Query Performace

Question

Hi,
The below two queries will do same work. i want to know which one is performance low.
UPDATE (SELECT p.ph_country_code,c.country_code
               FROM address a,
                    phone p,
                    country c
              WHERE a.user_id = p.user_id
                AND a.country_id = c.country_id) u
        SET u.ph_country_code = u.country_code;
DECLARE
  v_ph_country_code  country.country_code%TYPE;
  CURSOR cur_ph IS
      SELECT user_id
        FROM phone;
BEGIN
  FOR rec_ph IN cur_ph LOOP
      SELECT c.country_code
        INTO v_ph_country_code
        FROM address a,
             country c
       WHERE a.country_id = c.country_id
         AND a.user_id    = rec_ph.user_id;
      UPDATE phone
         SET ph_country_code   = v_ph_country_code
       WHERE user_id            = rec_ph.user_id;
  END LOOP;
END;   Thank you,
Vi

Answer 1

GHD wrote:
The below two queries will do same work. i want to know which one is performance low.Let's say there are a substantial volume of data - and the CBO can parallelise access.
This means that (under the right circumstances) your native SQL can be executed using parallel processing (Oracle supports both parallel DML and DDL).
The PL/SQL code is your manual attempt at doing the same thing as that single SQL statement. How are you going to scale it for processing huge volumes in parallel? The PL/SQL engine cannot do that for you. The SQL your code generates for the SQL engine to execute is row-by-row processing. So no way that the SQL engine can do that in parallel as your code is a single session expecting a single row at a time.
Performance is therefore not only faster using native SQL (as the others indicated with their responses), native SQL is also a lot more scalable than doing row-by-row processing using PL/SQL.
There is a simple mantra for this: Maximise SQL. Minimise PL/SQL.
SQL is the superior language to crunch data in the database. Not PL/SQL.