October 4, 2015 at 6:07 am
Hi SQL Folks,
I have two tables tabA (cola1, cola2, cola3) and tabB(colb1, colb2, colb3, colb4) which I need to join on all 3 columns of table A.
Of the 3 columns in tabA, few can be NULL, in that case I want to check the joining condition for the rest of the columns, so its conditional joining. Let me rephrase what I am trying to acheive - I need to check if the columns in joining condition is NULL in my 1st table (tabA), If so I need to check the joining condition for the rest of the two columns, if 2nd column is again NULL, I need to check the joining condition on the third column.
What I am trying to do is as below. Its working, but is very slow when one of the tables is huge. Can I optimize it or rewrite in a better way ?
--- First Create two tables
Create table tabA
(cola1 nvarchar(100), cola2 nvarchar(100), cola3 nvarchar(100))
Insert into tabA values (NULL,'A1','A2')
Select * from tabA
create table tabB
(colb1 nvarchar(100), colb2 nvarchar(100), colb3 nvarchar(100), colb4 nvarchar(100))
Insert into tabB values ('B1','B2','B3','B4')
Select * from tabB
------- Joining the two ---
Select A.cola1,B.*
from tabA as A INNER JOIN tabB as B
ON
(select case when (A.cola1) is null THEN 'a' else B.colb1 end)= (Select case When A.cola1 is null then 'a' else B.colb1 end )and
(select case when (A.cola2) is null THEN 'a' else B.colb2 end)= (Select case When A.cola2 is null then 'a' else B.colb2 end )and
(select case when (A.cola3) is null THEN 'a' else B.colb3 end)= (Select case When A.cola3 is null then 'a' else B.colb3 end )
October 4, 2015 at 6:20 am
By slow I mean its taking more than 10 minutes which is not acceptable as this would form the query for the dataset of a ssrs report.
October 4, 2015 at 10:09 am
October 4, 2015 at 10:43 am
The first table - tabA is being populated by Users and the columns are Nullable. So few records might have 1 NULL value, other records might have two NULL values etc.
cola1, cola2, Cola3
NULL, aa, abc
acd, NULL, sr
acc, NULL, NULL
acb, cbf, cbf
This table needs to be join with the other table tabB in Database which would have values for all these columns and I need to join the two with whatever available non-null columns I have in tabA. May be I am repeating what I have already mentioned in my first post on this thread but this is what I need to do.
Select <col list> from tabA A INNER JOIN tabB B
ON
A.cola1=B.colb1 and -- this condition shouldn't participate if A.cola1 IS NULL
A.cola2=B.colb2 and -- this condition shouldn't participate if A.cola2 IS NULL
A.cola3=B.colb3 -- this condition shouldn't participate if A.cola3 IS NULL
Thanks again.
October 4, 2015 at 11:21 am
Two possible solutions
😎
USE tempdb;
GO
SET NOCOUNT ON;
--- First Create two tables
IF OBJECT_ID(N'dbo.tabA') IS NOT NULL DROP TABLE dbo.tabA;
CREATE TABLE dbo.tabA
(
cola1 NVARCHAR(100) NULL
,cola2 NVARCHAR(100) NULL
,cola3 NVARCHAR(100) NULL
);
INSERT INTO dbo.tabA (cola1,cola2,cola3)
VALUES ( NULL, 'A1', 'A2')
,( 'A0', NULL, 'A2')
,( 'A0', 'A1', NULL)
,( 'A0', 'A1', 'A2')
IF OBJECT_ID(N'dbo.tabB') IS NOT NULL DROP TABLE dbo.tabB;
CREATE TABLE dbo.tabB
(
colb1 NVARCHAR(100) NOT NULL
,colb2 NVARCHAR(100) NOT NULL
,colb3 NVARCHAR(100) NOT NULL
,colb4 NVARCHAR(100) NOT NULL
);
INSERT INTO dbo.tabB (colb1,colb2,colb3,colb4)
VALUES ( 'B1', 'B2', 'B3', 'B4')
,( 'A0', 'A1', 'A2', 'THIS IS IT')
,( 'C9', 'B2', 'B3', 'B4')
-- IN THIS CASE THERE IS NO USE ADDING AN INDEX TO tabA, IT WILL NOT BE USED.
CREATE NONCLUSTERED INDEX NCLIDX_DBO_TABB_COLB1_COLB2_COLB3_INCL_COLB4 ON dbo.tabB (colb1 ASC,colb2 ASC,colb3 ASC) INCLUDE (colb4);
-- TWO METHODS
SELECT
TA.cola1
,TA.cola2
,TA.cola3
,TB.colb1
,TB.colb2
,TB.colb3
,TB.colb4
FROM dbo.tabA TA
INNER JOIN dbo.tabB TB
ON TB.colb1 = ISNULL(TA.cola1,TB.colb1)
AND TB.colb2 = ISNULL(TA.cola2,TB.colb2)
AND TB.colb3 = ISNULL(TA.cola3,TB.colb3)
;
----------------
SELECT
TA.cola1
,TA.cola2
,TA.cola3
,TB.colb1
,TB.colb2
,TB.colb3
,TB.colb4
FROM dbo.tabA TA
CROSS APPLY dbo.tabB TB
WHERE CASE WHEN TA.cola1 IS NULL THEN TB.colb1 ELSE TA.cola1 END = TB.colb1
AND CASE WHEN TA.cola2 IS NULL THEN TB.colb2 ELSE TA.cola2 END = TB.colb2
AND CASE WHEN TA.cola3 IS NULL THEN TB.colb3 ELSE TA.cola3 END = TB.colb3
;
Results
cola1 cola2 cola3 colb1 colb2 colb3 colb4
------ ------ ------ ------ ------ ------ -----------
NULL A1 A2 A0 A1 A2 THIS IS IT
A0 NULL A2 A0 A1 A2 THIS IS IT
A0 A1 NULL A0 A1 A2 THIS IS IT
A0 A1 A2 A0 A1 A2 THIS IS IT
October 5, 2015 at 2:59 am
A third possible solution:
SELECT
ta.cola1,
ta.cola2,
ta.cola3,
colb1 = COALESCE(tb1.colb1, tb2.colb1, tb3.colb1),
colb2 = COALESCE(tb1.colb2, tb2.colb2, tb3.colb2),
colb3 = COALESCE(tb1.colb3, tb2.colb3, tb3.colb3),
colb4 = COALESCE(tb1.colb4, tb2.colb4, tb3.colb4)
FROM dbo.tabA ta
LEFT JOIN dbo.tabB tb1 ON tb1.colb1 = ta.cola1
LEFT JOIN dbo.tabB tb2 ON tb2.colb2 = ta.cola2
LEFT JOIN dbo.tabB tb3 ON tb3.colb3 = ta.cola3
Also, see this article by Dwain Camps[/url].
For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden
October 5, 2015 at 1:52 pm
Those JOIN conditions will always be met, regardless of the actual values in A (unless if a B column is NULL). So just do a CROSS JOIN of the tables:
Select A.cola1,B.*
from tabA as A CROSS JOIN tabB as B
SQL DBA,SQL Server MVP(07, 08, 09) A socialist is someone who will give you the shirt off *someone else's* back.
October 6, 2015 at 4:17 am
Couldn't say a thanks yesterday as I was a bit occupied.
Better late than never, so a big thanks for you.
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply