[HIVE-21368] Vectorization: Unnecessary Decimal64 -> HiveDecimal conversion

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0.0-alpha-1
• Assignee: Teddy Choi
• Reporter: Gopal Vijayaraghavan
• Created: 01/Mar/19 18:02
• Updated: 17/Nov/22 08:51
• Resolved: 14/Mar/19 05:18

Description

Joins projecting Decimal64 have a suspicious cast in the inner loop

ConvertDecimal64ToDecimal(col 14:decimal(7,2)/DECIMAL_64) -> 24:decimal(7,2)'

create temporary table foo(x int , y decimal(7,2));
create temporary table bar(x int , y decimal(7,2));
set hive.explain.user=false;
explain vectorization detail select sum(foo.y) from foo, bar where foo.x = bar.x;

'                      Map Join Operator'
'                        condition map:'
'                             Inner Join 0 to 1'
'                        keys:'
'                          0 _col0 (type: int)'
'                          1 _col0 (type: int)'
'                        Map Join Vectorization:'
'                            bigTableKeyColumnNums: [0]'
'                            bigTableRetainedColumnNums: [3]'
'                            bigTableValueColumnNums: [3]'
'                            bigTableValueExpressions: ConvertDecimal64ToDecimal(col 1:decimal(7,2)/DECIMAL_64) -> 3:decimal(7,2)'
'                            className: VectorMapJoinInnerBigOnlyLongOperator'
'                            native: true'
'                            nativeConditionsMet: hive.mapjoin.optimized.hashtable IS true, hive.vectorized.execution.mapjoin.native.enabled IS true, hive.execution.engine tez IN [tez, spark] IS true, One MapJoin Condition IS true, No nullsafe IS true, Small table vectorizes IS true, Fast Hash Table and No Hybrid Hash Join IS true'
'                            projectedOutputColumnNums: [3]'