问题
I have two tables that I'm trying to join and insert into another table based on a combination of three columns. I'll explain.
Table M
| ANO | BNO | Timestamp | Duration
---------------------------------------------------------------
| 5612853 | 4732621 | 21.11.2013 09:50:58 | 196
| 4842988 | 5610953 | 21.11.2013 17:34:58 | 98
| 7765759 | 5612853 | 21.11.2013 20:48:00 | 377
| 2470321 | 2470263 | 21.11.2013 21:47:18 | 125
Table N
| ANO | BNO | Timestamp | Duration
---------------------------------------------------------------
| 5612853 | 4732621 | 21.11.2013 09:50:52 | 196
| 4842988 | 5610953 | 21.11.2013 17:34:53 | 98
| 7765759 | 5612853 | 21.11.2013 20:47:55 | 377
| 2470321 | 2470263 | 21.11.2013 21:47:13 | 125
Now these two tables have to be matched and inserted into Table MN based on
M.ANO=N.ANO and M.BNO=N.BNO and ((M.TIMESTAMP = N.TIMESTAMP+5/86400) or (M.TIMESTAMP = N.TIMESTAMP+6/86400))
So in theory, my output Table MN should be
| ANO | BNO | Timestamp | Duration || ANO | BNO | Timestamp | Duration
--------------------------------------------------------------------------------------------------------------
| 5612853 | 4732621 | 21.11.2013 09:50:58 | 196 || 5612853 | 4732621 | 21.11.2013 09:50:52 | 196
| 4842988 | 5610953 | 21.11.2013 17:34:58 | 98 || 4842988 | 5610953 | 21.11.2013 17:34:53 | 98
| 7765759 | 5612853 | 21.11.2013 20:48:00 | 377 || 7765759 | 5612853 | 21.11.2013 20:47:55 | 377
| 2470321 | 2470263 | 21.11.2013 21:47:18 | 125 || 2470321 | 2470263 | 21.11.2013 21:47:13 | 125
Table M has about 1.4 million records, and Table N has about 0.9 million.
I've tried to join the two tables based on the below two queries. But it takes hours to execute and that isn't feasible if I have to run this on a daily basis.
INSERT INTO MN_RECON (
SELECT M.*,N.* FROM M FULL OUTER JOIN N ON
M.ANO=N.ANO AND M.BNO=N.BNO AND
((M.TIMESTAMP=N.TIMESTAMP+5/86400) OR (M.TIMESTAMP=N.TIMESTAMP+6/86400))
INSERT INTO MN_RECON (
SELECT M.*,N.* FROM M FULL OUTER JOIN N ON
M.ANO=N.ANO AND M.BNO=N.BNO AND
(M.TIMESTAMP-N.TIMESTAMP IN (5/86400,6/86400)
When I run just the SELECT statement of the above 2 queries I get an output within a minute (just a few 100 sample lines) but with the INSERT added it takes a very long time. Is there a way to optimize what I want to do?
I need it to match on timestamp because there can be multiple occurrences of the same ANO - BNO combination during the day, with the timestamp being the unique identifier between them
And I need a full outer join because I need to focus on records that are not matched, as well as matched records with a difference in duration between the two tables.
Additional Oracle information Oracle Database 11g Enterprise Edition 11.2.0.3.0 64-bit Production
EXPLAIN PLAN
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
----------------------------------------------------------------------------------------------------
| 0 | INSERT STATEMENT | | 2386K| 530M| | 2395M (1)|999:59:59 |
| 1 | LOAD TABLE CONVENTIONAL | MN_RECON | | | | | |
| 2 | VIEW | | 2386K| 530M| | 2395M (1)|999:59:59 |
| 3 | UNION-ALL | | | | | | |
|* 4 | HASH JOIN RIGHT OUTER| | 1417K| 109M| 49M| 10143 (1)| 00:02:02 |
| 5 | TABLE ACCESS FULL | N_VOICE | 968K| 38M| | 1753 (1)| 00:00:22 |
| 6 | TABLE ACCESS FULL | M_VOICE | 1417K| 52M| | 2479 (1)| 00:00:30 |
|* 7 | FILTER | | | | | | |
| 8 | TABLE ACCESS FULL | N_VOICE | 968K| 38M| | 1754 (1)| 00:00:22 |
|* 9 | TABLE ACCESS FULL | M_VOICE | 1 | 29 | | 2479 (1)| 00:00:30 |
回答1:
A simple way to speed up the query is by creating a function-based index:
CREATE INDEX indexname1 ON N (timestamp+5/86400);
CREATE INDEX indexname2 ON N (timestamp+6/86400);
来源：https://stackoverflow.com/questions/20186616/how-to-join-two-tables-based-on-a-timestamp-with-variance-of-a-few-seconds