6 Embedded SQL

Output versus Input Host Variables

Depending on how they are used, host variables are called output or input host variables.

Host variables in the INTO clause of a SELECT or FETCH statement are called output host variables because they hold column values output by Oracle. Oracle assigns the column values to corresponding output host variables in the INTO clause.

All other host variables in a SQL statement are called input host variables because your program inputs their values to Oracle. For example, you use input host variables in the VALUES clause of an INSERT statement and in the SET clause of an UPDATE statement. They are also used in the WHERE, HAVING, and FOR clauses. Input host variables can appear in a SQL statement wherever a value or expression is allowed.

EXEC SQL SELECT ename, empno INTO :name,:number FROM emp ORDER BY :ord;

You cannot use input host variables to supply SQL keywords or the names of database objects. Thus, you cannot use input host variables in data definition statements such as ALTER, CREATE, and DROP. In the following example, the DROP TABLE statement is invalid:

char table_name[30]; 
 
printf("Table name? "); 
gets(table_name); 
 
EXEC SQL DROP TABLE :table_name;  -- host variable not allowed 

If you need to change database object names at runtime, use dynamic SQL. See also Chapter 13, "Oracle Dynamic SQL".

Before Oracle executes a SQL statement containing input host variables, your program must assign values to them. An example follows:

int     emp_number; 
char    temp[20];
VARCHAR emp_name[20]; 

/* get values for input host variables */ 
printf("Employee number? "); 
gets(temp);
emp_number = atoi(temp);
printf("Employee name? "); 
gets(emp_name.arr); 
emp_name.len = strlen(emp_name.arr); 
 
EXEC SQL INSERT INTO EMP (EMPNO, ENAME) 
    VALUES (:emp_number, :emp_name); 

Notice that the input host variables in the VALUES clause of the INSERT statement are prefixed with colons.

Insert NULLs

You can use indicator variables to INSERT NULLs. Before the INSERT, for each column you want to be NULL, set the appropriate indicator variable to -1, as shown in the following example:

set ind_comm = -1; 
 
EXEC SQL INSERT INTO emp (empno, comm) 
     VALUES (:emp_number, :commission:ind_comm); 

The indicator variable ind_comm specifies that a NULL is to be stored in the COMM column.

You can hard code the NULL instead, as follows:

EXEC SQL INSERT INTO emp (empno, comm) 
     VALUES (:emp_number, NULL); 

While this is less flexible, it might be more readable. Typically, you insert NULLs conditionally, as the next example shows:

printf("Enter employee number or 0 if not available: "); 
scanf("%d", &emp_number); 
 
if (emp_number == 0) 
    ind_empnum = -1; 
else 
    ind_empnum = 0; 
 
EXEC SQL INSERT INTO emp (empno, sal) 
     VALUES (:emp_number:ind_empnum, :salary); 

Returned NULLs

You can also use indicator variables to manipulate returned NULLs, as the following example shows:

EXEC SQL SELECT ename, sal, comm 
    INTO :emp_name, :salary, :commission:ind_comm 
    FROM emp 
    WHERE empno = :emp_number; 
 if (ind_comm == -1) 
    pay = salary;   /* commission is NULL; ignore it */ 
else 
    pay = salary + commission; 

Fetch NULLs

When DBMS=V7 or DBMS=V8, if you SELECT or FETCH NULLs into a host variable not associated with an indicator variable, Oracle issues the following error message:

ORA-01405: fetched column value is NULL 

Test for NULLs

You can use indicator variables in the WHERE clause to test for NULLs, as the following example shows:

EXEC SQL SELECT ename, sal 
INTO :emp_name, :salary 
FROM emp 
WHERE :commission INDICATOR :ind_comm IS NULL ... 

However, you cannot use a relational operator to compare NULLs with each other or with other values. For example, the following SELECT statement fails if the COMM column contains one or more NULLs:

EXEC SQL SELECT ename, sal 
INTO :emp_name, :salary 
FROM emp 
WHERE comm = :commission; 

The next example shows how to compare values for equality when some of them might be NULLs:

EXEC SQL SELECT ename, sal 
     INTO :emp_name, :salary 
     FROM emp 
     WHERE (comm = :commission) OR ((comm IS NULL) AND 
          (:commission INDICATOR :ind_comm IS NULL)); 

Truncated Values

When DBMS=V7 or V8, if you SELECT or FETCH a truncated column value into a host variable not associated with an indicator variable, a warning is generated instead of an error.

