5 Binding and Defining in OCI

This chapter contains these topics:

Overview of Binding in OCI
Advanced Bind Operations in OCI
Overview of Defining in OCI
Advanced Define Operations in OCI
Binding and Defining Arrays of Structures in OCI
Binding and Defining Multiple Buffers
DML with RETURNING Clause in OCI
Character Conversion in OCI Binding and Defining
PL/SQL REF CURSORs and Nested Tables in OCI
Runtime Data Allocation and Piecewise Operations in OCI

Overview of Binding in OCI

This chapter expands on the basic concepts of binding and defining, and provides more detailed information about the different types of binds and defines you can use in OCI applications. Additionally, this chapter discusses the use of arrays of structures, as well as other issues involved in binding, defining, and character conversions.

For example, given the INSERT statement

INSERT INTO emp VALUES
    (:empno, :ename, :job, :sal, :deptno)

and the following variable declarations

text     *ename, *job;
sword    empno, sal, deptno;

the bind step makes an association between the placeholder name and the address of the program variables. The bind also indicates the datatype and length of the program variables, as illustrated in Figure 5-1.

See Also:

The code that implements this example is found in the section "Steps Used in OCI Binding".

Figure 5-1 Using OCIBindByName() to Associate Placeholders with Program Variables

Description of "Figure 5-1 Using OCIBindByName() to Associate Placeholders with Program Variables"

If you change only the value of a bind variable, it is not necessary to rebind it in order to execute the statement again. Because the bind is by reference, as long as the address of the variable and handle remain valid, you can re-execute a statement that references the variable without rebinding.

Note:

At the interface level, all bind variables are considered at least IN and must be properly initialized. If the variable is a pure OUT bind variable, you can set the variable to 0. You can also provide a NULL indicator and set that indicator to -1 (NULL).

In the Oracle server, datatypes have been implemented for named datatypes, REFs and LOBs, and they may be bound as placeholders in a SQL statement.

Note:

For opaque datatypes (descriptors or locators) whose sizes are not known, pass the address of the descriptor or locator pointer. Set the size parameter to the size of the appropriate data structure, (sizeof(structure))

Named Binds and Positional Binds

The SQL statement in the previous section is an example of a named bind. Each placeholder in the statement has a name associated with it, such as 'ename' or 'sal'. When this statement is prepared and the placeholders are associated with values in the application, the association is made by the name of the placeholder using the OCIBindByName() call with the name of the placeholder passed in the placeholder parameter.

A second type of bind is known as a positional bind. In a positional bind, the placeholders are referred to by their position in the statement rather than their names. For binding purposes, an association is made between an input value and the position of the placeholder, using the OCIBindByPos() call.

Using the previous example for a positional bind:

INSERT INTO emp VALUES
    (:empno, :ename, :job, :sal, :deptno)

The five placeholders are then each bound by calling OCIBindByPos() and passing the position number of the placeholder in the position parameter. For example, the :empno placeholder would be bound by calling OCIBindByPos() with a position of 1, :ename with a position of 2, and so on.

In the case of a duplicate bind, only a single bind call may be necessary. Consider the following SQL statement, which queries the database for employees whose commission and salary are both greater than a given amount:

SELECT empno FROM emp
    WHERE sal > :some_value
    AND comm > :some_value

An OCI application could complete the binds for this statement with a single call to OCIBindByName() to bind the :some_value placeholder by name. In this case, the second placeholder inherits the bind information from the first placeholder.

OCI Array Interface

You can pass data to Oracle in various ways.

You can execute a SQL statement repeatedly using the OCIStmtExecute() routine and supply different input values on each iteration.

You can use the Oracle array interface and input many values with a single statement and a single call to OCIStmtExecute(). In this case you bind an array to an input placeholder, and the entire array can be passed at the same time, under the control of the iters parameter.

The array interface significantly reduces round trips to the database when you are updating or inserting a large volume of data. This reduction can lead to considerable performance gains in a busy client/server environment. For example, consider an application that needs to insert 10 rows into the database. Calling OCIStmtExecute() ten times with single values results in ten network round trips to insert all the data. The same result is possible with a single call to OCIStmtExecute() using an input array, which involves only one network round trip.

Note:

When using the OCI array interface to perform inserts, row triggers in the database are fired as each row is inserted.

The maximum number of rows allowed in an array DML statement is 4 gigabytes -1.

Binding Placeholders in PL/SQL

You process a PL/SQL block by placing the block in a string variable, binding any variables, and then executing the statement containing the block, just as you would with a single SQL statement.

When you bind placeholders in a PL/SQL block to program variables, you must use OCIBindByName() or OCIBindByPos() to perform the basic binds for host variables that are either scalars or arrays.

The following short PL/SQL block contains two placeholders, which represent IN parameters to a procedure that updates an employee's salary, given the employee number and the new salary amount:

char plsql_statement[] = "BEGIN\
                          RAISE_SALARY(:emp_number, :new_sal);\
                          END;" ;

These placeholders can be bound to input variables in the same way as placeholders in a SQL statement.

When processing PL/SQL statements, output variables are also associated with program variables using bind calls.

For example, in a PL/SQL block such as

BEGIN
    SELECT ename,sal,comm INTO :emp_name, :salary, :commission
    FROM emp
    WHERE empno = :emp_number;
END;

you would use OCIBindByName() to bind variables in place of the :emp_name, :salary, and :commission output placeholders, and in place of the input placeholder :emp_number.

Note:

All buffers, even pure OUT buffers, must be initialized by setting the buffer length to zero in the bind call, or by setting the corresponding indicator to -1.

See Also:

"Information for Named Datatype and REF Binds" for more information about binding PL/SQL placeholders

Steps Used in OCI Binding

Placeholders are bound in several steps. For a simple scalar or array bind, it is only necessary to specify an association between the placeholder and the data, by using OCIBindByName() or OCIBindByPos().

Once the bind is complete, the OCI library knows where to find the input data or where to put PL/SQL output data when the SQL statement is executed. Program input data does not need to be in the program variable when it is bound to the placeholder, but the data must be there when the statement is executed.

The following code example shows handle allocation and binding for each placeholder in a SQL statement.

...
/* The SQL statement, associated with stmthp (the statement handle)
by calling OCIStmtPrepare() */
text *insert = (text *) "INSERT INTO emp(empno, ename, job, sal, deptno)\
    VALUES (:empno, :ename, :job, :sal, :deptno)";
...

/*  Bind the placeholders in the SQL statement, one per bind handle. */
checkerr(errhp, OCIBindByName(stmthp, &bnd1p, errhp, (text *) ":ENAME",
    strlen(":ENAME"), (ub1 *) ename, enamelen+1, SQLT_STR, (void *) 0,
    (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));
checkerr(errhp, OCIBindByName(stmthp, &bnd2p, errhp, (text *) ":JOB",
    strlen(":JOB"), (ub1 *) job, joblen+1, SQLT_STR, (void *)
    &job_ind, (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));
checkerr(errhp, OCIBindByName(stmthp, &bnd3p, errhp, (text *) ":SAL",
    strlen(":SAL"), (ub1 *) &sal, (sword) sizeof(sal), SQLT_INT,
    (void *) &sal_ind, (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0,
    OCI_DEFAULT));
checkerr(errhp, OCIBindByName(stmthp, &bnd4p, errhp, (text *) ":DEPTNO",
    strlen(":DEPTNO"), (ub1 *) &deptno,(sword) sizeof(deptno), SQLT_INT,
    (void *) 0, (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));
checkerr(errhp, OCIBindByName(stmthp, &bnd5p, errhp, (text *) ":EMPNO",
    strlen(":EMPNO"), (ub1 *) &empno, (sword) sizeof(empno), SQLT_INT,
    (void *) 0, (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0,OCI_DEFAULT));

Note:

The checkerr() function evaluates the return code from an OCI application. The code for the function is listed in the section "OCI Programming Steps".

