Question

I have two tables, A and B. The structure looks like this:

CREATE TABLE A (
    w int NOT NULL,
    x int NOT NULL,
    y int NOT NULL,
    CONSTRAINT PK_A PRIMARY KEY (w, x, y)
)

CREATE TABLE B (
    w int NOT NULL,
    y int NULL,
    z int NOT NULL
)

I want to make sure for any set of values entered in table B, that w and y are in table A. If the y value in table B is null, I only want to make sure that w is in table A.

Some sample data, inserts, and expected results:

Table A
w  x  y
----------
1  1  1
1  1  2
1  2  1
1  3  2
2  1  1

INSERT INTO B (w, y, z) VALUES (1, 1, 3) -- good
INSERT INTO B (w, y, z) VALUES (1, NULL, 3) -- good
INSERT INTO B (w, y, z) VALUES (1, 1, 4) -- good
INSERT INTO B (w, y, z) VALUES (2, NULL, 3) -- good
INSERT INTO B (w, y, z) VALUES (1, 3, 1) -- fail
INSERT INTO B (w, y, z) VALUES (3, NULL, 1) -- fail

Any way for this to work? I’m using SQL Server 2000 if that comes into play.

Answer 1

Unfortunately, you cannot use a foreign key constraint on B.w and B.y because they would reference non-unique columns on A. But you can add this check via triggers:

create trigger check_w on B for insert, update
as
if not exists(select * from A join inserted on A.w = inserted.w) 
begin
  raiserror('W not in A!', 1, 1)
  rollback transaction
end

GO

create trigger check_y on B for insert, update
as
if 
(select y from inserted) is not null and 
not exists(select * from A join inserted on A.y = inserted.y) 
begin
  raiserror('Y not null and not in A!', 1, 1)
  rollback transaction
end

GO

You can definitely combine these two triggers in one.

Also note that you’ll need a trigger on A for the delete operation. You either prevent deletion if there’re matching rows on B, or you perform a cascade delete operation.