What's a clean way to join tables with multiple repetitive joins? - sql-server

Question

I am wondering if I'm doing this in the most efficient way possible.
I have a database table that houses custom data elements. To see what records use the data in the custom object table I have to join on 3 ID numbers, as well as specify the description. It's not that bad, but sometimes I can have up to 30 of these joins in my SQL script and that can be long. I was wondering if there is a cleaner way to code this? Maybe with a temp stored procedure or temp function?
SELECT
r.ID,
apples.status,
oranges.status,
bananas.status
FROM
record r
LEFT JOIN
custom_object apples ON apples.T_ID1 = r.T_ID1
AND apples.T_ID2 = r.T_ID2
AND apples.T_ID3 = r.T_ID3
AND apples.DESC = 'Apples'
LEFT JOIN
custom_object oranges ON oranges .ID1= r.T_ID1
AND oranges.T_ID2 = r.T_ID2
AND oranges.T_ID3 = r.T_ID3
AND oranges.DESC = 'Oranges'
LEFT JOIN
custom_object bananas ON bananas.T_ID1= r.T_ID1
AND bananas.T_ID2 = r.T_ID2
AND bananas.T_ID3 = r.T_ID3
AND bananas.DESC = 'Bananas'

Answer 1

Using a single join with a case expression seems like a much simpler approach here.
SELECT
r.ID,
AppleStatus = case when co.[DESC] = 'Apples' then co.status end,
OrangeStatus = case when co.[DESC]= 'Oranges' then co.status end,
BananaStatus = case when co.[DESC] = 'Bananas' then co.status end
FROM record r
LEFT JOIN custom_object co ON co.T_ID1 = r.T_ID1
AND co.T_ID2 = r.T_ID2
AND co.T_ID3 = r.T_ID3
AND co.[DESC] in ('Apples', 'Oanges', 'Bananas')

Answer 2

Conditional aggregation is a possibility:
SELECT
r.ID,
MAX(CASE WHEN co.[DESC] = 'Apples' THEN co.status END) AS applesStatus,
MAX(CASE WHEN co.[DESC] = 'Oranges' THEN co.status END) AS orangesStatus,
MAX(CASE WHEN co.[DESC] = 'Bananas' THEN co.status END) AS bananasStatus
FROM
record r
LEFT JOIN
custom_object co ON co.T_ID1 = r.T_ID1
AND co.T_ID2 = r.T_ID2
AND co.T_ID3 = r.T_ID3
AND co.[DESC] IN ('Apples', 'Oranges', 'Bananas') -- optional
GROUP BY r.ID
The same inside an OUTER APPLY will avoid the outer GROUP BY and allow more flexibility as you build out the rest of your query.
SELECT
r.ID,
statuses.*
FROM
record r
OUTER APPLY (
SELECT
MAX(CASE WHEN co.[DESC] = 'Apples' THEN co.status END) AS applesStatus,
MAX(CASE WHEN co.[DESC] = 'Oranges' THEN co.status END) AS orangesStatus,
MAX(CASE WHEN co.[DESC] = 'Bananas' THEN co.status END) AS bananasStatus
FROM custom_object co
WHERE co.T_ID1 = r.T_ID1
AND co.T_ID2 = r.T_ID2
AND co.T_ID3 = r.T_ID3
AND co.[DESC] IN ('Apples', 'Oranges', 'Bananas') -- optional
) statuses
A Common Table Expression (CTE) can also be used:
; WITH statuses AS (
SELECT
co.T_ID1, co.T_ID2, co.T_ID3,
MAX(CASE WHEN co.[DESC] = 'Apples' THEN co.status END) AS applesStatus,
MAX(CASE WHEN co.[DESC] = 'Oranges' THEN co.status END) AS orangesStatus,
MAX(CASE WHEN co.[DESC] = 'Bananas' THEN co.status END) AS bananasStatus
FROM custom_object co
WHERE co.[DESC] IN ('Apples', 'Oranges', 'Bananas') -- optional
GROUP BY co.T_ID1, co.T_ID2, co.T_ID3
)
SELECT
r.ID,
s.applesStatus,
s.orangesStatus,
s.bananasStatus
FROM
record r
LEFT JOIN
statuses s ON s.T_ID1 = r.T_ID1
AND s.T_ID2 = r.T_ID2
AND s.T_ID3 = r.T_ID3
The above all assume no duplicate properties per ID combination.
For best performance, you should define an index on custom_object(T_ID1, T_ID2, T_ID3, [DESC]).
Sample results:
ID
applesStatus
orangesStatus
bananasStatus
1
Apples1
Oranges1
Bananas1
2
Apples2
null
null
3
null
Oranges3
null
4
null
null
Bananas4
5
null
null
null
6
Apples6
null
Bananas6
See this db<>fiddle.