trigger to create parent element and retrieve id in postgresql - sql

Question

I have created the following tables:
CREATE TABLE IF NOT EXISTS public.teams (
id SERIAL PRIMARY KEY,
name VARCHAR(64) NOT NULL UNIQUE
) WITH (OIDS = FALSE);
CREATE TABLE IF NOT EXISTS public.submissions (
id SERIAL PRIMARY KEY,
team_id INTEGER REFERENCES public.teams NOT NULL,
records_num INTEGER NOT NULL,
timestamp TIMESTAMP NOT NULL
) WITH (OIDS = FALSE);
CREATE TABLE IF NOT EXISTS public.predictions (
id SERIAL PRIMARY KEY,
submission_id INTEGER REFERENCES public.submissions NOT NULL,
customer INTEGER REFERENCES public.real NOT NULL,
date DATE NOT NULL,
billing NUMERIC(20, 2) NOT NULL
) WITH (OIDS = FALSE);
CREATE TABLE IF NOT EXISTS public.real (
customer INTEGER PRIMARY KEY,
date DATE NOT NULL,
billing NUMERIC(20, 2) NOT NULL
) WITH (OIDS = FALSE);
The relation for submissions-predictions is one-to-many; users will submit predictions in packets of 1000 rows that should get the same submission id.
I am trying to create a trigger that runs BEFORE INSERT ON predictions that creates a submissions row. This is what I have so far:
CREATE OR REPLACE FUNCTION insert_submission() RETURNS TRIGGER AS
$$
BEGIN
INSERT INTO submissions(team_id, records_num, timestamp)
VALUES (1, 1, '2018-04-21 00:00:00'); /*example values, need to fill with dynamically assigned ones, specially for records_num and team_id*/
RETURN NULL;
END
$$ LANGUAGE plpgsql;
DROP TRIGGER trigger_submission ON public.predictions;
CREATE TRIGGER trigger_submission BEFORE INSERT ON predictions
EXECUTE PROCEDURE insert_submission();
So, my questions are:
How do I go about retrieving the newly created submissions.id for the row inserted by the trigger, in order to add it to all the rows inserted in predictions by the user? Do I have to run another trigger AFTER INSERT for this?
EDIT: to clarify following #bignose answer, the sequence of events would go like this:
User inserts 1000 rows into public.predictions:
INSERT INTO predictions(customer, date, billing)
VALUES
(1, '2018-01-05', 543.42),
(4, '2018-04-02', 553.21),
...
(423, '2019-11-18', 38.87) /* 1000th row */
He does not know which submission_id to insert in those rows and indeed, the submissions row for this packet of predictions doesn't exist yet so a trigger runs before to create a row in submissions that would execute something like this:
INSERT INTO public.submisssions(team_id, records_num, timestamp)
VALUES (
4, /* I will need something to retrieve team_id here */
1000, /* I will need something to count the rows of the insert that triggered this */
NOW() /* convert to timestamp */
)
This last query should return the public.submission.id value that it has just created to the insert the user requested so that it ends up being something like this:
INSERT INTO predictions(customer, date, billing)
VALUES
(#submission_id, 1, '2018-01-05', 543.42),
(#submission_id, 4, '2018-04-02', 553.21),
...
(#submission_id, 423, '2019-11-18', 38.87) /* 1000th row */
Where #submission_id should be the value retrieved from the trigger (and the some for all the 1000 rows)
How could I count the rows inserted by the user to use them as value for submissions.records_num?
How could I retrieve team.id to insert during the trigger execution, assuming I know team.name beforehand?
Thank you!
Kind regards

Answer 1

A trigger function, when used for a row-level trigger, has access to the old and new state of the table.
CREATE OR REPLACE FUNCTION insert_submission() RETURNS TRIGGER AS
$$
BEGIN
INSERT INTO submissions(team_id, records_num, timestamp)
VALUES (NEW.foo, NEW.bar, '2018-04-21 00:00:00');
RETURN NULL;
END
$$ LANGUAGE plpgsql;
It's not clear from the description, which fields you expect to retrieve from the row that triggers this function. So you'll need to substitute NEW.foo and NEW.bar with field references in the NEW row state.