The SELECT Statement

Querying the database is a common SQL operation. To issue a query you use the SELECT statement. In the following example, you query the EMP table:

EXEC SQL SELECT ename, job, sal + 2000 
INTO :emp_name, :job_title, :salary 
FROM emp 
WHERE empno = :emp_number; 

The column names and expressions following the keyword SELECT make up the select list. The select list in our example contains three items. Under the conditions specified in the WHERE clause (and following clauses, if present), Oracle returns column values to the host variables in the INTO clause.

The number of items in the select list should equal the number of host variables in the INTO clause, so there is a place to store every returned value.

In the simplest case, when a query returns one row, its form is that shown in the last example. However, if a query can return more than one row, you must FETCH the rows using a cursor or SELECT them into a host-variable array. Cursors and the FETCH statement are discussed later in this chapter. See also "Host Arrays"for information on array processing.

If a query is written to return only one row but might actually return several rows, the result of the SELECT is indeterminate. Whether this causes an error depends on how you specify the SELECT_ERROR option. The default value, YES, generates an error if more than one row is returned.

The INSERT Statement

Use the INSERT statement to add rows to a table or view. In the following example, you add a row to the EMP table:

EXEC SQL INSERT INTO emp (empno, ename, sal, deptno) 
VALUES (:emp_number, :emp_name, :salary, :dept_number); 

Each column you specify in the column list must belong to the table named in the INTO clause. The VALUES clause specifies the row of values to be inserted. The values can be those of constants, host variables, SQL expressions, SQL functions such as USER and SYSDATE, or user-defined PL/SQL functions.

The number of values in the VALUES clause must equal the number of names in the column list. However, you can omit the column list if the VALUES clause contains a value for each column in the table, in the order that they are defined in the table.

EXEC SQL INSERT INTO emp2 (empno, ename, sal, deptno) 
SELECT empno, ename, sal, deptno FROM emp  
WHERE job= :job_title ;

The UPDATE Statement

Use the UPDATE statement to change the values of specified columns in a table or view. In the following example, we update the SAL and COMM columns in the EMP table:

EXEC SQL UPDATE emp 
SET sal = :salary, comm = :commission 
WHERE empno = :emp_number;

Use the optional WHERE clause to specify the conditions under which rows are updated. See also "The WHERE Clause".

The SET clause lists the names of one or more columns for which you must provide values. You can use a subquery to provide the values, as the following example shows:

EXEC SQL UPDATE emp 
SET sal = (SELECT AVG(sal)*1.1 FROM emp WHERE deptno = 20) 
WHERE empno = :emp_number; 

The UPDATE statement has an optional returning clause, like the INSERT and DELETE statements. It is only allowed after the optional WHERE condition.

For more details, see also "UPDATE (Executable Embedded SQL)".

The DELETE Statement

Use the DELETE statement to remove rows from a table or view. In the following example, you delete all employees in a given department from the EMP table:

EXEC SQL DELETE FROM emp 
WHERE deptno = :dept_number ; 

We have used the optional WHERE clause to specify the condition under which rows are deleted.

The returning clause option can be used in DELETE statements also. It is allowed after the optional WHERE condition. In the earlier example, it is good practice to record the field values of each employee that is deleted.

The WHERE Clause

Use the WHERE clause to SELECT, UPDATE, or DELETE only those rows in a table or view that meet your search condition. The WHERE-clause search condition is a Boolean expression, which can include scalar host variables, host arrays (not in SELECT statements), subqueries, and user-defined stored functions.

If you omit the WHERE clause, all rows in the table or view are processed. If you omit the WHERE clause in an UPDATE or DELETE statement, Oracle sets sqlwarn[4] in the SQLCA to 'W' to warn that all rows were processed.

