Question

I have a table that contains, for example, two fields that I want to make unique within the database. For example:

create table Subscriber (
    ID int not null,
    DataSetId int not null,
    Email nvarchar(100) not null,
    ...
)

The ID column is the primary key and both DataSetId and Email are indexed.

What I want to be able to do is prevent the same Email and DataSetId combination appearing in the table or, to put it another way, the Email value must be unique for a given DataSetId.

I tried creating a unique index on the columns

CREATE UNIQUE NONCLUSTERED INDEX IX_Subscriber_Email
ON Subscriber (DataSetId, Email)

but I found that this had quite a significant impact on search times (when searching for an email address for example – there are 1.5 million rows in the table).

Is there a more efficient way of achieving this type of constraint?

Answer 1

but I found that this had quite a significant impact on search times
(when searching for an email address for example

The index you defined on (DataSetId, Email) cannot be used for searches based on email. If you would create an index with the Email field at the leftmost position, it could be used:

CREATE UNIQUE NONCLUSTERED INDEX IX_Subscriber_Email
   ON Subscriber (Email, DataSetId);

This index would server both as a unique constraint enforcement and as a means to quickly search for an email. This index though cannot be used to quickly search for a specific DataSetId.

The gist of it if is that whenever you define a multikey index, it can be used only for searches in the order of the keys. An index on (A, B, C) can be used to seek values on column A, for searching values on both A and B or to search values on all three columns A, B and C. However it cannot be used to search values on B or on C alone.