PL/SQL Block in an OCI Program

Perhaps the most common use for PL/SQL blocks in OCI is to call stored procedures or stored functions. Assume that there is a procedure named RAISE_SALARY stored in the database, and you embed a call to that procedure in an anonymous PL/SQL block, and then process the PL/SQL block.

The following program fragment shows how to embed a stored procedure call in an OCI application. The program passes an employee number and a salary increase as inputs to a stored procedure called raise_salary:

raise_salary (employee_num IN, sal_increase IN, new_salary OUT);

This procedure raises a given employee's salary by a given amount. The increased salary which results is returned in the stored procedure's variable, new_salary, and the program displays this value.

Note that the PL/SQL procedure argument, new_salary, although a PL/SQL OUT variable, must be bound, not defined. This is further explained in the section on OCI defines.

/* Define PL/SQL statement to be used in program. */
text *give_raise = (text *) "BEGIN\
                  RAISE_SALARY(:emp_number,:sal_increase, :new_salary);\
                     END;";
OCIBind  *bnd1p = NULL;                      /* the first bind handle */
OCIBind  *bnd2p = NULL;                     /* the second bind handle */
OCIBind  *bnd3p = NULL;                      /* the third bind handle */

static void checkerr();
sb4 status;

main()
{
  sword    empno, raise, new_sal;
  void     *tmp;
  OCISession *usrhp = (OCISession *)NULL; 
...
/* attach to database server, and perform necessary initializations
and authorizations */
...
      /* allocate a statement handle */
  checkerr(errhp, OCIHandleAlloc( (void *) envhp, (void **) &stmthp,
           OCI_HTYPE_STMT, 100, (void **) &tmp));

      /* prepare the statement request, passing the PL/SQL text
        block as the statement to be prepared */
checkerr(errhp, OCIStmtPrepare(stmthp, errhp, (text *) give_raise, (ub4) 
      strlen(give_raise), OCI_NTV_SYNTAX, OCI_DEFAULT));

      /* bind each of the placeholders to a program variable */
 checkerr( errhp, OCIBindByName(stmthp, &bnd1p, errhp, (text *) ":emp_number",
             -1, (ub1 *) &empno,
            (sword) sizeof(empno), SQLT_INT, (void *) 0,
             (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));

 checkerr( errhp, OCIBindByName(stmthp, &bnd2p, errhp, (text *) ":sal_increase",
             -1, (ub1 *) &raise,
             (sword) sizeof(raise), SQLT_INT, (void *) 0,
             (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));

      /* remember that PL/SQL OUT variable are bound, not defined */

checkerr( errhp, OCIBindByName(stmthp, &bnd3p, errhp, (text *) ":new_salary",
             -1, (ub1 *) &new_sal,
             (sword) sizeof(new_sal), SQLT_INT, (void *) 0,
             (ub2 *) 0, (ub2) 0, (ub4) 0, (ub4 *) 0, OCI_DEFAULT));

      /* prompt the user for input values */
printf("Enter the employee number: ");
scanf("%d", &empno); 
      /* flush the input buffer */
myfflush();

printf("Enter employee's raise: ");
scanf("%d", &raise);
      /* flush the input buffer */
myfflush();

  /* execute PL/SQL block*/
  checkerr(errhp, OCIStmtExecute(svchp, stmthp, errhp, (ub4) 1, (ub4) 0, 
      (OCISnapshot *) NULL, (OCISnapshot *) NULL, OCI_DEFAULT));

  /* display the new salary, following the raise */
printf("The new salary is %d\n", new_sal);
}

This example demonstrates how to perform a simple scalar bind where only a single bind call is necessary. In some cases, additional bind calls are needed to define attributes for specific bind datatypes or execution modes.

Advanced Bind Operations in OCI

The section "Binding Placeholders in OCI" discussed how a basic bind operation is performed to create an association between a placeholder in a SQL statement and a program variable using OCIBindByName() or OCIBindByPos(). This section covers more advanced bind operations, including multi-step binds, and binds of named datatypes and REFs.

In some cases, additional bind calls are necessary to define specific attributes for certain bind datatypes or certain execution modes.

The following sections describe these special cases, and the information about binding is summarized in Table 5-1.

Table 5-1 Information Summary for Bind Types

Type of Bind	Bind Datatype	Notes
Scalar	any scalar datatype	Bind a single scalar using `OCIBindByName()` or `OCIBindByPos()`.
Array of Scalars	any scalar datatype	Bind an array of scalars using `OCIBindByName()` or `OCIBindByPos()`.
Named Datatype	SQLT_NTY	Two bind calls are required: `OCIBindByName()` or `OCIBindByPos()` `OCIBindObject()`
REF	SQLT_REF	Two bind calls are required: `OCIBindByName()` or `OCIBindByPos()` `OCIBindObject()`
LOB `BFILE`	SQLT_BLOB SQLT_CLOB	Allocate the LOB locator using `OCIDescriptorAlloc()`, and then bind its address, `OCILobLocator` `**`, with `OCIBindByName()` or `OCIBindByPos()`, using one of the LOB datatypes.
Array of Structures or Static Arrays	varies	Two bind calls are required: `OCIBindByName()` or `OCIBindByPos()` `OCIBindArrayOfStruct()`
Piecewise Insert	varies	`OCIBindByName()` or `OCIBindByPos()` is required. The application may also need to call `OCIBindDynamic()` to register piecewise callbacks.
REF CURSOR variables	SQLT_RSET	Allocate a statement handle, `OCIStmt`, and then bind its address, `OCIStmt **`, using the SQLT_RSET datatype.

See Also:

"Named Datatype Binds" For information on binding named datatypes (objects)
"Binding REFs" for information on binding REFs

Binding LOBs

There are two ways of binding LOBs:

Bind the LOB locator, rather than the actual LOB values. In this case the LOB value is written or read by passing a LOB locator to the OCI LOB functions.
Bind the LOB value directly, without using the LOB locator.

Binding LOB Locators

Either a single locator or an array of locators can be bound in a single bind call. In each case, the application must pass the address of a LOB locator and not the locator itself. For example, if an application has prepared a SQL statement:

INSERT INTO some_table VALUES (:one_lob)

where one_lob is a bind variable corresponding to a LOB column, and has made the following declaration:

OCILobLocator * one_lob;

Then the following calls would be used to bind the placeholder and execute the statement:

/* initialize single locator */
one_lob = OCIDescriptorAlloc(...OCI_DTYPE_LOB...);
...
/* pass the address of the locator */
OCIBindByName(...,(void *) &one_lob,... SQLT_CLOB, ...);
OCIStmtExecute(...,1,...)                /* 1 is the iters parameter */

You can also insert an array using the same SQL INSERT statement. In this case, the application would include the following code:

OCILobLocator * lob_array[10];
...
for (i=0; i<10, i++)
    lob_array[i] = OCIDescriptorAlloc(...OCI_DTYPE_LOB...);
                                     /* initialize array of locators */
...
OCIBindByName(...,(void *) lob_array,...);
OCIBindArrayOfStruct(...);
OCIStmtExecute(...,10,...);               /* 10 is the iters parameter */

You must allocate descriptors with the OCIDescriptorAlloc() routine before they can be used. In the case of an array of locators, you must initialize each array element using OCIDescriptorAlloc(). Use OCI_DTYPE_LOB as the type parameter when allocating BLOBs, CLOBs, and NCLOBs. Use OCI_DTYPE_FILE when allocating BFILEs.

Restrictions on Binding LOB Locators

Piecewise and callback INSERT or UPDATE operations are not supported.
When using a FILE locator as a bind variable for an INSERT or UPDATE statement, you must first initialize the locator with a directory object and filename, using OCILobFileSetName() before issuing the INSERT or UPDATE statement.

