Reputation: 34160
I want to union 2 select statements that the result will be distinct record-wise however I want to omit duplicate results in second select statement (considering some columns)
select id,name,[type],[parent] from table1 where [type] = 1
union
select * from table2 // but exclude results from this table where
// a record with the same type and parent exists
// in the first select
I have thought of this (not tested):
select * from(
select *,rank() over(order by [type],[parent]) [rank] from(
select id,name,[type],[parent] from table1 where [type] = 1
union
select * from table2) t
) a where rank = 1
but it doesn't seem right, is there a better way to exclude duplicate from second select?
Edit:
each item can have add-ons. and add-ons are created in 2 ways:
1.specifically created add-ons in table1
2.publicly defining that an item of type x must have add-on
first select gets the list of addons that are specifically created for the Items , table2 create a list of add-ons for all the Items now there will be a duplicate add-on if there were an add-on specifically created for an Item.
Upvotes: 13
Views: 22109
Reputation: 866
I just saw this question, and I think that instead of not exist you can use left join, it has better performance, so the query should look like below:
select * from table1
union
select * from table2
left join table1 exists
on table2.parent = exists.parent
and table2.type = exists.type
where exists.id is null
Upvotes: 0
Reputation: 5367
try this:
;WITH cte AS (
SELECT *, 1 AS SetID FROM table1 WHERE [Type] = 1
UNION ALL
SELECT *, 2 AS SetID FROM table2
)
,cte2 as (
SELECT *,
RANK() OVER (PARTITION BY [Type], [Parent] ORDER BY SetID) FROM cte) rk
FROM cte
)
SELECT * FROM cte2 WHERE rk = 1
Upvotes: 1
Reputation: 9933
try
select * from table1
union
select * from table2
where not exists(select 1 from table1
where table2.parent = table1.parent
and table2.type = table1.type)
Upvotes: 13