The DECLARE CURSOR Statement

You use the DECLARE CURSOR statement to define a cursor by giving it a name and associating it with a query, as the following example shows:

EXEC SQL DECLARE emp_cursor CURSOR FOR 
     SELECT ename, empno, sal 
     FROM emp 
     WHERE deptno = :dept_number; 

The cursor name is an identifier used by the precompiler, not a host or program variable, and should not be defined in the Declare Section. Therefore, cursor names cannot be passed from one precompilation unit to another. Cursor names cannot be hyphenated. They can be any length, but only the first 31 characters are significant. For ANSI compatibility, use cursor names no longer than 18 characters.

The precompiler option CLOSE_ON_COMMIT is provided for use in the command line or in a configuration file. Any cursor not declared with the WITH HOLD clause is closed after a COMMIT or ROLLBACK when CLOSE_ON_COMMIT=YES. See "WITH HOLD Clause in DECLARE CURSOR Statements", and "CLOSE_ON_COMMIT".

If MODE is specified at a higher level than CLOSE_ON_COMMIT, then MODE takes precedence. The defaults are MODE=ORACLE and CLOSE_ON_COMMIT=NO. If you specify MODE=ANSI then any cursors not using the WITH HOLD clause will be closed on COMMIT. The application will run more slowly because cursors are closed and re-opened many times. Setting CLOSE_ON_COMMIT=NO when MODE=ANSI results in performance improvement. To see how macro options such as MODE affect micro options such as CLOSE_ON_COMMIT, see "Precedence of Option Values".

The SELECT statement associated with the cursor cannot include an INTO clause. Rather, the INTO clause and list of output host variables are part of the FETCH statement.

Because it is declarative, the DECLARE CURSOR statement must physically (not just logically) precede all other SQL statements referencing the cursor. That is, forward references to the cursor are not allowed. In the following example, the OPEN statement is misplaced:

... 
EXEC SQL OPEN emp_cursor; 
*    -- MISPLACED OPEN STATEMENT
EXEC SQL DECLARE emp_cursor CURSOR FOR 
SELECT ename, empno, sal 
    FROM emp 
    WHERE ename = :emp_name; 

The cursor control statements (DECLARE, OPEN, FETCH, CLOSE) must all occur within the same precompiled unit. For example, you cannot DECLARE a cursor in file A, then OPEN it in file B.

Your host program can DECLARE as many cursors as it needs. However, in a given file, every DECLARE statement must be unique. That is, you cannot DECLARE two cursors with the same name in one precompilation unit, even across blocks or procedures, because the scope of a cursor is global within a file.

For users of MODE=ANSI or CLOSE_ON_COMMIT=YES, the WITH HOLD clause can be used in a DECLARE section to override the behavior defined by the two options. With these options set, the behavior will be for all cursors to be closed when a COMMIT is issued. This can have performance implications due to the overhead of re-opening the cursor to continue processing. The careful use of WITH HOLD can speed up programs that need to conform to the ANSI standard for precompilers in most respects.

If you will be using many cursors, you might want to specify the MAXOPENCURSORS option.

The OPEN Statement

You use the OPEN statement to execute the query and identify the active set. In the following example, you OPEN a cursor named emp_cursor:

EXEC SQL OPEN emp_cursor;
 

OPEN zeroes the rows-processed count kept by the third element of SQLERRD in the SQLCA. However, none of the rows are visible to the application at this point. That is handled by the FETCH statement.

OPEN positions the cursor just before the first row of the active set. It also zeroes the rows-processed count kept by the third element of SQLERRD in the SQLCA. However, none of the rows is actually retrieved at this point. That will be done by the FETCH statement.

Once you OPEN a cursor, the query's input host variables are not re-examined until you reOPEN the cursor. Thus, the active set does not change. To change the active set, you must reOPEN the cursor.

Generally, you should CLOSE a cursor before reOPENing it. However, if you specify MODE=ORACLE (the default), you need not CLOSE a cursor before reOPENing it. This can increase performance.

The amount of work done by OPEN depends on the values of three precompiler options: HOLD_CURSOR, RELEASE_CURSOR, and MAXOPENCURSORS.

