Distinct clause on large join - sql

Question

Following this question, suppose now I've set up the indexes, and now I want only to return certain field, without duplicates:
Select distinct A.cod
from A join B
on A.id1=B.id1 and
A.id2=B.id2
where A.year=2016
and B.year=2016
the problem now is I'm getting something like 150k cod, with only 1000 distinct values, so my query is very inefficient.
Question: how can I improve that? i.e, how can I tell the DB, for every row on A, to stop joining that row as soon as a match is found?
Thank you in advance!

Answer 1

I'm basing my answer on your question:
how can I tell the DB, for every row on A, to stop joining that row as soon as a match is found?
with the EXISTS clause, once it sees a match it will stop and check for the next record to be checked.
adding the DISTINCT will filter out any duplicate CODs (in case there is one).
select DISTINCT cod
from A ax
where year = 2016
and exists ( select 1
from B bx
WHERE Ax.ID1 = Bx.ID1
AND Ax.ID2 = Bx.ID2
AND Ax.YEAR = Bx.YEAR);
EDIT: Was curious which solution (IN or EXISTS) will give me a better Explain plan
Create the 1st Table Definition
Create table A
(
ID1 number,
ID2 number,
cod varchar2(100),
year number
);
insert 4000000 sequential numbers
BEGIN
FOR i IN 1..4000000 loop
insert into A (id1, id2, cod, year)
values (i, i , i, i);
end loop;
END;
commit;
Create Table B and insert the same data into to it
Create table B
as
select *
from A;
Reinsert Data from Table A to make duplicates
insert into B
select *
from A
Build the Indexes mentioned in the Previous Post Index on join and where
CREATE INDEX A_IDX ON A(year, id1, id2);
CREATE INDEX B_IDX ON B(year, id1, id2);
Update a bunch of rows to make it fetch multiple rows with the year 2016:
update B
set year = 2016
where rownum < 20000;
update A
set year = 2016
where rownum < 20000;
commit;
Check Explain plan using EXISTS
Plan hash value: 1052726981
----------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 44 | 7 (15)| 00:00:01 |
| 1 | HASH UNIQUE | | 1 | 44 | 7 (15)| 00:00:01 |
| 2 | NESTED LOOPS SEMI | | 1 | 44 | 6 (0)| 00:00:01 |
| 3 | TABLE ACCESS BY INDEX ROWID| A | 1 | 26 | 4 (0)| 00:00:01 |
|* 4 | INDEX RANGE SCAN | A_IDX | 1 | | 3 (0)| 00:00:01 |
|* 5 | INDEX RANGE SCAN | B_IDX | 2 | 36 | 2 (0)| 00:00:01 |
----------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
4 - access("YEAR"=2016)
5 - access("BX"."YEAR"=2016 AND "AX"."ID1"="BX"."ID1" AND "AX"."ID2"="BX"."ID2")
filter("AX"."YEAR"="BX"."YEAR")
Check Explain plan using IN
Plan hash value: 3002464630
----------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 44 | 7 (15)| 00:00:01 |
| 1 | HASH UNIQUE | | 1 | 44 | 7 (15)| 00:00:01 |
| 2 | NESTED LOOPS | | 1 | 44 | 6 (0)| 00:00:01 |
| 3 | TABLE ACCESS BY INDEX ROWID| A | 1 | 26 | 4 (0)| 00:00:01 |
|* 4 | INDEX RANGE SCAN | A_IDX | 1 | | 3 (0)| 00:00:01 |
|* 5 | INDEX RANGE SCAN | B_IDX | 1 | 18 | 2 (0)| 00:00:01 |
----------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
4 - access("YEAR"=2016)
5 - access("YEAR"=2016 AND "ID1"="ID1" AND "ID2"="ID2")
Although my test case is limited, i'm guessing that both the IN and EXISTS clause have nearly the same execution.

Answer 2

On the face of it, what you are actually trying to do should be done like this:
select distinct cod
from A
where year = 2016
and (id1, id2) in (select id1, id2 from B where year = 2016)
The subquery in the WHERE condition is a non-correlated query, so it will be evaluated only once. And the IN condition is evaluated using short-circuiting; instead of a complete join, it will search through the results of the subquery only until a match is found.
EDIT: As Migs Isip points out, there may be duplicate codes in the original table, so a "distinct" may still be needed. I edited my code to add it back after Migs posted his answer.

Answer 3

Not sure about your existing indexes but you can improve your query a bit by adding another JOIN condition like
Select distinct A.cod
from A join B
on A.id1=B.id1 and
A.id2=B.id2 and
A.year = B.year // this one
where A.year=2016;