Inserting the Identity of another Insert from Select? - sql

Question

Is This Possible?
Here is something I'm looking for, executed together:
First, it would execute the INSERT based on how many rows in the SELECT
INSERT INTO TABLE2 (xID, NAME)
SELECT xID, NAME FROM TABLE
Then getting the ##IDENTITY of each INSERTED ROW, it would create a new Insert including the same data of the first SELECT statement:
INSERT INTO TABLE3 (xID, NAME, ID)
SELECT xID, NAME, ID as Scope_IdentitY()
If not, what the best way without using cursor or while?

Answer 1

You have, at least, two options:
1) The OUTPUT...INTO target_table clause (SQL2005+)
2) Or you could write composable DML(SQL2008+).
Example:
DECLARE #Table2 TABLE(
ID INT IDENTITY PRIMARY KEY, --IDENTITY
xID INT NOT NULL,
NAME VARCHAR(25) NOT NULL
);
DECLARE #Table3 TABLE(
ID INT PRIMARY KEY, --No IDENTITY
xID INT NOT NULL,
NAME VARCHAR(25) NOT NULL
);
--First solution: OUTPUT ... INTO
INSERT INTO #Table2 (xID, NAME)
OUTPUT inserted.xID, inserted.NAME, inserted.ID INTO #Table3(xID, NAME, ID)
SELECT t.Col1, t.Col2
FROM (SELECT 11,'A' UNION ALL SELECT 22,'B' UNION ALL SELECT 33,'C') AS t(Col1,Col2);
--Second solution: composable DML
INSERT INTO #Table3(xID, NAME, ID)
SELECT src.xID, src.NAME, src.ID
FROM
(
INSERT INTO #Table2 (xID, NAME)
OUTPUT inserted.xID, inserted.NAME, inserted.ID
SELECT t.Col1, t.Col2
FROM (VALUES(44,'D'),(55,'E'),(66,'F')) AS t(Col1,Col2)
) src
SELECT * FROM #Table2
SELECT * FROM #Table3

Answer 2

INSERT INTO TABLE2 (xID, NAME)
OUTPUT
INSERTED.xID, INSERTED.NAME, INSERTED.ID
INTO TABLE3 (xID, NAME, ID)
SELECT xID, NAME FROM [TABLE]

Answer 3

You can declare a table variable and store the output of the rows inserted into dbo.Table2 in this variable and use the table variable as the input for table dbo.Table3.
CREATE TABLE dbo.Table1
(
xid int NOT NULL
, name varchar(30) NOT NULL
);
CREATE TABLE dbo.Table2
(
id int NOT NULL IDENTITY
, xid int NOT NULL
, name varchar(30) NOT NULL
);
CREATE TABLE dbo.Table3
(
id int NOT NULL
, xid int NOT NULL
, name varchar(30) NOT NULL
);
INSERT INTO dbo.Table1 (xid, name) VALUES
(195, 'abc'),
(242, 'def'),
(332, 'ghi');
GO
DECLARE #tempTable table
( id int
, xid int
, name varchar(30)
);
INSERT dbo.Table2
OUTPUT INSERTED.id, INSERTED.xid, INSERTED.name
INTO #tempTable
SELECT xid, name FROM dbo.Table1;
INSERT dbo.Table3 (id, xid, name)
SELECT id, xid, name FROM #tempTable;
SELECT id, xid, name FROM dbo.Table2;
SELECT id, xid, name FROM dbo.Table3;
GO

Answer 4

OK, based on your comments below, try this:
INSERT INTO TABLE2 (xID, NAME)
SELECT xID, NAME FROM TABLE;
INSERT INTO TABLE3 (xID, NAME, ID)
SELECT xID, NAME, ##identity
FROM TABLE2;

Answer 5

Assuming these table structures:
TABLE_A
-----------
X_ID
NAME
TABLE_B
----------------
TABLE_B_ID [PK]
X_ID
NAME
TABLE_C
----------------
TABLE_C_ID [PK]
X_ID
NAME
TABLE_B_ID [FK]
Then wouldn't this work (best in a transaction)?:
-- Grab data from TABLE_A and INSERT INTO TABLE_B
INSERT INTO TABLE_B (
X_ID,
NAME
)
SELECT
X_ID,
NAME
FROM
TABLE_A
-- Grab data from TABLE_B that matches the data imported from TABLE_A
-- and INSERT that data into TABLE_C (incl. the PK from TABLE_B)
INSERT INTO TABLE_C (
X_ID,
NAME,
TABLE_B_ID
)
SELECT
b.X_ID,
b.NAME,
b.TABLE_B_ID
FROM
TABLE_B b
INNER JOIN
TABLE_A a ON a.X_ID = b.X_ID