Cfrtjryk

I have a Person table that has a created_by column that references the primary id of the table itself. So, it could be an employee that adds another employee to the database. It works fine.

But People can also add themselves (signup). So the value in the created_by column should be the auto-incremented value of the id column. But that value is obviously not available until after the insert.

So I could either (a) make the reference not to check the values, (b) add a default value in the beginning or (c) make the column nullable. All options seem bad to me.

The MySQL's dialect has this:

SET FOREIGN_KEY_CHECKS = 0;

INSERT INTO EMPLOYEE VALUES('12345','67890'),('67890','12345');

SET FOREIGN_KEY_CHECKS = 1;

...but I could not find something similar for SQL Server's T-SQL.

It looks like you are using an Identity for your Primary Key. If you need flexibility with your primary key I would recommend using a sequence. it would look something like this.

CREATE SEQUENCE SQ_temp AS INT INCREMENT BY 1 START WITH 1



CREATE TABLE Users 

( ID INT PRIMARY KEY DEFAULT NEXT VALUE FOR SQ_temp

, UserName VARCHAR(30)

, createdBy INT REFERENCES Users (ID)

)





INSERT INTO dbo.Users

(

    ID

    , UserName

    , createdBy

)

VALUES

(NEXT VALUE FOR SQ_temp,'User1', NEXT VALUE FOR SQ_temp)

, (NEXT VALUE FOR SQ_temp,'User2', NEXT VALUE FOR SQ_temp)

, (NEXT VALUE FOR SQ_temp,'User3', NEXT VALUE FOR SQ_temp)





SELECT *

FROM dbo.Users AS u

Insert self referencing entry into SQL server

For the general question as per the title, you can add a direct circular reference in a simple insert such as

INSERT node 

       (id , name  , parent_id) 

VALUES (123, 'Test', 123      )

because the constraint is enforced considering all the new data: as long as the value the FX references exists once the statement is complete all is well. This is the same as inserting several values in a linked list:

INSERT node 

       (id , name   , parent_id) 

VALUES (101, 'Test1', 100      )

     , (102, 'Test2', 101      )

     , (103, 'Test3', 102      )

or an indirect circular reference:

INSERT node 

       (id , name   , parent_id) 

VALUES (201, 'Test5', 203      )

     , (202, 'Test6', 201      )

     , (203, 'Test7', 202      )

So the value in the created_by column should be the auto-incremented value of the id column

This poses a problem because you don't know what the generated is before you insert, in fact there is no reliable way of knowing. For inserting individual rows you can use SCOPE_IDENTITY() to immediately update the new row:

INSERT node 

       (name  ) 

VALUES ('Test')

-- and now make the circular reference

UPDATE node 

SET    parent_id = SCOPE_IDENTITY()  

WHERE  id        = SCOPE_IDENTITY()

In the above example if parent_id is a required column (declared NOT NULL with no DEFAULT) then you'll need to provide a dummy temporary valid value instead of leaving it out of the initial INSERT statement, like so:

INSERT node 

       (name  , parent_id) 

VALUES ('Test', 0        )

-- and now make the circular reference

UPDATE node 

SET    parent_id = SCOPE_IDENTITY()  

WHERE  id        = SCOPE_IDENTITY()

For dealing with inserts of multiple rows (unlikely in the circumstance you describe but common elsewhere) you can use the OUTPUT clause to read the IDs created for each row for further reference. I'll not go over that here as it is overkill for the current question, see https://docs.microsoft.com/en-us/sql/t-sql/queries/output-clause-transact-sql for more detail.

What about temporaryly disabling the FK Constraint check?

CREATE TABLE dbo.person (id int NOT NULL PRIMARY KEY, createdby int  CONSTRAINT FK_TEST FOREIGN KEY REFERENCES dbo.person(id))



ALTER TABLE dbo.PERSON

NOCHECK CONSTRAINT FK_TEST



INSERT INTO  dbo.person(id, createdby) values (1,2)

INSERT INTO  dbo.person(id, createdby) values (2,1)



    ALTER TABLE dbo.PERSON

WITH CHECK CHECK CONSTRAINT FK_TEST

Result

SELECT * FROM dbo.person



id  createdby

1   2

2   1