See Also:

Chapter 7, "LOB and BFILE Operations" for more information about the OCI LOB functions

Binding LOB Data

Oracle allows nonzero binds for INSERTs and UPDATEs of any size LOB. So you can bind data into a LOB column using OCIBindByPos(), OCIBindByName(), and PL/SQL binds.

The bind of more than 4 kilobytes of data to a LOB column uses space from the temporary tablespace. Make sure that your temporary tablespace is big enough to hold at least the amount of data equal to the sum of all the bind lengths for LOBs. If your temporary tablespace is extendable, it will be extended automatically after the existing space is fully consumed. Use the following command to create an extendable temporary tablespace:

CREATE TABLESPACE ... AUTOEXTENT ON ... TEMPORARY ...;

Restrictions on Binding LOB Data

If a table has both LONG and LOB columns, then you can have binds of greater than 4 kilobytes for either the LONG column or the LOB columns, but not both in the same statement.
In an INSERT AS SELECT operation, Oracle does not allow binding of any length data to LOB columns.
A special consideration applies on the maximum size of bind variables that are neither LONG or LOB, but which appear after any LOB or LONG bind variable in the SQL statement. You will receive an ORA-24816 error from the Oracle Database if the maximum size for such bind variables exceeds 4000 bytes. To avoid this, you need to set OCI_ATTR_MAXDATA_SIZE to 4000 bytes for any such binds whose maximum size may exceed 4000 bytes on the server side after character set conversion. Alternatively, reorder the binds so that such binds are placed before any LONG or LOBs in the bind list.

See Also:
"Using OCI_ATTR_MAXDATA_SIZE Attribute"

Oracle does not do implicit conversions, such as HEX to RAW or RAW to HEX, for data of size more than 4000 bytes. The following PL/SQL code illustrates this:

create table t (c1 clob, c2 blob);
declare
  text   varchar(32767);
  binbuf raw(32767);
begin
  text := lpad ('a', 12000, 'a');
  binbuf := utl_raw.cast_to_raw(text);

  -- The following works:
  insert into t values (text, binbuf);

  -- The following won't work because Oracle won't do implicit 
  -- hex to raw conversion.
  insert into t (c2) values (text);  

  -- The following won't work because Oracle won't do implicit 
  -- raw to hex conversion.
  insert into t (c1) values (binbuf);  

  -- The following won't work because we can't combine the
  -- utl_raw.cast_to_raw() operator with the >4k bind.
  insert into t (c2) values (utl_raw.cast_to_raw(text));

end;
/

If you bind more than 4000 bytes of data to a BLOB or a CLOB, and the data is filtered by a SQL operator, then Oracle will limit the size of the result to at most 4000 bytes.

For example:

create table t (c1 clob, c2 blob);
-- The following command inserts only 4000 bytes because the result of
-- LPAD is limited to 4000 bytes
insert into t(c1) values (lpad('a', 5000, 'a')); 

-- The following command inserts only 2000 bytes because the result of
-- LPAD is limited to 4000 bytes, and the implicit hex to raw conversion
-- converts it to 2000 bytes of RAW data.
insert into t(c2) values (lpad('a', 5000, 'a'));

Examples of Binding LOB Data

Consider the following SQL statements which will be used in the examples that follow:

CREATE TABLE foo (a INTEGER );
CREATE TYPE lob_typ AS OBJECT (A1 CLOB );
CREATE TABLE lob_long_tab (C1 CLOB, C2 CLOB, CT3 lob_typ, L LONG);

Example1: Binding LOBs

void insert()                 /* A function in an OCI program */
{
   /* The following is allowed */
   ub1 buffer[8000];
   text *insert_sql = (text *) "INSERT INTO lob_long_tab (C1, C2, L) \
                       VALUES (:1, :2, :3)";
   OCIStmtPrepare(stmthp, errhp, insert_sql, strlen((char*)insert_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[1], errhp, 2, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[2], errhp, 3, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example2: Binding LOBs

void insert()
{
   /* The following is allowed */
   ub1 buffer[8000];
   text *insert_sql = (text *) "INSERT INTO lob_long_tab (C1, L) \
                      VALUES (:1, :2)";
   OCIStmtPrepare(stmthp, errhp, insert_sql, strlen((char*)insert_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[1], errhp, 2, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example3: Binding LOBs

void update()
{
   /* The following is allowed, no matter how many rows it updates */
   ub1 buffer[8000];
   text *update_sql = (text *)"UPDATE lob_long_tab SET \
                              C1 = :1, C2=:2, L=:3";
   OCIStmtPrepare(stmthp, errhp, update_sql, strlen((char*)update_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[1], errhp, 2, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[2], errhp, 3, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example4: Binding LOBs

void update()
{
   /* The following is allowed, no matter how many rows it updates */
   ub1 buffer[8000];
   text *update_sql = (text *)"UPDATE lob_long_tab SET \
                               C1 = :1, C2=:2, L=:3";
   OCIStmtPrepare(stmthp, errhp, update_sql, strlen((char*)update_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[1], errhp, 2, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[2], errhp, 3, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                               (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example5: Binding LOBs

void insert()
{
   /* Piecewise, callback and array insert/update operations similar to 
    * the allowed regular insert/update operations are also allowed */
}

Example6: Binding LOBs

void insert()
{
   /* The following is NOT allowed because we try to insert >4000 bytes
    * into both LOB and LONG columns */
   ub1 buffer[8000];
   text *insert_sql = (text *)"INSERT INTO lob_long_tab (C1, L) \
                               VALUES (:1, :2)";
   OCIStmtPrepare(stmthp, errhp, insert_sql, strlen((char*)insert_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[1], errhp, 2, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example7: Binding LOBs

void insert()
{
   /* Insert of data into LOB attributes is allowed */
   ub1 buffer[8000];
   text *insert_sql = (text *)"INSERT INTO lob_long_tab (CT3) \
                               VALUES (lob_typ(:1))";
   OCIStmtPrepare(stmthp, errhp, insert_sql, strlen((char*)insert_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 2000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Example8: Binding LOBs

void insert()
{
   /* The following is NOT allowed because we try to do insert as
    * select character data into LOB column */
   ub1 buffer[8000];
   text *insert_sql = (text *)"INSERT INTO lob_long_tab (C1) SELECT \
                               :1 from FOO";
   OCIStmtPrepare(stmthp, errhp, insert_sql, strlen((char*)insert_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIBindByPos(stmthp, &bindhp[0], errhp, 1, (void *)buffer, 8000,
                SQLT_LNG, 0, 0, 0, 0, 0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *) NULL,
                              (OCISnapshot *) NULL, OCI_DEFAULT);
}

Binding in OCI_DATA_AT_EXEC Mode

If the mode parameter in a call to OCIBindByName() or OCIBindByPos() is set to OCI_DATA_AT_EXEC, an additional call to OCIBindDynamic() is necessary if the application will use the callback method for providing data at runtime. The call to OCIBindDynamic() sets up the callback routines, if necessary, for indicating the data or piece provided. If the OCI_DATA_AT_EXEC mode is chosen, but the standard OCI piecewise polling method will be used instead of callbacks, the call to OCIBindDynamic() is not necessary.

When binding RETURN clause variables, an application must use OCI_DATA_AT_EXEC mode, and it must provide callbacks.

See Also:

"Runtime Data Allocation and Piecewise Operations in OCI" for more information about piecewise operations

Binding REF CURSOR Variables

REF CURSORs are bound to a statement handle with a bind datatype of SQLT_RSET.

Overview of Defining in OCI

Query statements return data from the database to your application. When processing a query, you must define an output variable or an array of output variables for each item in the select-list for retrieving data. The define step creates an association that determines where returned results are stored, and in what format.

