Using Bulk Binds

Oracle Database Tips by Donald Burleson

In Pre 8.1, execution of every SQL statement embedded in the PL/SQL programs involves a context switch between the PL/SQL engine and SQL engine. This results in performance penalty. If we could avoid this switch or maximize processing during the switch, efficiencies are gained. By using BULK BIND we can maximize the performance of time spent in the SQL context.

The new Bulk Binds feature in PL/SQL 8.1 allows us to operate on multiple rows in a single DML statement.

Typical Performance Improvement achieved by WITH BULK BINDS feature vs WITHOUT BULK BINDS are listed below :

Note : The following figures are representative of the gains that might be had by passing parameters WITH BULK BINDS mechanism verses WITHOUT BULK BINDS mechanism. Actual timings can vary depending on the machine type and load etc.

Task	WITHOUT BULK BINDS	WITH BULK BINDS
Time Units	( Secs )	( Secs )
Inserting 10000 Records	77	8
Querying 10000 Records	19	12

Technique:

The example below use the following type definitions for the nested tables into which the bulk binds are performed:

/*--------------------------------------------------------------------*/
/*  Nested Table type for storing opening,closing and high values     */
/*--------------------------------------------------------------------*/

CREATE TYPE nt_values AS TABLE OF NUMBER(8,2)
/
 

/*--------------------------------------------------------------------*/
/*  Nested Table type for storing volume                              */
/*--------------------------------------------------------------------*/ 

CREATE TYPE nt_volume AS TABLE OF NUMBER(12)

/

/*--------------------------------------------------------------------*/
/*  Nested Table type for storing stock names                         */
/*--------------------------------------------------------------------*/

CREATE TYPE nt_stknames AS TABLE OF VARCHAR2(30)
/

Typical parameter definitions would then become:

stock_name nt_stknames,
stk_open_vals nt_values,
high nt_values,
stk_close_vals nt_values,
vol nt_volume,

To use the bulk binds feature in Oracle8i you will convert code of the form:

FOR l_i IN 1..stk_open_vals.COUNT
       LOOP
 INSERT INTO Stock_History VALUES (stock_name(l_i),stk_open_vals(l_i),
      high(l_i), stk_close_vals(l_i), vol(l_i),NULL, SYSDATE - 1);
       END LOOP;

Into code using the bulk binds feature such as:

      FORALL l_i IN 1..stk_open_vals.COUNT
 INSERT INTO Stock_History VALUES (stock_name(l_i),stk_open_vals(l_i),
     high(l_i), stk_close_vals(l_i), vol(l_i),NULL, SYSDATE);

Both code fragments are essentially identical except the non-bulk bind example uses a standard FOR loop construct while the bulk bind example replaces this with the FORALL loop.

Similar code techniques can be used for SELECT, UPDATE and DELETE as well, for example for SELECT not using BULK bind:

        CURSOR l_c1 IS
    SELECT stock_name,AVG(VOLUME) FROM Stock_History
    GROUP BY stock_name
     ORDER BY stock_name;
    BEGIN
       OPEN l_c1;
       LOOP
   l_averages.EXTEND;
   l_stock_names.EXTEND;
           FETCH l_c1 INTO l_stock_names(l_i),l_averages(l_i);
   EXIT WHEN l_c1%NOTFOUND;
   l_i := l_i + 1;
       END LOOP;
       CLOSE l_c1;
    END;

For SELECT using BULK bind:

      CURSOR l_c1 IS
    SELECT stock_name,AVG(VOLUME) FROM Stock_History
    GROUP BY stock_name
    ORDER BY stock_name;
    BEGIN
       /* Use BULK FETCHING  */
       OPEN l_c1;
       FETCH l_c1 BULK COLLECT INTO l_stock_names,l_averages;
       CLOSE l_c1;
   END;

Notice in the SELECT example the bulk bind is implemented through the FETCH?BULK COLLECT INTO construct.

In the next example the use of FORALL and an UPDATE statement is demonstrated:

BEGIN
       /* Bulk Update Statement */
        FORALL l_j IN 1..100
           UPDATE stock_history
           SET stock_group = 'L'
           WHERE volume <= 500 AND stock_name = all_stock_names(l_j) AND
           to_char(Trade_Date,'DD-MON-YYYY') = to_char(SYSDATE,'DD-MON-YYYY');