PostgreSQL multi INSERT...RETURNING with multiple columns

Question

I`m building a database with Postgres 9.3 as a backend, having 3 tables:

table1 (user_id, username, name, surname, emp_date)
table2 (pass_id, user_id, password)
table3 (user_dt_id, user_id, adress, city, phone)

As can be seen table2 and table3 are child tables of table1.
I can extract the user_id of a newly inserted row in table1 (parent):

INSERT INTO "table1" (default,'johnee','john','smith',default) RETURNING userid;

I need to insert the newly extracted id (from table1) into user_id columns of table2 and table3 along with other data unique for those tables. Basically 3 X INSERT ...
How do I do that?

Answer 1

Use data-modifying CTEs to chain your three INSERTs. Something like this:

WITH ins1 AS (
   INSERT INTO table1 (username, name,  surname)
   VALUES ('johnee','john','smith')
   RETURNING user_id
   )
, ins2 AS (
   INSERT INTO table2 (user_id, password)
   SELECT ins1.user_id, 'secret'
   FROM   ins1                            -- nothing to return here
   )
INSERT INTO table3 (user_id, adress, city, phone)
SELECT ins1.user_id, ...
FROM   ins1
RETURNING user_id;

• It's typically best to add a column definition list for INSERTs (except for special cases). Else, if the table structure changes, your code might break in surprising ways.

• I omitted columns where you would just enter DEFAULT. Defaults are inserted automatically. Shorter, same result.

• The final, optional RETURNING returns the resulting user_id - obviously from a sequence or some other default. It's actually the user_id from table3, but that's the same unless you have some triggers or other magic interfering.

More about data-modifying (a.k.a. "writable") CTEs:

• Are SELECT type queries the only type that can be nested?