SQL Insert into 2 tables, passing the new PK from one table as the FK in the other

Question

How can I achieve an insert query on 2 tables that will insert the primary key set from one table as a foreign key into the second table.

Here's a quick example of what I'm trying to do, but I'd like this to be one query, perhaps a join.

INSERT INTO Table1 (col1, col2) VALUES ( val1, val2 )
INSERT INTO Table2 (foreign_key_column) VALUES (parimary_key_from_table1_insert)

I'd like this to be one join query. I've made some attempts but I can't get this to work correctly.

Answer 1

This is not possible to do with a single query.

The record in the PK table needs to be inserted before the new PK is known and can be used in the FK table, so at least two queries are required (though normally 3, as you need to retrieve the new PK value for use).

The exact syntax depends on the database being used, which you have not specified.

If you need this set of inserts to be atomic, use transactions.

Answer 2

Despite what others have answered, this absolutely is possible, although it takes 2 queries made consecutively with the same connection (to maintain the session state).

Here's the mysql solution (with executable test code below):

INSERT INTO Table1 (col1, col2) VALUES ( val1, val2 );
INSERT INTO Table2 (foreign_key_column) VALUES (LAST_INSERT_ID());

Note: These should be executed using a single connection.

Here's the test code:

create table tab1 (id int auto_increment primary key, note text);
create table tab2 (id int auto_increment primary key, tab2_id int references tab1, note text);
insert into tab1 values (null, 'row 1');
insert into tab2 values (null, LAST_INSERT_ID(), 'row 1');
select * from tab1;
select * from tab2;
mysql> select * from tab1;
+----+-------+
| id | note  |
+----+-------+
|  1 | row 1 |
+----+-------+
1 row in set (0.00 sec)

mysql> select * from tab2;
+----+---------+-------+
| id | tab2_id | note  |
+----+---------+-------+
|  1 |       1 | row 1 |
+----+---------+-------+
1 row in set (0.00 sec)

Answer 3

There may be a few ways to accomplish this. Probably the most straight forward is to use a stored procedure that accepts as input all the values you need for both tables, then inserts to the first, retrieves the PK, and inserts to the second.

If your DB supports it, you can also tell the first INSERT to return a value:

INSERT INTO table1... RETURNING primary_key;

This at least saves the SELECT step that would otherwise be necessary. If you go with a stored procedure approach, you'll probably want to incorporate this into that stored procedure.

It could also possibly be done with a combination of views and triggers--if supported by your DB. This is probably far messier than it's worth, though. I believe this could be done in PostgreSQL, but I'd still advise against it. You'll need a view that contains all of the columns represented by both table1 and table2 , then you need an ON INSERT DO INSTEAD trigger with three parts--the first part inserts to the new table, the second part retrieves the PK from the first table and updates the NEW result, and the third inserts to the second table. (Note: This view doesn't even have to reference the two literal tables, and would never be used for queries--it only has to contain columns with names/data types that match the real tables)

Of course all of these methods are just complicated ways of getting around the fact that you can't really do what you want with a single command.

Answer 4

From your example, if the tuple (col1, col2) can be considered unique, then you could do:

INSERT INTO table1 (col1, col2) VALUES (val1, val2);
INSERT INTO table2 (foreign_key_column) VALUES (SELECT id FROM Table1 WHERE col1 = val1 AND col2 = val2);