Count(*) – Explaining different behaviour in Joins

by

Observations : 

  1. Count(1) or Count(*) – This is never expanded on each column individually so will work perfectly fine on complete data. 
    1. Count(1) is more optimized then Count(*)
  1. Count(source.*) – source represents “Left table” of “Left Outer Join”:
    1. This will be evaluated as Count(source.col1, source.col2, …. source.colN )
    2. So, if any column has NULL, then the complete row will get filtered out in the count result.
    3. Rows, in which all columns from the Source query has “NOT NULLs” will be counted only
  1. Count(target.*) – target represents “Right table” of “Left Outer Join”
    1. Same explanation as above.
    2. This will be evaluated as Count(target.col1, target.col2, …. target.colN )
    3. So, if any column has NULL, then the complete row will get filtered out in the count result.
    4. Rows, in which all columns from the Target query has “NOT NULLs” will be counted only
    5. Solution to count target table – Use any column (not limited to key column with non NULL values) for which we will have no NULLs (e.g. bookmark/last_modified_ts column, as this, is linked to each row and will have current_timestamp value)

The above was analysed from the Query plan. Please refer to the below queries :

Source Table — stg.nik_test_1
Target Table — stg.nik_test_2

DDLs:
create table stg.nik_test_1 (id int, first string, last string);
create table stg.nik_test_2 (id int, first string, last string);

insert into stg.nik_test_1 values(1,’NN’,’A1′);
insert into stg.nik_test_1 values(2,NULL,’A1′);
insert into stg.nik_test_1 values(NULL,’M1′,NULL);

insert into stg.nik_test_2 values(4,’Y1′,’S1′);
insert into stg.nik_test_2 values(5,NULL,’S1′);
insert into stg.nik_test_2 values(NULL,’N1′,NULL);
insert into stg.nik_test_2 values(2,’Y1′,’S1′);

Case -1:

explain using json select count(*) from
(select * from stg.nik_test_1) source
left outer join
(select * from stg.nik_test_2) target
on source.id = target.id
;
 
-- OUTPUT
{"GlobalStats":{"partitionsTotal":2,"partitionsAssigned":2,"bytesAssigned":3072},
"Operations":[[
{"id":0,"operation":"Result","expressions":["COUNT(*)"]},
{"id":1,"parent":0,"operation":"Aggregate","expressions":["aggExprs: [COUNT(*)]"]},
{"id":2,"parent":1,"operation":"LeftOuterJoin","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":3,"parent":2,"operation":"TableScan","objects":["STG.NIK_TEST_1"],"expressions":["ID"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536},
{"id":4,"parent":2,"operation":"Filter","expressions":["NIK_TEST_2.ID IS NOT NULL"]},
{"id":5,"parent":4,"operation":"JoinFilter","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":6,"parent":5,"operation":"TableScan","objects":["STG.NIK_TEST_2"],"expressions":["ID"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536}
]]}

Case-2:

explain using json select count(source.*) from
(select * from stg.nik_test_1) source
left outer join
(select * from stg.nik_test_2) target
on source.id = target.id
;
{"GlobalStats":{"partitionsTotal":2,"partitionsAssigned":2,"bytesAssigned":3072},"Operations":[[
{"id":0,"operation":"Result","expressions":["COUNT(COUNT(COUNT(NIK_TEST_1.ID, NIK_TEST_1.FIRST, NIK_TEST_1.LAST)))"]},
{"id":1,"parent":0,"operation":"LeftOuterJoin","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":2,"parent":1,"operation":"TableScan","objects":["STG.NIK_TEST_1"],"expressions":["ID","FIRST","LAST"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536},
{"id":3,"parent":1,"operation":"Filter","expressions":["NIK_TEST_2.ID IS NOT NULL"]},
{"id":4,"parent":3,"operation":"JoinFilter","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":5,"parent":4,"operation":"TableScan","objects":["STG.NIK_TEST_2"],"expressions":["ID"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536}
]]}

Case-3:

explain using json select count(target.*) from
(select * from stg.nik_test_1) source
left outer join
(select * from stg.nik_test_2) target
on source.id = target.id
;
{"GlobalStats":{"partitionsTotal":2,"partitionsAssigned":2,"bytesAssigned":3072},"Operations":[[
{"id":0,"operation":"Result","expressions":["COUNT(NIK_TEST_2.ID, NIK_TEST_2.FIRST, NIK_TEST_2.LAST)"]},
{"id":1,"parent":0,"operation":"Aggregate","expressions":["aggExprs: [COUNT(NIK_TEST_2.ID, NIK_TEST_2.FIRST, NIK_TEST_2.LAST)]"]},
{"id":2,"parent":1,"operation":"LeftOuterJoin","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":3,"parent":2,"operation":"TableScan","objects":["STG.NIK_TEST_1"],"expressions":["ID"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536},
{"id":4,"parent":2,"operation":"Filter","expressions":["NIK_TEST_2.ID IS NOT NULL"]},
{"id":5,"parent":4,"operation":"JoinFilter","expressions":["joinKey: (NIK_TEST_1.ID = NIK_TEST_2.ID)"]},
{"id":6,"parent":5,"operation":"TableScan","objects":["STG.NIK_TEST_2"],"expressions":["ID","FIRST","LAST"],"partitionsAssigned":1,"partitionsTotal":1,"bytesAssigned":1536}
]]}
Views: 2,793

Leave a Comment