如何将自动生成的主键ID保存在同一表的外键列中

Question

Following is the table structure:

CREATE TABLE [User] (
        [Id] bigint identity(1,1) not null,
        [FirstName] nvarchar(100) not null,
        [LastName] nvarchar(100) not null,
        [Title] nvarchar(5) null,
        [UserName] nvarchar(100) not null,
        [Password] nvarchar(100) not null,      
        [Inactive] bit null,
        [Created] Datetime not null,
        [Creator] bigint not null,
        [Modified] DateTime null,
        [Modifier] bigint null
        CONSTRAINT [PK_User] PRIMARY KEY CLUSTERED
        (
            [Id] Asc
        )
    );

IF NOT EXISTS (SELECT * FROM sys.foreign_keys WHERE object_id = OBJECT_ID(N'[FK_User_Creator]') AND parent_object_id = OBJECT_ID(N'[User]'))
    ALTER TABLE [User] ADD CONSTRAINT [FK_User_Creator] FOREIGN KEY([Creator]) REFERENCES [User]([Id])
GO


INSERT INTO [User] (Creator) Values ([Id] ?)

This is a case when table is empty and first user is going to add in table. Otherwise I don't have issue.

How can I insert Id in creator column with insert statement at the same time?

Answer 1

One way could be using Sequence instead of identity column. The below script might serve the same purpose:

CREATE SEQUENCE dbo.useridsequence  
    AS int  
    START WITH 1  
    INCREMENT BY 1 ;  
GO

CREATE TABLE [User] (
        [Id] bigint DEFAULT (NEXT VALUE FOR dbo.useridsequence) ,
        [FirstName] nvarchar(100) not null,
        [LastName] nvarchar(100) not null,
        [Title] nvarchar(5) null,
        [UserName] nvarchar(100) not null,
        [Password] nvarchar(100) not null,      
        [Inactive] bit null,
        [Created] Datetime not null,
        [Creator] bigint DEFAULT NEXT VALUE FOR dbo.useridsequence ,
        [Modified] DateTime null,
        [Modifier] bigint null
        CONSTRAINT [PK_User] PRIMARY KEY CLUSTERED
        (
            [Id] Asc
        )
    );

IF NOT EXISTS (SELECT * FROM sys.foreign_keys WHERE object_id = OBJECT_ID(N'[FK_User_Creator]') AND parent_object_id = OBJECT_ID(N'[User]'))
    ALTER TABLE [User] ADD CONSTRAINT [FK_User_Creator] FOREIGN KEY([Creator]) REFERENCES [User]([Id])
GO


INSERT INTO [User]
(
    -- Id -- this column value is auto-generated
    FirstName,
    LastName,
    Title,
    UserName,
    [Password],
    Inactive,
    Created,
    Creator,
    Modified,
    Modifier
)
VALUES
(
    'Foo',
    'Bar',
    'Title',
    'UserName ',
    'Password',
     0,
    GETDATE(),
    DEFAULT,
    GETDATE(),
    1 
)

SELECT * FROM [User] AS u

Result :

Answer 2

The short answer is that you can't do this. And I suggest your model is logically flawed in the first place. Do you intend to define all actual database users (eg, create user ... for login ...) as rows in [Users]? You need to think about that - but the typical answer is no. If the answer is yes, then you don't need the creator column at all because it is redundant. All you need is the created date - for which you probably should have defined a default.

But if you want to do this, you will need to do it in two steps (and you will need to make the column nullable). You insert a row (or rows) with values for the "real" data columns. Then update those same rows with the identity values generated for id. An example showing different ways to do this

use tempdb;
set nocount on;
CREATE TABLE dbo.[user] (
        [user_id] smallint identity(3,10) not null primary key,
        [name] nvarchar(20) not null,
        [active] bit  not null default (1),
        [created] Datetime not null default (current_timestamp),
        [creator] smallint  null  
    );
ALTER TABLE dbo.[user] ADD CONSTRAINT [fk_user] FOREIGN KEY(creator) REFERENCES dbo.[user](user_id);
GO
-- add first row
insert dbo.[user] (name) values ('test');
update dbo.[user] set creator = SCOPE_IDENTITY() where user_id = SCOPE_IDENTITY()

-- add two more rows
declare @ids table (user_id smallint not null);
insert dbo.[user] (name) output inserted.user_id into @ids
values ('nerk'), ('pom');
update t1 set creator = t1.user_id
from @ids as newrows inner join dbo.[user] as t1 on newrows.user_id = t1.user_id;
select * from dbo.[user] order by user_id;

-- mess things up a bit
delete dbo.[user] where name = 'pom';

-- create an error, consume an identity value
insert dbo.[user](name) values (null);

-- add 2 morerows
delete @ids; 
insert dbo.[user] (name) output inserted.user_id into @ids
values ('nerk'), ('pom');
update t1 set creator = t1.user_id
from @ids as newrows inner join dbo.[user] as t1 on newrows.user_id = t1.user_id;

select * from dbo.[user] order by user_id;
drop table dbo.[user];

And I changed the identity specification to demonstrate something few developers realize. It isn't always defined as (1,1) and the next inserted value can jump for many reasons - errors and caching/restarts for example. Lastly, I think you will regret naming a table with a reserved word since references to it will require the use of delimiters. Reduce the pain.