I’m using Oracle 11g.
I have 2 related tables: stored values (A) and new values to insert (B). Both are related between them with an id of 3 columns (client, group and personcode). Each table has about 20 other columns (let’s call them attributes).
I have to match them so I can know which values are new (id in B and not in A) so I insert them in A, which are equals (id in B and in A with the same attributes) and which are not in the new values (id in A but not in B anymore), so I delete them from the stored values (A).
For instance:
A: client | group | personcode | name | surname _________________________________________________ 1 | 1 | 1 | joe | doe 1 | 1 | 2 | carl | john 1 | 1 | 3 | john | john
B:
client | group | personcode | name | surname _________________________________________________ 1 | 1 | 1 | joe | doe 1 | 1 | 3 | john | john 1 | 1 | 4 | mary | con
In this example, person 4 is new, person 2 should be deleted and 1 and 3 remains the same.
So, I need a query which returns the following results:
client | group | personcode | action _________________________________________ 1 | 1 | 1 | equal 1 | 1 | 2 | remove 1 | 1 | 3 | equal 1 | 1 | 4 | new
What I’ve made is the following query:
WITH
A AS (
-- select from A table
),
B AS
(
-- select from B table
),
delete AS
(
-- select from A WHERE NOT EXISTS (B.id = A.ID)
),
news AS
(
-- select from B WHERE NOT EXISTS (A.id = B.ID)
),
eq AS
(
-- select A.* from A, B WHERE A.id = B.id AND A.attributes = B.attributes
)
select action.client, action.group, action.personcode, 'remove' from delete action
UNION ALL
select action.client, action.group, action.personcode, 'new' from news action
UNION ALL
select action.client, action.group, action.personcode, 'equal' from eq action
;
The problem is that, although each of those 3 lasts selects runs in less than 10 seconds, when I merge them using UNION or UNION ALL, the complete query lasts about 90 seconds, even if delete or new or equal are empty. It could be more than 3000 rows in A or in B.
Is there any way to get this results in a better, faster way?
Well, thanks all for your reply.
I’ve finally made a view to which I pass some parameters to filter the first two queries, using the strategy described in this blog
The complete process lasts 30 secs now, and 0 if there are no rows at A or B (before, it lasts 90 secs always).
This is the solution which less affects my current procedures.