SQL JOIN between A, B and C mixing full and left join - sql

Question

I'd like to know if there is a better way to create my left join in the below example:
SELECT TOP 10 COALESCE(A.COD_PRODUCT, B.COD_PRODUCT),
COALESCE(A.COD_FAMILY, B.COD_FAMILY),
COALESCE(A.DATE_EXTRACT, B.DATE_EXTRACT),
A.MASS
B.VOLUME
C.PRICE
FROM FIRSTTABLE A FULL JOIN SECONDTABLE B ON B.COD_PRODUCT = A.COD_PRODUCT
AND B.COD_FAMILY = A.COD_FAMILY
AND B.DATE_EXTRACT = A.DATE_EXTRACT
LEFT JOIN THIRDTABLE C ON C.COD_PRODUCT = COALESCE(A.COD_PRODUCT,B.COD_PRODUCT)
AND C.COD_FAMILY = COALESCE(A.COD_FAMILY, B.COD_FAMILY)
AND C.DATE_EXTRACT = COALESCE(A.DATE_EXTRACT, B.DATE_EXTRACT)
That kind of jointure takes long time and I suspect it to be highly expensive and improvable
EDIT: I'd like to improve this SELECT FROM JOIN statement in a View.

Answer 1

You could split the query into two: gather all data matching FIRSTTABLE. And then union it with all data matching SECONDTABLE that is not in FIRSTTABLE.
That should allow SQL Server to use the indexes on these tables better.
SELECT A.COD_PRODUCT,
A.COD_FAMILY,
A.DATE_EXTRACT,
A.MASS,
B.VOLUME,
C.PRICE
FROM FIRSTTABLE A
LEFT OUTER JOIN SECONDTABLE B
ON B.COD_PRODUCT = A.COD_PRODUCT
AND B.COD_FAMILY = A.COD_FAMILY
AND B.DATE_EXTRACT = A.DATE_EXTRACT
LEFT OUTER JOIN THIRDTABLE C
ON C.COD_PRODUCT = A.COD_PRODUCT
AND C.COD_FAMILY = A.COD_FAMILY
AND C.DATE_EXTRACT = A.DATE_EXTRACT
UNION ALL
SELECT B.COD_PRODUCT,
B.COD_FAMILY,
B.DATE_EXTRACT,
NULL AS MASS,
B.VOLUME,
C.PRICE
FROM SECONDTABLE B
LEFT OUTER JOIN THIRDTABLE C
ON C.COD_PRODUCT = B.COD_PRODUCT
AND C.COD_FAMILY = B.COD_FAMILY
AND C.DATE_EXTRACT = B.DATE_EXTRACT
WHERE NOT EXISTS (SELECT 1
FROM FIRSTTABLE A
WHERE A.COD_PRODUCT = B.COD_PRODUCT
AND A.COD_FAMILY = B.COD_FAMILY
AND A.DATE_EXTRACT = B.DATE_EXTRACT)

Answer 2

Compare the execution plan with this replacement join:
LEFT JOIN THIRDTABLE C
ON (C.COD_PRODUCT = A.COD_PRODUCT or C.COD_PRODUCT = B.COD_PRODUCT)
AND (C.COD_FAMILY = A.COD_FAMILY or C.COD_FAMILY = B.COD_FAMILY)
AND (C.DATE_EXTRACT =A.DATE_EXTRACT or C.DATE_EXTRACT = B.DATE_EXTRACT)

