execute sql inside plpgsql function

Question

Using postgreSQL 9.1.9, I have some SQL requests like this:

 INSERT INTO "my_table" VALUES(10,'James','California');

Below a simplified table of the original names lookup table...

  names

  name_id      name
  ---          -----
  3            James

but in fact I don't have to enter (into "my_table") the text value provided by the SQL request (name of person, name of the state) but its corresponding ID located in another table (ex: names table)

So i was thinking on creating a trigger, calling a function that should execute the select SQL request and return the modified row to be inserted.

Is that possible?

CREATE FUNCTION insert_by_ID() RETURNS TRIGGER AS '

        BEGIN
            --Can we execute this query, store the result in a variable to be added
              to the NEW record?

            SELECT name_id from names where name=NEW.name;

                           --this request would be...
                           --SELECT name_id from names where name='James'     
          RETURN NEW;
        END;
        ' LANGUAGE plpgsql;

CREATE TRIGGER insert_info_ID
        BEFORE INSERT ON my_table
        FOR EACH ROW
        EXECUTE PROCEDURE insert_by_ID();

It is possible, of course, but may be redundant. Complete your question with *my_table* and *names* definitions. — klin, May 14 '14 at 16:59
Run `\d names` in psql and post the output. Plus, as *always* your version of Postgres. And please clarify the fuzzy explanation. — Erwin Brandstetter, May 14 '14 at 17:07
Your question is still unclear. I can guess you probably need an updateable view. Please read this article: http://vibhorkumar.wordpress.com/2011/10/28/instead-of-trigger/ — klin, May 15 '14 at 12:32
Are we dealing with names that can *always* be found in table `names` or do you want to insert a name if not found and proceed with the newly created `name_id`? If so, are we dealing with heavy concurrent write load? — Erwin Brandstetter, May 15 '14 at 21:29

score 1 · Accepted Answer · edited Jun 20 '20 at 09:12

My educated guess is you want a solution for the "SELECT or INSERT if not exists" problem.

I suggest you create a plpgsql function like this:

CREATE OR REPLACE FUNCTION f_name_id(_name text, OUT _name_id int) AS
$func$
BEGIN

LOOP
   BEGIN

   SELECT name_id FROM names WHERE name = _name FOR SHARE   -- lock!
   INTO  _name_id;

   IF NOT FOUND THEN
      INSERT INTO names (name) VALUES (_name) RETURNING name_id
      INTO  _name_id;
   END IF;

   EXCEPTION WHEN UNIQUE_VIOLATION THEN      -- inserted concurrently!
      RAISE NOTICE 'It actually happened!';  -- hardly ever happens
   END;

   EXIT WHEN _name_id IS NOT NULL;           -- else try again
END LOOP;

END
$func$ LANGUAGE plpgsql;

This is a simple version of (more explanation and links over there):
Is SELECT or INSERT in a function prone to race conditions?

INSERT

Then your INSERT statement can look like this:

INSERT INTO my_table(my_table_id, name_id, state) -- always with column list
VALUES (10, f_name_id('James'), 'California');

execute sql inside plpgsql function

1 Answers1

INSERT

Linked