For example, if your program processes the following statement you would normally need to define two output variables, one to receive the value returned from the name column, and one to receive the value returned from the ssn column:

SELECT name, ssn FROM employees
    WHERE empno = :empnum

If you were only interested in retrieving values from the name column, you would not need to define an output variable for ssn. If the SELECT statement being processed returns more than a single row for a query, the output variables you define can be arrays instead of scalar values.

Depending on the application, the define step can take place before or after an execute. If you know the datatypes of select-list items at compile time, the define can take place before the statement is executed. If your application is processing dynamic SQL statements entered by you at runtime or statements that do not have a clearly defined select-list, the application must execute the statement to retrieve describe information. After the describe information is retrieved, the type information for each select-list item is available for use in defining output variables.

OCI processes the define call locally on the client side. In addition to indicating the location of buffers where results should be stored, the define step determines what data conversions must take place when data is returned to the application.

Note:

Output buffers must be 2-byte aligned.

The dty parameter of the OCIDefineByPos() call specifies the datatype of the output variable. OCI is capable of a wide range of data conversions when data is fetched into the output variable. For example, internal data in Oracle DATE format can be automatically converted to a String datatype on output.

See Also:

Chapter 3, "Datatypes" For more information about datatypes and conversions
"Describing Select-list Items" for more information

Steps Used in OCI Defining

A basic define is done with a position call, OCIDefineByPos(). This step creates an association between a select-list item and an output variable. Additional define calls may be necessary for certain datatypes or fetch modes. Once the define step is complete, the OCI library determines where to put retrieved data. You can make your define calls again to redefine the output variables without having to re-prepare or re-execute the SQL statement.

The following example shows a scalar output variable being defined following an execute and a describe.

SELECT department_name FROM departments WHERE department_id = :dept_input

   /* The input placeholder was bound earlier, and the data comes from the
   user input below */

  printf("Enter employee dept: ");
  scanf("%d", &deptno);

 /* Execute the statement. If OCIStmtExecute() returns OCI_NO_DATA, meaning that
    no data matches the query, then the department number is invalid. */

  if ((status = OCIStmtExecute(svchp, stmthp, errhp, 0, 0, (OCISnapshot *) 0, (OCISnapshot *) 0,
       OCI_DEFAULT))
      && (status != OCI_NO_DATA))
  {
    checkerr(errhp, status);
    return OCI_ERROR;
  }
  if (status == OCI_NO_DATA) {
    printf("The dept you entered doesn't exist.\n");
    return 0;
  }

   /* The next two statements describe the select-list item, dname, and
   return its length */
  checkerr(errhp, OCIParamGet((void *)stmthp, (ub4) OCI_HTYPE_STMT, errhp, (void **)&parmdp, (ub4) 1));
  checkerr(errhp, OCIAttrGet((void*) parmdp, (ub4) OCI_DTYPE_PARAM,
        (void*) &deptlen, (ub4 *) &sizelen, (ub4) OCI_ATTR_DATA_SIZE,
        (OCIError *) errhp  ));

  /* Use the retrieved length of dname to allocate an output buffer, and
   then define the output variable. If the define call returns an error,
   exit the application */
  dept = (text *) malloc((int) deptlen + 1);
  if (status = OCIDefineByPos(stmthp, &defnp, errhp,
             1, (void *) dept, (sb4) deptlen+1,
             SQLT_STR, (void *) 0, (ub2 *) 0,
             (ub2 *) 0, OCI_DEFAULT))
  {
    checkerr(errhp, status);
    return OCI_ERROR;
  }

See Also:

"Describing Select-list Items" for an explanation of the describe step

Advanced OCI Defines

In some cases the define step requires more than just a call to OCIDefineByPos(). There are additional calls that define the attributes of an array fetch, OCIDefineArrayOfStruct(), or a named datatype fetch, OCIDefineObject(). For example, to fetch multiple rows with a column of named datatypes, all three calls must be invoked for the column; but to fetch multiple rows of scalar columns, OCIDefineArrayOfStruct() and OCIDefineByPos() are sufficient.

Oracle also provides pre-defined C datatypes that map object type attributes.

See Also:

Advanced Define Operations in OCI

This section covers advanced defined operations, including multi-step defines, and defines of named datatypes and REFs.

In some cases the define step requires additional calls that define the attributes of an array fetch, OCIDefineArrayOfStruct(), or a named datatype fetch, OCIDefineObject(). For example, to fetch multiple rows with a column of named datatypes, all the three calls must be invoked for the column. To fetch multiple rows of scalar columns only OCIDefineArrayOfStruct() and OCIDefineByPos() are sufficient.

Defining LOB Output Variables

There are two ways of defining LOBs:

Define as a LOB locator, rather than the actual LOB values. In this case the LOB value is written or read by passing a LOB locator to the OCI LOB functions.
Define as a LOB value directly, without using the LOB locator.

Defining LOB Locators

Either a single locator or an array of locators can be defined in a single define call. In each case, the application must pass the address of a LOB locator and not the locator itself. For example, if an application has prepared a SQL statement like:

SELECT lob1 FROM some_table;

where lob1 is the LOB column and one_lob is a define variable corresponding to a LOB column with the following declaration:

OCILobLocator * one_lob;

The following sequence of steps bind the placeholder, and execute the statement:

/* initialize single locator */
one_lob = OCIDescriptorAlloc(...OCI_DTYPE_LOB...);
...
/* pass the address of the locator */
OCIDefineByPos(...,1, ...,(void *) &one_lob,... SQLT_CLOB, ...);
OCIStmtExecute(...,1,...);                /* 1 is the iters parameter */

You can also do an array select using the same SQL SELECT statement. In this case, the application would include the following steps:

OCILobLocator * lob_array[10];
...
for (i=0; i<10, i++)
   lob_array[i] = OCIDescriptorAlloc(...OCI_DTYPE_LOB...);
                                         /* initialize array of locators */
...
OCIDefineByPos(...,1, (void *) lob_array,... SQLT_CLOB, ...);
OCIDefineArrayOfStruct(...);
OCIStmtExecute(...,10,...);               /* 10 is the iters parameter */

Note that you must allocate descriptors with the OCIDescriptorAlloc() routine before they can be used. In the case of an array of locators, you must initialize each array element using OCIDescriptorAlloc(). Use OCI_DTYPE_LOB as the type parameter when allocating BLOBs, CLOBs, and NCLOBs. Use OCI_DTYPE_FILE when allocating BFILEs.

Defining LOB Data

Oracle allows nonzero defines for SELECTs of any size LOB. So you can select up to the maximum allowed size of data from a LOB column using OCIDefineByPos(), and PL/SQL defines. Because there can be multiple LOBs in a row, you can select the maximum size of data from each one of those LOBs in the same SELECT statement.

