Question

How do I do an inverse join with more than one key column?

In this baby-toy SqlServer example, I have the following

CREATE TABLE [dbo].[CarList](
 [myID] [int] IDENTITY(1,1) NOT NULL,
 [CarColour] [varchar](32) NOT NULL,
 [CarName] [varchar](128) NOT NULL,
 [CarCompany] [varchar](32) NOT NULL,

 CONSTRAINT [PK_CarList] PRIMARY KEY CLUSTERED(
  [myID] ASC,
  [CarColour] ASC,
  [CarName] ASC,
  [CarCompany] ASC
 )
)

GO

INSERT INTO CarList (CarColour, CarName, CarCompany)
VALUES('blue', 'Abe', 'Ford')

Elsewhere in the DB I have a table like

CREATE TABLE [dbo].[NewCars](
 [CarColour] [varchar](32) NOT NULL,
 [CarName] [varchar](128) NOT NULL,
 [CarCompany] [varchar](32) NOT NULL,
)

GO

INSERT INTO NewCars (CarColour, CarName, CarCompany)
SELECT 'blue', 'Abe', 'Ford'
 UNION ALL
SELECT 'blue', 'Abe', 'GM'
 UNION ALL
SELECT 'blue', 'Betty', 'Ford'
 UNION ALL
SELECT 'green', 'Abe', 'Honda'

Now I want to insert cars I don’t already have in the CarList table

Something like…

INSERT INTO CarList ( CarColour, CarName, CarCompany)
 SELECT DISTINCT new.CarColour, new.CarName, new.CarCompany
 FROM    NewCars new, CarList old
 WHERE   new.CarColour  <> old.CarColour
     AND new.CarName    <> old.CarName
     AND new.CarCompany <> old.CarCompany

Which doesn’t work because the “blue’, ‘Betty’, ‘Ford’ row will get filtered out…

If this were just a single ID of some kind it would be really easy

INSERT INTO myTable (myID, param1, param2, etc)
SELECT param1, param2, etc
FROM someOtherTable new
WHERE new.myID NOT IN (SELECT myID FROM myTable)

For reasons I don’t really want to get into, I cannot remove rows from NewCars that match CarList. I also need to do this in one pass if possible.

[edit]
Thanks guys!

Answer 1

thanks for DDL and DML

Here is one way

INSERT INTO CarList ( CarColour, CarName, CarCompany)
 SELECT DISTINCT *
 FROM    NewCars n
where not exists (select 1 from CarList c where c.CarColour =n.CarColour
and c.CarName = n.CarName
and c.CarCompany = n.CarCompany)

There are at least 4 different way to do this

• NOT IN (will not work for more than 1 column like you have)
• NOT EXISTS
• LEFT and RIGHT JOIN
• OUTER APPLY (2005+)
• EXCEPT (2005+)

Read Select all rows from one table that don’t exist in another table