[sql] IF EXISTS, THEN SELECT ELSE INSERT AND THEN SELECT

How do you say the following in Microsoft SQL Server 2005:

IF EXISTS (SELECT * FROM Table WHERE FieldValue='') THEN
   SELECT TableID FROM Table WHERE FieldValue=''
ELSE
   INSERT INTO TABLE(FieldValue) VALUES('')
   SELECT TableID FROM Table WHERE TableID=SCOPE_IDENTITY()
END IF

What I'm trying to do is to see if there is a blank fieldvalue already, and if there is then return that TableID, else insert a blank fieldvalue and return the corresponding primary key.

You need to do this in transaction to ensure two simultaneous clients won't insert same fieldValue twice:

SET TRANSACTION ISOLATION LEVEL SERIALIZABLE
BEGIN TRANSACTION
    DECLARE @id AS INT
    SELECT @id = tableId FROM table WHERE fieldValue=@newValue
    IF @id IS NULL
    BEGIN
       INSERT INTO table (fieldValue) VALUES (@newValue)
       SELECT @id = SCOPE_IDENTITY()
    END
    SELECT @id
COMMIT TRANSACTION

you can also use Double-checked locking to reduce locking overhead

DECLARE @id AS INT
SELECT @id = tableID FROM table (NOLOCK) WHERE fieldValue=@newValue
IF @id IS NULL
BEGIN
    SET TRANSACTION ISOLATION LEVEL SERIALIZABLE
    BEGIN TRANSACTION
        SELECT @id = tableID FROM table WHERE fieldValue=@newValue
        IF @id IS NULL
        BEGIN
           INSERT INTO table (fieldValue) VALUES (@newValue)
           SELECT @id = SCOPE_IDENTITY()
        END
    COMMIT TRANSACTION
END
SELECT @id

As for why ISOLATION LEVEL SERIALIZABLE is necessary, when you are inside a serializable transaction, the first SELECT that hits the table creates a range lock covering the place where the record should be, so nobody else can insert the same record until this transaction ends.

Without ISOLATION LEVEL SERIALIZABLE, the default isolation level (READ COMMITTED) would not lock the table at read time, so between SELECT and UPDATE, somebody would still be able to insert. Transactions with READ COMMITTED isolation level do not cause SELECT to lock. Transactions with REPEATABLE READS lock the record (if found) but not the gap.

You just have to change the structure of the if...else..endif somewhat:

if exists(select * from Table where FieldValue='') then begin
  select TableID from Table where FieldValue=''
end else begin
  insert into Table (FieldValue) values ('')
  select TableID from Table where TableID = scope_identity()
end

You could also do:

if not exists(select * from Table where FieldValue='') then begin
  insert into Table (FieldValue) values ('')
end
select TableID from Table where FieldValue=''

Or:

if exists(select * from Table where FieldValue='') then begin
  select TableID from Table where FieldValue=''
end else begin
  insert into Table (FieldValue) values ('')
  select scope_identity() as TableID
end

IF EXISTS (SELECT 1 FROM Table WHERE FieldValue='') 
BEGIN
   SELECT TableID FROM Table WHERE FieldValue=''
END
ELSE
BEGIN
   INSERT INTO TABLE(FieldValue) VALUES('')
   SELECT SCOPE_IDENTITY() AS TableID
END

See here for more information on IF ELSE

Note: written without a SQL Server install handy to double check this but I think it is correct

Also, I've changed the EXISTS bit to do SELECT 1 rather than SELECT * as you don't care what is returned within an EXISTS, as long as something is I've also changed the SCOPE_IDENTITY() bit to return just the identity assuming that TableID is the identity column

DECLARE @t1 TABLE (
    TableID     int         IDENTITY,
    FieldValue  varchar(20)
)

--<< No empty string
IF EXISTS (
    SELECT *
    FROM @t1
    WHERE FieldValue = ''
) BEGIN
    SELECT TableID
    FROM @t1
    WHERE FieldValue=''
END
ELSE BEGIN
    INSERT INTO @t1 (FieldValue) VALUES ('')
    SELECT SCOPE_IDENTITY() AS TableID
END

--<< A record with an empty string already exists
IF EXISTS (
    SELECT *
    FROM @t1
    WHERE FieldValue = ''
) BEGIN
    SELECT TableID
    FROM @t1
    WHERE FieldValue=''
END
ELSE BEGIN
    INSERT INTO @t1 (FieldValue) VALUES ('')
    SELECT SCOPE_IDENTITY() AS TableID
END

It sounds like your table has no key. You should be able to simply try the INSERT: if it’s a duplicate then the key constraint will bite and the INSERT will fail. No worries: you just need to ensure the application doesn't see/ignores the error. When you say 'primary key' you presumably mean IDENTITY value. That's all very well but you also need a key constraint (e.g. UNIQUE) on your natural key.

Also, I wonder whether your procedure is doing too much. Consider having separate procedures for 'create' and 'read' actions respectively.

You were close:

IF EXISTS (SELECT * FROM Table WHERE FieldValue='')
   SELECT TableID FROM Table WHERE FieldValue=''
ELSE
BEGIN
   INSERT INTO TABLE (FieldValue) VALUES ('')
   SELECT TableID FROM Table WHERE TableID=SCOPE_IDENTITY()
END

create schema tableName authorization dbo
go
IF OBJECT_ID ('tableName.put_fieldValue', 'P' ) IS NOT NULL 
drop proc tableName.put_fieldValue
go
create proc tableName.put_fieldValue(@fieldValue int) as
declare @tableid int = 0
select @tableid = tableid from table where fieldValue=''
if @tableid = 0 begin
   insert into table(fieldValue) values('')
   select @tableid = scope_identity()
end
return @tableid
go
declare @tablid int = 0
exec @tableid = tableName.put_fieldValue('')