Consider the following SQL statements which will be used in the examples that follow:

CREATE TABLE lob_tab (C1 CLOB, C2 CLOB);

Example1: Defining LOBs Before Execution

void select_define_before_execute()      /* A function in an OCI program */
{
  /* The following is allowed */
   ub1 buffer1[8000];
   ub1 buffer2[8000];
   text *select_sql = (text *)"SELECT c1, c2 FROM lob_tab";

   OCIStmtPrepare(stmthp, errhp, select_sql, (ub4)strlen((char*)select_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIDefineByPos(stmthp, &defhp[0], errhp, 1, (void *)buffer1, 8000,
                SQLT_LNG, (void *)0, (ub2 *)0, (ub2 *)0, (ub4) OCI_DEFAULT);
   OCIDefineByPos(stmthp, &defhp[1], errhp, 2, (void *)buffer2, 8000,
                SQLT_LNG, (void *)0, (ub2 *)0, (ub2 *)0, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 1, 0, (OCISnapshot *)0, 
                 (OCISnapshot *)0, OCI_DEFAULT);
}

Example2: Defining LOBs after Execution

void select_execute_before_define()
{
  /* The following is allowed */
   ub1 buffer1[8000];
   ub1 buffer2[8000];
   text *select_sql = (text *)"SELECT c1, c2 FROM lob_tab";

   OCIStmtPrepare(stmthp, errhp, select_sql, (ub4)strlen((char*)select_sql),
                 (ub4) OCI_NTV_SYNTAX, (ub4) OCI_DEFAULT);
   OCIStmtExecute(svchp, stmthp, errhp, 0, 0, (OCISnapshot *)0,
                 (OCISnapshot *)0, OCI_DEFAULT);
   OCIDefineByPos(stmthp, &defhp[0], errhp, 1, (void *)buffer1, 8000,
                 SQLT_LNG, (void *)0, (ub2 *)0, (ub2 *)0, (ub4) OCI_DEFAULT);
   OCIDefineByPos(stmthp, &defhp[1], errhp, 2, (void *)buffer2, 8000,
                 SQLT_LNG, (void *)0, (ub2 *)0, (ub2 *)0, (ub4) OCI_DEFAULT);
   OCIStmtFetch(stmthp, errhp, 1, OCI_FETCH_NEXT, OCI_DEFAULT);
}

Defining PL/SQL Output Variables

Do not use the define calls to define output variables for select-list items in a SQL SELECT statement inside a PL/SQL block. Use OCI bind calls instead.

See Also:

"Information for Named Datatype and REF Defines, and PL/SQL OUT Binds" for more information about defining PL/SQL output variables.

Defining for a Piecewise Fetch

A piecewise fetch requires an initial call to OCIDefineByPos(). An additional call to OCIDefineDynamic() is necessary if the application will use callbacks rather than the standard polling mechanism.

Binding and Defining Arrays of Structures in OCI

Defining arrays of structures requires an initial call to OCIDefineByPos(). An additional call to OCIDefineArrayOfStruct() is necessary to set up each additional parameter, including the skip parameter necessary for arrays of structures operations.

Using arrays of structures can simplify the processing of multi-row, multi-column operations. You can create a structure of related scalar data items, and then fetch values from the database into an array of these structures, or insert values into the database from an array of these structures.

For example, an application may need to fetch multiple rows of data from columns NAME, AGE, and SALARY. The application can include the definition of a structure containing separate fields to hold the NAME, AGE and SALARY data from one row in the database table. The application would then fetch data into an array of these structures.

In order to perform a multi-row, multi-column operation using an array of structures, associate each column involved in the operation with a field in a structure. This association, which is part of OCIDefineArrayOfStruct() and OCIBindArrayOfStruct() calls, specifies where data is stored.

Skip Parameters

When you split column data across an array of structures, it is no longer stored contiguously in the database. The single array of structures stores data as though it were composed of several arrays of scalars. For this reason, you must specify a skip parameter for each field you are binding or defining. This skip parameter is the number of bytes that need to be skipped in the array of structures before the same field is encountered again. In general, this will be equivalent to the byte size of one structure.

Figure 5-2 shows how a skip parameter is determined. In this case the skip parameter is the sum of the sizes of the fields field1, field2, and field3, which is 8 bytes. This equals the size of one structure.

Figure 5-2 Determining Skip Parameters

Description of "Figure 5-2 Determining Skip Parameters"

On some operating systems it may be necessary to set the skip parameter to sizeof(one_array_element) rather than sizeof(struct), because some compilers insert extra bytes into a structure.

Consider an array of C structures consisting of two fields, a ub4 and a ub1:

struct demo {
    ub4 field1;
    ub1 field2;
};
struct demo demo_array[MAXSIZE];

Some compilers insert three bytes of padding after the ub1 so that the ub4 which begins the next structure in the array is properly aligned. In this case, the following statement may return an incorrect value:

skip_parameter = sizeof(struct demo);

On some operating systems this will produce a proper skip parameter of eight. On other systems, skip_parameter will be set to five bytes by this statement. In this case, use the following statement to get the correct value for the skip parameter:

skip_parameter = sizeof(demo_array[0]);

Skip Parameters for Standard Arrays

Arrays of structures are an extension of binding and defining arrays of single variables. When specifying a single-variable array operation, the related skip will be equal to the size of the datatype of the array under consideration. For example, for an array declared as:

text emp_names[4][20];

the skip parameter for the bind or define operation will be 20. Each data element in the array is then recognized as a separate unit, rather than being part of a structure.

OCI Calls Used with Arrays of Structures

Two OCI calls must be used when performing operations involving arrays of structures:

OCIBindArrayOfStruct() for binding fields in arrays of structures for input variables
OCIDefineArrayOfStruct() for defining arrays of structures for output variables.

Note:
When binding or defining for arrays of structures, multiple calls are required. A call to OCIBindByName() or OCIBindByPos() must precede a call to OCIBindArrayOfStruct(), and a call to OCIDefineByPos() must precede a call to OCIDefineArrayOfStruct().

Arrays of Structures and Indicator Variables

The implementation of arrays of structures also supports the use of indicator variables and return codes. You can declare parallel arrays of column-level indicator variables and return codes that correspond to the arrays of information being fetched, inserted, or updated. These arrays can have their own skip parameters, which are specified during OCIBindArrayOfStruct() or OCIDefineArrayOfStruct() calls.

You can set up arrays of structures of program values and indicator variables in many ways. Consider an application that fetches data from three database columns into an array of structures containing three fields. You can set up a corresponding array of indicator variable structures of three fields, each of which is a column-level indicator variable for one of the columns being fetched from the database. A one-to-one relationship between the fields in an indicator struct and the number of select-list items is not necessary.

See Also:

"Indicator Variables" for more information about indicator variables.

Binding and Defining Multiple Buffers

You can specify multiple buffers for use with a single bind or define call. Performance is improved because the number of round trips is decreased, data stored at different non-contiguous addresses is not copied to one contiguous location. CPU time spent and memory used are thus reduced.

The datatype OCIIOV is defined as:

typedef struct OCIIOV
{
  void *bfp;  /* The pointer to a buffer for the data    */
  ub4  bfl;   /* The size of the buffer                  */
}OCIIOV;

The value OCI_IOV for the mode parameter is used in the OCIBindByPos() and OCIBindByName() functions for binding multiple buffers. If this value of mode is specified, the address of OCIIOV must be passed in parameter valuep. The size of the datatype must be passed in the parameter valuesz. For example:

OCIIOV vecarr[NumBuffers];
...
/* For bind at position 1 with dataype int     */
OCIBindByPos(stmthp, bindp, errhp, 1, (void *)&vecarr[0],
             sizeof(int), ... OCI_IOV);
...

The value OCI_IOV for the mode parameter is used in the OCIDefineByPos() function for defining multiple buffers. If this value of mode is specified, the address of OCIIOV is passed in parameter valuep. The size of the datatype must be passed in the parameter valuesz.

See Also:

Example of Using Multiple Bind and Define Buffers

Here is an example that illustrates the use of the structure OCIIOV and its mode values:

/* The following macros mention the maximum length of the data in the 
 * different buffers. */
 
#define    LENGTH_DATE      10
#define    LENGTH_EMP_NAME  100
 
 
/* These 2 macros represent the number of elements in each bind and define 
   array */
#define  NUM_BIND     30
#define  NUM_DEFINE   45
 
/* The bind buffers for inserting dates */
char  buf_1[NUM_BIND][LENGTH_DATE], 
char  buf_2[NUM_BIND * 2][LENGTH_DATE], 
 
/* The bind buffer for inserting emp name */
char  buf_3[NUM_BIND * 3][LENGTH_EMP_NAME], 
 
/* The define buffers */
char  buf_4[NUM_DEFINE][LENGTH_EMP_NAME];
char  buf_5[NUM_DEFINE][LENGTH_EMP_NAME];
 
/* The size of data value for buffers corresponding to the same column must be
   the same and that value is passed in the OCIBind/Define calls.
   buf_4 and buf_5 above have the same data values ie. LENGTH_EMP_NAME though
   the  number of elements are different in the two buffers.
   
