Ok, I'm trying to make an indexed view that is against a simple table that stores the results of what people think is good/bad for a post. This is the results of a thumbs up / thumbs down, voting on posts.
So here's my pseduo fake table :-
HelpfulPostId INT IDENTITY(1,1) NOT NULL PRIMARY KEY,
PostId INT NOT NULL,
IsHelpful BIT NOT NULL,
UserId INT NOT NULL
So a user can only have one vote per post. It's either a 1 (helpful) or 0 (unhelpful) <-- not sure of a better way to handle that, if there is a better way.
Ok. What i'm trying to do is get a view that looks like the following.
HelpfulPostId INT IDENTITY(1,1) NOT NULL PRIMARY KEY,
PostId INT NOT NULL,
IsHelpfulCount COUNT_BIG (WHERE IsHelpful = 1)
IsNotHelpfulCount COUNT_BIG (WHERE IsHelpful = 0)
And finally, i'll need to make it schemabindable so i can add an index on the PK and then an index on the PostId.
I have no idea about the sql to make the view. Any suggestions?
Cheers :)
Thoughts:
There are other limitations of indexed views
CREATE VIEW dbo.Example
WITH SCHEMABINDING
AS
SELECT
PostId,
SUM(CAST(IsHelpful AS bigint)) AS IsHelpfulCount,
SUM(CAST(1-IsHelpful AS bigint)) AS IsNotHelpfulCount,
COUNT_BIG(*) AS Dummy --Used to satisfy requirement
FROM
dbo.bob
GROUP BY
PostId
GO
CREATE UNIQUE CLUSTERED INDEX IXC_Test ON dbo.Example (PostId)
GO
Edit: Removed the Identity field, which was accidently added to the original question/post.
Edit 2 (gbn): I forgot that any aggregate in an indexed view also needs a COUNT_BIG(*). So, simply add one as a dummy column. I've tested this.
If the view definition uses an aggregate function, the SELECT list must also include COUNT_BIG (*).