: Major
> Labels: ignite-3
> Fix For: 3.1
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Currently, type coercion is not applied to quantify operators (ALL, SOME),
> resulting in a wrong plan.
> For example, let's take a look at output for following queries:
> {code}
> CREATE TABLE test (id INT PRIMARY KEY, int_c INT, byte_c TINYINT);
> EXPLAIN PLAN FOR SELECT byte_c = SOME(SELECT int_c FROM test) FROM test;
> ===============
> Project
> fields: [EXPR$0]
> exprs: [OR(AND(IS NULL(BYTE_C), null, <>(c, 0)), AND(IS NOT NULL(i),
> <>(c, 0), IS NOT NULL(BYTE_C)), AND(<(ck, c), null, <>(c, 0), IS NOT
> NULL(BYTE_C), IS NULL(i)))]
> est. row count: 3008500
> HashJoin
> condition: =(BYTE_C, INT_C)
> joinType: left
> est. row count: 3008500
> NestedLoopJoin
> condition: true
> joinType: inner
> est. row count: 10000
> Exchange
> distribution: single
> est. row count: 10000
> TableScan
> table: [PUBLIC, TEST]
> fields: [BYTE_C]
> est. row count: 10000
> Project
> fields: [c, ck]
> exprs: [CAST(f0_0):BIGINT NOT NULL, CAST(f0_1):BIGINT NOT NULL]
> est. row count: 1
> ReduceSortAggregate
> rowType: RecordType(BIGINT f0_0, BIGINT f0_1)
> group: []
> aggs: [$SUM0(_ACC0), $SUM0(_ACC1)]
> collation: []
> est. row count: 1
> Exchange
> distribution: single
> est. row count: 1
> MapSortAggregate
> group: []
> aggs: [COUNT(), COUNT(INT_C)]
> collation: []
> est. row count: 1
> TableScan
> table: [PUBLIC, TEST]
> fields: [INT_C]
> est. row count: 10000
> ColocatedHashAggregate
> group: [INT_C]
> aggs: [LITERAL_AGG(true)]
> est. row count: 2000
> Exchange
> distribution: single
> est. row count: 10000
> TableScan
> table: [PUBLIC, TEST]
> fields: [INT_C]
> est. row count: 10000
> {code}
> The problem is missed cast of {{byte_c}} to {{INTEGER}} before {{HashJoin}}
> operation resulting in a wrong outcome.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