*/
OCIBind     *bndhp1 = (OCIBind   *)0;
OCIBind     *bndhp2 = (OCIBind   *)0;
OCIDefine   *defhp  = (OCIDefine *)0;
OCIStmt     *stmthp = (OCIStmt   *)0;
OCIError    *errhp  = (OCIError  *)0;
 
OCIIOV  bvec[2], dvec[2];
 
/* 
Example of how to use indicators and return codes with this feature, 
showing the allocation when using with define. We allocate memory 
for indicator; return code, and the length buffer as one chunk of 
NUM_DEFINE * 2 elements.
*/
short *indname[NUM_DEFINE*2];            /* indicators */
ub4   *alenname[NUM_DEFINE*2];           /* return lengths */
ub2   *rcodename[NUM_DEFINE*2];          /* return codes */
 
static text *insertstr  = 
             "INSERT INTO EMP (EMP_NAME, JOIN_DATE) VALUES (:1, :2)";
static text *selectstr  = "SELECT EMP_NAME FROM EMP";
 
/* Allocate environment, error handles etc and initialize the 
   environment  */
...
/* Prepare the statement with the insert query in order to show the 
   binds */
OCIStmtPrepare  (stmthp, errhp, insertstr,
                 (ub4)strlen((char *)insertstr),
                 (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
 
/* Populate buffers with values. The following represents the simplest
 * way of populating the buffers. However, in an actual scenario 
 * these buffers may  have been populated by data received from different
 * sources. */
 
/* Store the date into the bind buffers for the date. */
strcpy(buf_1[0], "21-SEP-02");
...
strcpy(buf_1[NUM_BIND - 1], "21-OCT-02");
...
strcpy(buf_2[0], "22-OCT-02");
...
strcpy(buf_2[2*NUM_BIND - 1], "21-DEC-02");
...
memset(bvec[0], 0, sizeof(OCIIOV));
memset(bvec[1], 0, sizeof(OCIIOV));
 
/* Set up the addresses in the IO Vector structure */
bvec[0].bfp = buf_1[0];                       /* Buffer address of the data */
bvec[0].bfl = NUM_BIND*LENGTH_DATE;           /* Size of the buffer */
 
/* And so on for other structures as well. */
bvec[1].bfp = buf_2[0];                       /* Buffer address of the data */
bvec[1].bfl = NUM_BIND*2*LENGTH_DATE;         /* Size of the buffer  */ 
 
/* Do the bind for date, using OCIIOV */
OCIBindByPos (stmthp, &bindhp2, errhp, 2, (void *)&bvec[0], 
              sizeof(buf_1[0]), SQLT_STR, 
              (void *)inddate, (ub2 *)alendate, (ub2 *)rcodedate, 0,
              (ub4 *)0,  OCI_IOV);
 
/* Store the employee names into the bind buffers 3 for the names */
strcpy (buf_3[0], "JOHN ");
...
strcpy (buf_3[NUM_BIND *3 - 1], "HARRY");
 
/* Do the bind for employee name */
OCIBindByPos  (stmthp,  &bindhp1, errhp, 1, buf_3[0], sizeof(buf_3[0]),
      SQLT_STR, (void *)indemp, (ub2 *)alenemp, (ub2 *)rcodeemp, 0, 
      (ub4 *)0, OCI_DEFAULT);
 
OCIStmtExecute (svchp, stmthp, errhp, NUM_BIND*3, 0, 
               (OCISnapshot *)0, (OCISnapshot *)0, OCI_DEFAULT);
 
...
/* Now the statement to depict defines */
/* Prepare the statement with the select query in order to show the 
   defines */
OCIStmtPrepare(stmthp, errhp, selectstr,(ub4)strlen((char *)selectstr),
               (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
 
memset(dvec[0], 0, sizeof(OCIIOV);
memset(dvec[1], 0, sizeof(OCIIOV));
 
/* Set up the define vector */
dvec[0].bfp = buf_4[0];
dvec[0].bfl = NUM_DEFINE*LENGTH_EMP_NAME;
 
dvec[1].bfp = buf_5[0];
dvec[1].bfl = NUM_DEFINE*LENGTH_EMP_NAME;
 
/* 
Pass the buffers for the indicator, length of the data and the 
return code.  We need to note that, the buffer where we receive
the data is split into 2 locations, 
each having NUM_DEFINE number of elements. However the indicator
buffer, the actual length buffer, and the return code buffer is one 
single chunk of NUM_DEFINE * 2 elements.
*/
OCIDefineByPos (stmthp, &defhp, errhp, 1, (void *)&dvec[0], 
                sizeof(buf_4[0]), SQLT_STR, (void *)indname, 
                (ub2 *)alenname, (ub2 *)rcodename, OCI_IOV);
 
OCIStmtExecute (svchp, stmthp, errhp, NUM_DEFINE*2, 0, 
                (OCISnapshot*)0, 
                (OCISnapshot*)0, OCI_DEFAULT);
...

DML with RETURNING Clause in OCI

OCI supports the use of the RETURNING clause with SQL INSERT, UPDATE, and DELETE statements. This section outlines the rules for correctly implementing DML statements with the RETURNING clause.

See Also:

For a complete examples, see the demonstration programs included with your Oracle installation. For additional information, refer to Appendix B, "OCI Demonstration Programs"
Oracle Database SQL Language Reference. or more information about the use of the RETURNING clause with INSERT, UPDATE, or DELETE statements

Using DML with RETURNING Clause

Using the RETURNING clause with a DML statement enables you to combine two SQL statements into one, possibly saving you a server round trip. This is accomplished by adding an extra clause to the traditional UPDATE, INSERT, and DELETE statements. The extra clause effectively adds a query to the DML statement.

In OCI, values are returned to the application as OUT bind variables. In the following examples, the bind variables are indicated by a preceding colon, ":". These examples assume the existence of table1, a table that contains columns col1, col2, and col3.

The following statement inserts new values into the database and then retrieves the column values of the affected row from the database, for manipulating inserted rows.

INSERT INTO table1 VALUES (:1, :2, :3)
     RETURNING col1, col2, col3
     INTO :out1, :out2, :out3

The next example updates the values of all columns where the value of col1 falls within a given range, and then returns the affected rows which were modified.

UPDATE table1 SET col1 = col1 + :1, col2 = :2, col3 = :3
     WHERE col1 >= :low AND col1 <= :high
     RETURNING col1, col2, col3
     INTO :out1, :out2, :out3

The DELETE statement deletes the rows where col1 value falls within a given range, and then returns the data from those rows.

DELETE FROM table1 WHERE col1 >= :low AND col2 <= :high 
     RETURNING col1, col2, col3
     INTO :out1, :out2, :out3

Binding RETURNING...INTO variables

Because both the UPDATE and DELETE statements can affect multiple rows in the table, and a DML statement can be executed multiple times in a single OCIExecute() call, how much data will be returned may not be known at runtime. As a result, the variables corresponding to the RETURNING...INTO placeholders must be bound in OCI_DATA_AT_EXEC mode. An application must define its own dynamic data handling callbacks rather than using a polling mechanism.

The returning clause can be particularly useful when working with LOBs. Normally, an application must insert an empty LOB locator into the database, and then SELECT it back out again to operate on it. Using the RETURNING clause, the application can combine these two steps into a single statement:

INSERT INTO some_table VALUES (:in_locator)
    RETURNING lob_column
    INTO :out_locator

An OCI application implements the placeholders in the RETURNING clause as pure OUT bind variables. However, all binds in the RETURNING clause are initially IN and must be properly initialized. To provide a valid value, you can provide a NULL indicator and set that indicator to -1.

