NOEXPAND Required in 2012 Enterprise

Question

SQL Server 2012 Enterprise, build 11.0.2332.0. I've created a view that joins two tables, and has a computed column that includes components from both underlying tables. I've created a unique clustered index on the view on two columns, the first of which
is the computed column. When I query the view like this:
SELECT * FROM schema.View WHERE ComputedColumn = '<some value>';
the optimizer does not use the view index, but instead goes directly against the underlying tables (slow). If I add WITH (NOEXPAND) to the query, it uses the view index and is very fast. I thought that NOEXPAND should never be required when using Enterprise
edition for the view index to be considered. Interestingly, if I simply query for the count(*), the view index is used without specifying NOEXPAND.
What am I missing?
Thanks
Vern Rabe

Answer 1

There are definitely cases where the optimizer expands a view when it shouldn't. 
To reiterate what David said: the optimizer never expands a view. It's the algebrizer that expands the view, unless you specify NOEXPAND. If you are on Enterprise Edition, the optimizer
may consider replacing parts of the query with the indexed view.
From this follows that the optimizer:
1) May entirely neglect this particular view (view matching is fairly expensive.)
2) Discard the view (because of  misestimation.)
Unfortunately, I don't know if there is a way to see if an indexed view is considered. Way back in SQL 2000, the optimizer generate a ExecutionWarning event with this information, but I am unable to create this on SQL 2012.
Whether you could do anything, except adding NOEXPAND? No idea, since I have no information about tables, indexes etc. But UPDATE STATISTICS WITH FULLSCAN on the view itself and the underlying tables.
But from the description you give of the view, I would place my bets that this falls under category 1. The optimizer is not even trying to match this the view.
Erland Sommarskog, SQL Server MVP, [email protected]