PL/SQL usage in Datawarehousing projects

Question

Hi,
In Datawarehousing projects, most of the processing can be done through SQL code directly.
I found from few articles that we should strongly avoid PL/SQL Scripts in DWH projects.
But, if situation warrants, can we use PL/SQL for processing in D/W projects, is it advisable?
Please suggest, thank you.
Regards,
DR

Answer 1

PL/SQL is used instead of single SQL statements due to the "BULK COLLECT/FORALL" functionality which reduces the size of redo operations.It uses more redo, not less.
Tested in 11.1.0.6:
SQL> @desc one_million_rows
Name                                      Null?    Type
ID                                                 NUMBER
VAL                                                VARCHAR2(30)
SQL> set autotrace on statistics
SQL> insert into test_redo select * from one_million_rows;
1000000 rows created.
Statistics
          0  recursive calls
      42508  db block gets
      16959  consistent gets
       5827  physical reads
   48187952  redo size
        920  bytes sent via SQL*Net to client
       1017  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          1  sorts (memory)
          0  sorts (disk)
    1000000  rows processed
SQL> roll
Rollback complete.
SQL> set autotrace off
SQL> SELECT s.value FROM v$mystat s JOIN v$statname n USING (statistic#)
  2  WHERE  n.name = 'redo size';
     VALUE
317239240
1 row selected.
SQL> DECLARE
  2      CURSOR c IS SELECT * FROM one_million_rows;
  3      TYPE t IS TABLE OF one_million_rows%ROWTYPE;
  4      v_rows t;
  5  BEGIN
  6      OPEN c;
  7      LOOP
  8          FETCH c BULK COLLECT INTO v_rows LIMIT 500;
  9
10          FORALL i IN v_rows.FIRST..v_rows.LAST
11              INSERT INTO test_redo VALUES v_rows(i);
12
13          EXIT WHEN c%NOTFOUND;
14      END LOOP;
15  END;
16  /
PL/SQL procedure successfully completed.
SQL> SELECT s.value FROM v$mystat s JOIN v$statname n USING (statistic#)
  2  WHERE  n.name = 'redo size';
     VALUE
365731476
1 row selected.
SQL> select 365731476 - 317239240 redo from dual;
      REDO
  48492236
1 row selected.The PL/SQL block used 304,284 bytes more redo, which is not much but still more than the SQL version.