问题
I have a table in SQL Server 2005 with a foreign key, and I want that foreign key to be a unique value or null. I've set it up as a unique key, but it won't allow me to have multiple nulls in the same table. Is it possible to do what I want?
回答1:
This is a long time complaint about SQL Server's Unique constraints/indexes. The best solution is to create a view with schemabinding and then put a unique index on that column:
Create View dbo.MyUniqueColView
With SchemaBinding
As
Select MyColToBeUnique
From MyTable
Where MyColToBeUnique Is Not Null
GO
Create Unique Clustered Index IX_MyTable_MyColToBeUnique On MyUniqueColView ( MyColToBeUnique )
回答2:
You cannot enforce this with a table level constraint in SQL Server 2005 but you can create a view with definition SELECT Col FROM t WHERE Col IS NOT NULL and then create a unique clustered index on that.
回答3:
In SQL Server 2008, you could create a filtered index on the values that are not null - unfortunately, that is a new feature in the 2008 release, so in 2005, you're not able to do anything like that, I'm afraid.
回答4:
Always an annoyance of mine. I prefer the "other" interpretation of the standard.
See Creating a unique constraint that ignores nulls in SQL Server. It covers three different approaches.
Solution 1: Filtered indexes.
This seems pretty common, judging from other replies. I have not tried this approach -- it does require SQL Server 2008 though.
Solution 2: Constraint on computed column
I have used this, and it does work. However, relationships may need to be adjusted and/or secondary non-unique covering indexed added (one for unique constraint, one for index cover).
Solution 3: UDF Check Constraint (Not recommended)
Ahh, the power of TSQL (read: something else to deal with). I have not gone down this route.
Happy coding.
来源：https://stackoverflow.com/questions/5982933/sql-server-2005-unique-key-with-null-value