Question

I am trying to find a way to increment a second primary key column in a table automatically when a new entry is added for the first primary key column. I suppose an example would be best here so here goes.

Suppose I have a table:

CREATE TABLE T
(
     SecNum INT NOT NULL,
     EntryID INT NOT NULL,
     Value FLOAT,
) CONSTRAINT [PK_T] PRIMARY KEY CLUSTERED 
(
    [SecNum] ASC,
    [EntryID] ASC
)

I would run the following statement:

INSERT INTO T (SecNum, Value) VALUES (0, 10)

My table should look like:

SECNUM | ENTRYID | VALUE
-------------------------
   0         0       10

I would run the following statement:

INSERT INTO T (SecNum, Value) VALUES (0, 10)

My table should look like:

SECNUM | ENTRYID | VALUE
-------------------------
   0         0       10
   0         1       10

I would run the following statement:

INSERT INTO T (SecNum, Value) VALUES (1, 20)

My table should look like:

SECNUM | ENTRYID | VALUE
-------------------------
   0         0       10
   0         1       10
   1         0       20

Answer 1

This is possible using an INSTEAD OF trigger:

CREATE TRIGGER TriggerName
ON T
INSTEAD OF INSERT
AS
    -- THIS TOP BIT IS OPTIONAL, IT WILL ALLOW ENTRY ID TO BE OVERRIDDEN IF 
    -- IT IS SUPPLIED TO THE INSERT AND WILL NOT VIOLATE THE PRIMARY KEY
    IF NOT EXISTS 
        (   SELECT  1
            FROM    T
                    INNER JOIN inserted i
                        ON i.SecNum = T.secNum
                        AND i.EntryID = T.EntryID
            UNION
            SELECT  1
            FROM    inserted
            WHERE   EntryID IS NULL
        )
        BEGIN
            INSERT T (SecNum, EntryID, Value)
            SELECT  SecNum, EntryID, Value
            FROM    inserted
        END
    ELSE
    -- IF OVERRIDE ABILITY IS NOT REQUIRED JUST USE THE BELOW INSERT
        BEGIN
            INSERT T (SecNum, EntryID, Value)
            SELECT  i.SecNum, COALESCE(LastID, 0), i.Value
            FROM    inserted I
                    LEFT JOIN 
                    (   SELECT  SecNum, MAX(T.EntryID) + 1 [LastID]
                        FROM    T
                        GROUP BY SecNum
                    ) T
                        ON T.SecNum = i.SecNum

        END

Example here

HOWEVER this is not very elegant. It could be worth asking is it really necessary? Could you get away with using a surrogate primary key, and use ROW_NUMBER() to create Entry ID’s on the fly?