An application must adhere to the following rules when working with bind variables in a RETURNING clause:

Bind RETURNING clause placeholders in OCI_DATA_AT_EXEC mode using OCIBindByName() or OCIBindByPos(), followed by a call to OCIBindDynamic() for each placeholder.
When binding RETURNING clause placeholders, supply a valid OUT bind function as the ocbfp parameter of the OCIBindDynamic() call. This function must provide storage to hold the returned data.
The icbfp parameter of OCIBindDynamic() call should provide a default function which returns NULL values when called.
The piecep parameter of OCIBindDynamic() must be set to OCI_ONE_PIECE.

No duplicate binds are allowed in a DML statement with a RETURNING clause, and no duplication between bind variables in the DML section and the RETURNING section of the statement is allowed.

Note:

OCI only supports the callback mechanism for RETURNING clause binds. The polling mechanism is not supported.

OCI Error Handling

The OUT bind function provided to OCIBindDynamic() must be prepared to receive partial results of a statement in the event of an error. If the application has issued a DML statement that is executed 10 times, and an error occurs during the fifth iteration, the server returns the data from iterations 1 through 4. The callback function is still called to receive data for the first four iterations.

DML with RETURNING REF...INTO Clause in OCI

The RETURNING clause can also be used to return a REF to an object which is being inserted into or updated in the database:

UPDATE extaddr e SET e.zip = '12345', e.state ='AZ'
    WHERE e.state = 'CA' AND e.zip = '95117'
    RETURNING REF(e), zip
    INTO :addref, :zip

The preceding statement updates several attributes of an object in an object table and returns a REF to the object (and a scalar ZIP code) in the RETURNING clause.

Binding the Output Variable

Binding the REF output variable in an OCI application requires three steps:

The initial bind information is set using OCIBindByName()
Additional bind information for the REF (including the TDO) is set with OCIBindObject()
A call to OCIBindDynamic()

The following pseudocode shows a function which performs the binds necessary for the preceding example.

sword bind_output(stmthp, bndhp, errhp)
OCIStmt *stmthp;
OCIBind *bndhp[];
OCIError *errhp;
{
  ub4 i;
                                  /* get TDO for BindObject call */
  if (OCITypeByName(envhp, errhp, svchp, (CONST text *) 0,
                   (ub4) 0, (CONST text *) "ADDRESS_OBJECT",
                   (ub4) strlen((CONST char *) "ADDRESS_OBJECT"),
                   (CONST text *) 0, (ub4) 0,
                    OCI_DURATION_SESSION, OCI_TYPEGET_HEADER, &addrtdo))
  {
    return OCI_ERROR;
  }

                         /* initial bind call for both variables */
  if (OCIBindByName(stmthp, &bndhp[2], errhp,
                       (text *) ":addref", (sb4) strlen((char *) ":addref"),
                       (void *) 0, (sb4) sizeof(OCIRef *), SQLT_REF,
                       (void *) 0, (ub2 *)0, (ub2 *)0,
                       (ub4) 0, (ub4 *) 0, (ub4) OCI_DATA_AT_EXEC)
  ||  OCIBindByName(stmthp, &bndhp[3], errhp,
                       (text *) ":zip", (sb4) strlen((char *) ":zip"),
                       (void *) 0, (sb4) MAXZIPLEN, SQLT_CHR,
                       (void *) 0, (ub2 *)0, (ub2 *)0,
                       (ub4) 0, (ub4 *) 0, (ub4) OCI_DATA_AT_EXEC))
  {
    return OCI_ERROR;
  }

                                 /* object bind for REF variable */
  if (OCIBindObject(bndhp[2], errhp, (OCIType *) addrtdo,
          (void **) &addrref[0], (ub4 *) 0, (void **) 0, (ub4 *) 0))
  {
    return OCI_ERROR;
  }


  for (i = 0; i < MAXCOLS; i++)
    pos[i] = i;
                    /* dynamic binds for both RETURNING variables */
  if (OCIBindDynamic(bndhp[2], errhp, (void *) &pos[0], cbf_no_data,
                    (void *) &pos[0], cbf_get_data)
  ||  OCIBindDynamic(bndhp[3], errhp, (void *) &pos[1], cbf_no_data,
                    (void *) &pos[1], cbf_get_data))
  {
    return OCI_ERROR;
  }

  return OCI_SUCCESS;
}

Additional Notes About OCI Callbacks

When a callback function is called, the OCI_ATTR_ROWS_RETURNED attribute of the bind handle tells the application the number of rows being returned in that particular iteration. During the first callback of an iteration you can allocate space for all rows that are returned for that bind variable. During subsequent callbacks of the same iteration, you merely increment the buffer pointer to the correct memory within the allocated space.

Array Interface for DML RETURNING Statements in OCI

OCI provides additional functionality for single-row DML and array DML operations in which each iteration returns more than one row. To take advantage of this feature, you must specify an OUT buffer in the bind call that is at least as big as the iteration count specified by the OCIStmtExecute() call. This is in addition to the bind buffers provided through callbacks.

If any of the iteration returns more than one row, then the application receives an OCI_SUCCESS_WITH_INFO return code. In this case, the DML operation is successful. At this point the application may choose to roll back the transaction or ignore the warning.

Character Conversion in OCI Binding and Defining

This section discusses issues involving character conversions between the client and the server.

Choosing Character Set

If a database column containing character data is defined to be an NCHAR or NVARCHAR2 column, then a bind or define involving that column must take into account special considerations for dealing with character set specifications.

These considerations are necessary in case the width of the client character set is different from the server character set, and also for proper character conversion. During conversion of data between different character sets, the size of the data may increase or decrease by a factor of four. Insure that buffers provided to hold the data are of sufficient size.

In some cases, it may also be easier for an application to deal with NCHAR or NVARCHAR2 data in terms of numbers of characters, rather than numbers of bytes, which is the usual case.

Character Set Form and ID

Each OCI bind and define handle has OCI_ATTR_CHARSET_FORM and OCI_ATTR_CHARSET_ID attributes associated. An application can set these attributes with the OCIAttrSet() call in order to specify the character form and character set ID of the bind/define buffer.

The csform attribute (OCI_ATTR_CHARSET_FORM) indicates the character set of the client buffer, for binds, and the character set in which to store fetched data for defines. It has two possible values:

SQLCS_IMPLICIT - default value, indicates database character set ID for the bind or define buffer and the character buffer data are converted to the server database character set
SQLCS_NCHAR - indicates that the national character set ID for the bind or define buffer and the client buffer data are converted to the server national character set.

If the character set ID attribute, OCI_ATTR_CHARSET_ID, is not specified, either the default value of the database or the national character set ID of the client is used, depending on the value of csform. They are the values specified in the NLS_LANG and NLS_NCHAR environment variables, respectively

Note:

The data is converted and inserted into the database according to the server's database character set ID or national character set ID, regardless of the client-side character set id.
OCI_ATTR_CHARSET_ID must never be set to 0.
The define handle attributes OCI_ATTR_CHARSET_FORM and OCI_ATTR_CHARSET_ID do not affect the LOB types. LOB locators fetched from the server retain their original csforms. There is no CLOB/NCLOB conversion as part of define conversion based on these attributes.

See Also:

Oracle Database Reference for more information about NCHAR data

Implicit Conversion Between CHAR and NCHAR

As the result of implicit conversion between database character sets and national character sets, OCI can support cross binding and cross defining between CHAR and NCHAR. Even though the OCI_ATTR_CHARSET_FORM attribute is set to SQLCS_NCHAR, OCI enables conversion of data to the database character set if the data is inserted into a CHAR column.