Answer 3

You might can achieve the same result with a UNION or UNION ALL
WITH cte AS (
SELECT
A.COD_PRODUCT,
A.COD_FAMILY,
A.DATE_EXTRACT,
A.MASS,
NULL as VOLUME
FROM
FIRSTTABLE A
UNION
SELECT
B.COD_PRODUCT,
B.COD_FAMILY,
B.DATE_EXTRACT,
NULL,
B.VOLUME
FROM
FIRSTTABLE A
)
SELECT
*
FROM
cte AB
LEFT JOIN
THIRDTABLE C ON C.COD_PRODUCT = AB.COD_PRODUCT
AND C.COD_FAMILY = AB.COD_FAMILY
AND C.DATE_EXTRACT = AB.DATE_EXTRACT
If you can have Mass and Volume for the product,family,extract combination you can use aggregate to join A and B together
WITH cte AS (
SELECT
COD_PRODUCT,
COD_FAMILY,
DATE_EXTRACT,
MAX(MASS) MASS,
MAX(VOLUME) VOLUME
FROM (
SELECT
A.COD_PRODUCT,
A.COD_FAMILY,
A.DATE_EXTRACT,
A.MASS,
NULL as VOLUME
FROM
FIRSTTABLE A
UNION ALL
SELECT
B.COD_PRODUCT,
B.COD_FAMILY,
B.DATE_EXTRACT,
NULL,
B.VOLUME
FROM
FIRSTTABLE A
) T
GROUP BY
COD_PRODUCT,
COD_FAMILY,
DATE_EXTRACT
)
SELECT
*
FROM
cte AB
LEFT JOIN
THIRDTABLE C ON C.COD_PRODUCT = AB.COD_PRODUCT
AND C.COD_FAMILY = AB.COD_FAMILY
AND C.DATE_EXTRACT = AB.DATE_EXTRACT

Answer 4

You can try this:
--- isolate the full join data from a and b into a temp table
SELECT
COD_PRODUCT= COALESCE(A.COD_PRODUCT, B.COD_PRODUCT),
COD_FAMILY= COALESCE(A.COD_FAMILY, B.COD_FAMILY),
DATE_EXTRACT= COALESCE(A.DATE_EXTRACT, B.DATE_EXTRACT),
MASS= A.MASS,
VOLUME= B.VOLUME
INTO #TEMP
FROM FIRSTTABLE A
FULL JOIN SECONDTABLE B
ON B.COD_PRODUCT = A.COD_PRODUCT
AND B.COD_FAMILY = A.COD_FAMILY
AND B.DATE_EXTRACT = A.DATE_EXTRACT
-- add index clustered onto table (covering index)
CREATE CLUSTERED INDEX ix_tempCIndex ON #Temp ([COD_PRODUCT],[COD_FAMILY],[DATE_EXTRACT],[MASS],[VOLUME]);
-- left join C to this temp table
SELECT TOP 10
T.*, C.PRICE
FROM #TEMP T
LEFT JOIN THIRDTABLE C
ON C.COD_PRODUCT = T.COD_PRODUCT
AND C.COD_FAMILY = T.COD_FAMILY
AND C.DATE_EXTRACT = T.DATE_EXTRACT
-- drop temp table
DROP TABLE #TEMP

Answer 5

A coalesce or an OR inside of a join is going to slow down the query a lot. Depending on the size of the table, a better answer might be to join to table c twice. Once on table a and again on table B, then coalesce inside of your select clause.
SELECT TOP 10 COALESCE(A.COD_PRODUCT, B.COD_PRODUCT,c.COD_PRODUCT,ca.COD_PRODUCT),
COALESCE(A.COD_FAMILY, B.COD_FAMILY,c.COD_FAMILY,ca.COD_FAMILY),
COALESCE(A.DATE_EXTRACT, B.DATE_EXTRACT,c.DATE_EXTRACT,ca.DATE_EXTRACT),
A.MASS
B.VOLUME
C.PRICE
FROM FIRSTTABLE A
FULL JOIN SECONDTABLE B
ON B.COD_PRODUCT = A.COD_PRODUCT
AND B.COD_FAMILY = A.COD_FAMILY
AND B.DATE_EXTRACT = A.DATE_EXTRACT
LEFT JOIN THIRDTABLE C
ON C.COD_PRODUCT = A.COD_PRODUCT
AND C.COD_FAMILY = A.COD_FAMILY
AND C.DATE_EXTRACT = A.DATE_EXTRACT
LEFT JOIN THIRDTABLE Ca
ON Ca.COD_PRODUCT = b.COD_PRODUCT
AND Ca.COD_FAMILY = b.COD_FAMILY
AND Ca.DATE_EXTRACT = b.DATE_EXTRACT