The FETCH Statement

You use the FETCH statement to retrieve rows from the active set and specify the output host variables that will contain the results. Recall that the SELECT statement associated with the cursor cannot include an INTO clause. Rather, the INTO clause and list of output host variables are part of the FETCH statement. In the following example, you FETCH INTO three host variables:

EXEC SQL FETCH emp_cursor 
INTO :emp_name, :emp_number, :salary;
 

The cursor must have been previously DECLAREd and OPENed. The first time you execute FETCH, the cursor moves from before the first row in the active set to the first row. This row becomes the current row. Each subsequent execution of FETCH advances the cursor to the next row in the active set, changing the current row. The cursor can only move forward in the active set. To return to a row that has already been FETCHed, you must reOPEN the cursor, then begin again at the first row of the active set.

If you want to change the active set, you must assign new values to the input host variables in the query associated with the cursor, then reOPEN the cursor. When MODE=ANSI, you must CLOSE the cursor before reOPENing it.

As the next example shows, you can FETCH from the same cursor using different sets of output host variables. However, corresponding host variables in the INTO clause of each FETCH statement must have the same datatype.

EXEC SQL DECLARE emp_cursor CURSOR FOR 
SELECT ename, sal FROM emp WHERE deptno = 20; 
... 
EXEC SQL OPEN emp_cursor; 
 
EXEC SQL WHENEVER NOT FOUND GOTO ... 
for (;;) 
{ 
    EXEC SQL FETCH emp_cursor INTO :emp_name1, :salary1; 
    EXEC SQL FETCH emp_cursor INTO :emp_name2, :salary2; 
    EXEC SQL FETCH emp_cursor INTO :emp_name3, :salary3; 
    ... 
} 

If the active set is empty or contains no more rows, FETCH returns the "no data found" error code to sqlcode in the SQLCA, or to the SQLCODE or SQLSTATE status variables. The status of the output host variables is indeterminate. (In a typical program, the WHENEVER NOT FOUND statement detects this error.) To reuse the cursor, you must reOPEN it.

It is an error to FETCH on a cursor under the following conditions:

• Before OPENing the cursor

• After a "no data found" condition

• After CLOSEing it

The CLOSE Statement

When done FETCHing rows from the active set, you CLOSE the cursor to free the resources, such as storage, acquired by OPENing the cursor. When a cursor is closed, parse locks are released. What resources are freed depends on how you specify the HOLD_CURSOR and RELEASE_CURSOR options. In the following example, you CLOSE the cursor named emp_cursor:

EXEC SQL CLOSE emp_cursor; 

You cannot FETCH from a closed cursor because its active set becomes undefined. If necessary, you can reOPEN a cursor (with new values for the input host variables, for example).

When MODE=ORACLE, issuing a COMMIT or ROLLBACK closes cursors referenced in a CURRENT OF clause. Other cursors are unaffected by COMMIT or ROLLBACK and if open, remain open. However, when MODE=ANSI, issuing a COMMIT or ROLLBACK closes all explicit cursors.

Using Scrollable Cursors

The following statements let you define and manipulate a scrollable cursor.

EXEC SQL DECLARE emp_cursor SCROLL CURSOR FOR
SELECT ename, sal FROM emp WHERE deptno=20;
...
EXEC SQL OPEN emp_cursor;
EXEC SQL FETCH LAST emp_cursor INTO :emp_name, :sal;
EXEC SQL CLOSE emp_cursor;

The CLOSE_ON_COMMIT Precompiler Option

The CLOSE_ON_COMMIT micro precompiler option provides the ability to choose whether or not to close all cursors when a COMMIT is executed and the macro option MODE=ANSI. When MODE=ANSI, CLOSE_ON_COMMIT has the default value YES. Explicitly setting CLOSE_ON_COMMIT=NO results in better performance because cursors will not be closed when a COMMIT is executed, removing the need to re-open the cursors and incur extra parsing.

