Question

Let’s say have have two tables, A and B.
A contains fields: id, a1, a2.
B contains fields: id, b1, b2.

Asume a one to many relationship from table A to B based on the id field.

Now let’s say I want to create a unique constraint on table b that prohibits me from inserting duplicate records based on b1, b2, and a1. Is that possible?

Answer 1

You would have to have a copy of a1 appear in B in order to enforce this as a constraint. If you don’t want this visible to your users, you would create a table (say, _B) that contains all of Bs columns + a1, and then create a view (with suitable insert triggers to populate a1) called B that hides this column.

You should also create a superkey of A over id and a1, and use a foreign key constraint (possibly with ON UPDATE CASCADE) from B over both of these columns to ensure that this copied column correctly matches the corresponding value in A.

Whether you then decide that this should be the only foreign key, or continue to maintain the “correct” one (over just id) in addition is a matter of taste.

---

E.g.:

CREATE TABLE dbo.A (ID int not null,a1 int not null,a2 int not null,
    constraint PK_A PRIMARY KEY (id),
    constraint UQ_A_a1_check UNIQUE (id,a1)
)
CREATE TABLE dbo._B (ID int not null,a1 int not null,b1 int not null,b2 int not null,
    constraint PK_B PRIMARY KEY (a1,b1,b2),
    constraint FK_B_A FOREIGN KEY (id) references A (id),
    constraint DRI_B_A_a1_check FOREIGN KEY (id,a1) references A (id,a1) on update cascade
)
GO
CREATE VIEW dbo.B
WITH SCHEMABINDING
AS
    select id,b1,b2 from dbo._B
GO
CREATE TRIGGER T_B_I on dbo.B instead of insert
AS
    INSERT INTO dbo._B (id,a1,b1,b2)
    SELECT i.id,a.a1,i.b1,i.b2
    FROM
       inserted i
          inner join
       dbo.A a
          on
             i.id = a.id
GO