[jira] [Commented] (HIVE-22210) Incorrect results in some cases with vectorization
[
https://issues.apache.org/jira/browse/HIVE-22210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16931273#comment-16931273
]
Zoltan Haindrich commented on HIVE-22210:
-
I suppose IdentityExpression's output may not be freed; as its directly relays
the column...there were already a tweak for this:
https://github.com/apache/hive/blob/62f0c3fe89d7112d2cbc5c033290492be98a5556/ql/src/java/org/apache/hadoop/hive/ql/exec/vector/VectorizationContext.java#L2358
but in case a filter expression was created this was missing...
> Incorrect results in some cases with vectorization
> --
>
> Key: HIVE-22210
> URL: https://issues.apache.org/jira/browse/HIVE-22210
> Project: Hive
> Issue Type: Bug
>Reporter: Zoltan Haindrich
>Assignee: Zoltan Haindrich
>Priority: Major
> Attachments: HIVE-22210.01.patch
>
>
> running the following test with TestMiniLlapLocalCliDriver leads to an
> unexpected results; the coalesce calculated inside the subquery has a value
> of 1 instead of the correct(922) value.
> {code}
> drop table if exists u_table_4;
> create table u_table_4(smallint_col_22 smallint, int_col_5 int);
> insert into u_table_4 values(238,922);
> drop table u_table_7;
> create table u_table_7 ( bigint_col_3 bigint, int_col_10 int);
> insert into u_table_7 values (571,198);
> drop table u_table_19;
> create table u_table_19 (bigint_col_18 bigint ,int_col_19 int, STRING_COL_7
> string);
> insert into u_table_19 values (922,5,'500');
> set hive.mapjoin.full.outer=true;
> set hive.auto.convert.join=true;
> set hive.query.results.cache.enabled=false;
> set hive.merge.nway.joins=true;
> set hive.vectorized.execution.enabled=true;
> --explain analyze
> SELECT
> a5.int_col,
> 922 as expected,
> COALESCE(a5.int_col, a5.aa) as expected2,
> a5.int_col_3 as reality
> FROMu_table_19 a1
> FULL OUTER JOIN
> (
>SELECT a2.int_col_5 AS int_col,
> a2.smallint_col_22 as aa,
> COALESCE(a2.int_col_5, a2.smallint_col_22) AS
> int_col_3
>FROM u_table_4 a2
> ) a5
> ON (
> a1.bigint_col_18) = (a5.int_col_3)
> INNER JOIN
> (
> SELECT a3.bigint_col_3
> AS int_col,
> Cast (COALESCE(a3.bigint_col_3,
> a3.bigint_col_3, a3.int_col_10) AS BIGINT) * Cast (a3.bigint_col_3 AS BIGINT)
> AS int_col_3
> FROM u_table_7 a3
> WHEREbigint_col_3=571
> ) a4
> ON (a1.int_col_19=5)
> OR ((a5.int_col_3) IN (a4.int_col, 10))
> where
> a1.STRING_COL_7='500'
> ORDER BYint_col DESC nulls last limit 100
> ;
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
[jira] [Commented] (HIVE-22210) Incorrect results in some cases with vectorization
[
https://issues.apache.org/jira/browse/HIVE-22210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16931274#comment-16931274
]
Zoltan Haindrich commented on HIVE-22210:
-
I suppose IdentityExpression's output may not be freed; as its directly relays
the column...there were already a tweak for this:
https://github.com/apache/hive/blob/62f0c3fe89d7112d2cbc5c033290492be98a5556/ql/src/java/org/apache/hadoop/hive/ql/exec/vector/VectorizationContext.java#L2358
but in case a filter expression was created this was missing...
> Incorrect results in some cases with vectorization
> --
>
> Key: HIVE-22210
> URL: https://issues.apache.org/jira/browse/HIVE-22210
> Project: Hive
> Issue Type: Bug
>Reporter: Zoltan Haindrich
>Assignee: Zoltan Haindrich
>Priority: Major
> Attachments: HIVE-22210.01.patch
>
>
> running the following test with TestMiniLlapLocalCliDriver leads to an
> unexpected results; the coalesce calculated inside the subquery has a value
> of 1 instead of the correct(922) value.
> {code}
> drop table if exists u_table_4;
> create table u_table_4(smallint_col_22 smallint, int_col_5 int);
> insert into u_table_4 values(238,922);
> drop table u_table_7;
> create table u_table_7 ( bigint_col_3 bigint, int_col_10 int);
> insert into u_table_7 values (571,198);
> drop table u_table_19;
> create table u_table_19 (bigint_col_18 bigint ,int_col_19 int, STRING_COL_7
> string);
> insert into u_table_19 values (922,5,'500');
> set hive.mapjoin.full.outer=true;
> set hive.auto.convert.join=true;
> set hive.query.results.cache.enabled=false;
> set hive.merge.nway.joins=true;
> set hive.vectorized.execution.enabled=true;
> --explain analyze
> SELECT
> a5.int_col,
> 922 as expected,
> COALESCE(a5.int_col, a5.aa) as expected2,
> a5.int_col_3 as reality
> FROMu_table_19 a1
> FULL OUTER JOIN
> (
>SELECT a2.int_col_5 AS int_col,
> a2.smallint_col_22 as aa,
> COALESCE(a2.int_col_5, a2.smallint_col_22) AS
> int_col_3
>FROM u_table_4 a2
> ) a5
> ON (
> a1.bigint_col_18) = (a5.int_col_3)
> INNER JOIN
> (
> SELECT a3.bigint_col_3
> AS int_col,
> Cast (COALESCE(a3.bigint_col_3,
> a3.bigint_col_3, a3.int_col_10) AS BIGINT) * Cast (a3.bigint_col_3 AS BIGINT)
> AS int_col_3
> FROM u_table_7 a3
> WHEREbigint_col_3=571
> ) a4
> ON (a1.int_col_19=5)
> OR ((a5.int_col_3) IN (a4.int_col, 10))
> where
> a1.STRING_COL_7='500'
> ORDER BYint_col DESC nulls last limit 100
> ;
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