The PREFETCH Precompiler Option

The precompiler option PREFETCH allows for more efficient queries by pre-fetching a given number of rows. This decreases the number of server round trips needed and reduces overall memory usage. The number of rows set by the PREFETCH option value is used for all queries involving explicit cursors, subject to the standard precedence rules. When used inline, the PREFETCH option must precede any of these cursor statements:

• EXEC SQL OPEN cursor

• EXEC SQL OPEN cursor USING host_var_list

• EXEC SQL OPEN cursor USING DESCRIPTOR desc_name

When an OPEN is executed, the value of PREFETCH gives the number of rows to be pre-fetched when the query is executed. You can set the value from 0 (no pre-fetching) to 65535. The default value is 1.

Issuing Hints

You can issue an optimizer hint inside a C or C++ style comment, immediately after a SELECT, DELETE, or UPDATE command. You indicate that the comment contains one or more hints by following the comment opener with a plus sign, leaving no space between the opener and the '+'. For example, the following statement uses the ALL_ROWS hint to let the cost-based approach optimize the statement for the goal of best throughput:

EXEC SQL SELECT /*+ ALL_ROWS (cost-based) */ empno, ename, sal, job 
    INTO :emp_rec FROM emp 
    WHERE deptno = :dept_number; 

As shown in this statement, the comment can contain optimizer hints as well as other comments.

For more information about the cost-based optimizer, and optimizer hints, see Oracle Database Advanced Application Developer's Guide.

SQL File

The generated file name has the following format:

<filename>_<filetype>.sql

In Pro*C, for the file "abc.pc", the generated SQL file will be abc_pc.sql.

Generated file format

If the outlnprefix option is not used, then the format of the unique identifier used as outline name and comment is:

<category_name>_<filename>_<filetype>_<sequence no.>

If the outlnprefix option is used (outlnprefix=<prefix_name>), then the format of the unique identifier used as outline name and comment is:

<prefix_name>_<sequence no.>

If outline=yes, which is the default category, then <category_name> will be DEFAULT and outline name will be:

DEFAULT_<filename>_<filetype>_<sequence no.>

or

<prefix_name>_<sequence no.>

The allowed range for <sequence no.> is 0000 to 9999.

SQL in the generated precompiled file will have the comment appended to it as it appears in the outline for that SQL.

EXEC SQL select * from emp where empno=:var;
EXEC SQL select * from dept;

sqlstm.stmt = select * from emp where empno=:b1 /* mycat1_abc_pc_0000 */;

sqlstm.stmt = select * from dept /* mycat1_abc_pc_0001 */;

EXEC SQL select * from emp where empno=:var;
EXEC SQL select * from dept;

sqlstm.stmt = select * from emp where empno=:b1 /* myprefix_0000 */;

sqlstm.stmt = select * from dept /* myprefix_0001 */;

EXEC SQL select * from emp where empno=:var;
EXEC SQL select * from dept;

sqlstm.stmt = "select * from emp where empno=:b1 /* myprefix_0000 */;

sqlstm.stmt = "select * from dept /* myprefix_0001 */";

LOG File

The generated file name has the following format:

<filename>_<filetype>.log

In Pro*C, for the file "abc.pc", the generated LOG file will be abc_pc.log.

Consider the following example.

Example 1

If abc.pc has the statements

EXEC SQL select * from emp;

Contents of abc_pc.log

CATEGORY <Category_name> 
     Source SQL_0
        SELECT * FROM emp
     OUTLINE NAME
        abc_pc_0000
     OUTLINE SQL_0
        Select * from emp /* abc_pc_0000 */

Restrictions

You cannot use CURRENT OF clause on an index-organized table.

Explicit FOR UPDATE OF clauses or implicit FOR UPDATE clauses acquire exclusive row locks. All rows are locked at the OPEN, not as they are FETCHed, and are released when you COMMIT or ROLLBACK. Therefore, you cannot FETCH from a FOR UPDATE cursor after a COMMIT. If you try to do this, Oracle returns a 1002 error code.

Also, you cannot use host arrays with the CURRENT OF clause. For an alternative, see also "Mimicking CURRENT OF".

Furthermore, you cannot reference multiple tables in an associated FOR UPDATE OF clause, which means that you cannot do joins with the CURRENT OF clause.

Finally, you cannot use dynamic SQL with the CURRENT OF clause.