Setting Client Character Sets in OCI

You can set the character sets through the OCIEnvNlsCreate() function parameters charset and ncharset. Both of these parameters can be set as OCI_UTF16ID. The charset parameter controls coding of the metadata and CHAR data. ncharset controls coding of NCHAR data. The function OCINlsEnvironmentVariableGet() returns the character set from NLS_LANG and the national character set from NLS_NCHAR.

Here is an example of the use of these functions (OCI provides a typedef called utext to facilitate binding and defining of UTF-16 data):

OCIEnv *envhp; 
ub2 ncsid = 2; /* we8dec */ 
ub2 hdlcsid, hdlncsid; 
OraText thename[20]; 
utext *selstmt = L"SELECT ename FROM emp"; /* UTF16 statement */ 
OCIStmt *stmthp; 
OCIDefine *defhp; 
OCIError *errhp; 
OCIEnvNlsCreate(OCIEnv **envhp, ..., OCI_UTF16ID, ncsid); 
... 
OCIStmtPrepare(stmthp, ..., selstmt, ...); /* prepare UTF16 statement */ 
OCIDefineByPos(stmthp, defnp, ..., 1, thename, sizeof(thename), SQLT_CHR,...); 
OCINlsEnvironmentVariableGet(&hdlcsid, (size_t)0, OCI_NLS_CHARSET_ID, (ub2)0,
     (size_t*)NULL);
OCIAttrSet(defnp, ..., &hdlcsid, 0, OCI_ATTR_CHARSET_ID, errhp); 
           /* change charset ID to NLS_LANG setting*/ 
...

See Also:

Using OCI_ATTR_MAXDATA_SIZE Attribute

Update or insert operations are done through variable binding. When binding variables, specify OCI_ATTR_MAXCHAR_SIZE and OCI_ATTR_MAXDATA_SIZE in the bind handle to indicate character and byte constraints used when inserting data on the server.

These attributes are defined as:

OCI_ATTR_MAXCHAR_SIZE sets the maximum number of characters allowed in the buffer on the server side.
OCI_ATTR_MAXDATA_SIZE sets the maximum number of bytes allowed in the buffer on the server side.

Every bind handle has an OCI_ATTR_MAXDATA_SIZE attribute that specifies the number of bytes allocated on the server to accommodate client-side bind data after character set conversions.

An application will typically set OCI_ATTR_MAXDATA_SIZE to the maximum size of the column or the size of the PL/SQL variable, depending on how it is used. Oracle issues an error if OCI_ATTR_MAXDATA_SIZE is not large enough to accommodate the data after conversion, and the operation will fail.

For IN/INOUT binds, when OCI_ATTR_MAXDATA_SIZE attribute is set, the bind buffer must be large enough to hold the number of characters multiplied by the bytes in each character of the character set.

If OCI_ATTR_MAXCHAR_SIZE is set to a nonzero value such as 100, then if the character set has 2 bytes in each character, the minimum possible allocated size is 200 bytes.

The following scenarios demonstrate some examples of the use of the OCI_ATTR_MAXDATA_SIZE attribute:

Scenario 1: CHAR (source data) -> non-CHAR (destination column)

There are implicit bind conversions of the data. The recommended value of OCI_ATTR_MAXDATA_SIZE is the size of the source buffer multiplied by the worst-case expansion factor between the client and server character sets.
Scenario 2: CHAR (source data) -> CHAR (destination column) or non-CHAR (source data) -> CHAR (destination column)

The recommended value of OCI_ATTR_MAXDATA_SIZE is the size of the column.
Scenario 3: CHAR (source data) -> PL/SQL variable

In this case, the recommended value of OCI_ATTR_MAXDATA_SIZE is the size of the PL/SQL variable.

Using OCI_ATTR_MAXCHAR_SIZE Attribute

OCI_ATTR_MAXCHAR_SIZE enables processing to work with data in terms of number of characters, rather than number of bytes.

For binds, the OCI_ATTR_MAXCHAR_SIZE attribute sets the number of characters reserved on the server to store the bind data.

For example, if OCI_ATTR_MAXDATA_SIZE is set to 100, and OCI_ATTR_MAXCHAR_SIZE is set to 0, then the maximum possible size of the data on the server after conversion is 100 bytes. However, if OCI_ATTR_MAXDATA_SIZE is set to 300, and OCI_ATTR_MAXCHAR_SIZE is set to a nonzero value, such as 100, then if the character set has 2 bytes/character, the maximum possible allocated size is 200 bytes.

For defines, the OCI_ATTR_MAXCHAR_SIZE attribute specifies the maximum number of characters that the client application allows in the return buffer. Its derived byte length overrides the maxlength parameter specified in the OCIDefineByPos() call.

Note:

Regardless of the value of the attribute OCI_ATTR_MAXCHAR_SIZE, the buffer lengths specified in a bind or define call are always in terms of bytes. The actual length values sent and received by you are also in bytes.

Buffer Expansion During OCI Binding

Do not set OCI_ATTR_MAXDATA_SIZE for OUT binds or for PL/SQL binds. Only set OCI_ATTR_MAXDATA_SIZE for INSERT or UPDATE statements.

If neither of these two attributes is set, OCI expands the buffer using its best estimates.

IN Binds

If the underlying column was created using character length semantics, then it is preferable to specify the constraint using OCI_ATTR_MAXCHAR_SIZE. As long as the actual buffer contains less characters than specified in OCI_ATTR_MAXCHAR_SIZE, no constraints are violated at OCI level.

If the underlying column was created using byte length semantics, then use OCI_ATTR_MAXDATA_SIZE in the bind handle to specify the byte constraint on the server. If you also specify an OCI_ATTR_MAXCHAR_SIZE value, then this constraint is imposed when allocating the receiving buffer on the server side.

Dynamic SQL

For dynamic SQL, you can use the explicit describe to get OCI_ATTR_DATA_SIZE and OCI_ATTR_CHAR_SIZE in parameter handles, as a guide for setting OCI_ATTR_MAXDATA_SIZE and OCI_ATTR_MAXCHAR_SIZE attributes in bind handles. It is a good practice to specify OCI_ATTR_MAXDATA_SIZE and OCI_ATTR_MAXCHAR_SIZE to be no more than the actual column width in bytes, or characters.

Buffer Expansion During Inserts

You should avoid unexpected behavior caused by buffer expansion during inserts.

Consider what happens when the database column has character length semantics, and the user tries to insert data using OCIBindByPos() or OCIBindByName() while setting only the OCI_ATTR_MAXCHAR_SIZE to 3000 bytes. The database character set is UTF8 and the client character set is ASCII. Then, in this case although 3000 characters will fit in a buffer of size 3000 bytes for the client, on the server side it might expand to more than 4000 bytes. Unless the underlying column is a LONG or a LOB type, the server will return an error. You can get around this problem by specifying the OCI_ATTR_MAXDATA_SIZE to be 4000, to guarantee that the data will never exceed 4000 bytes.

Constraint Checking During Defining

To select data from columns into client buffers, OCI uses defined variables. You can set an OCI_ATTR_MAXCHAR_SIZE value on the define buffer to impose an additional character length constraint. There is no OCI_ATTR_MAXDATA_SIZE attribute for define handles since the buffer size in bytes serves as the limit on byte length. The define buffer size provided in the OCIDefineByPos() call can be used as the byte constraint.

Dynamic SQL Selects

When sizing buffers for dynamic SQL, always use the OCI_ATTR_DATA_SIZE value in the implicit describe to avoid data loss through truncation. If the database column is created using character length semantics known through OCI_ATTR_CHAR_USED attribute, then you can use the OCI_ATTR_MAXCHAR_SIZE value to set an additional constraint on the define buffer. A maximum number of OCI_ATTR_MAXCHAR_SIZE characters is put in the buffer.

Return Lengths

The following length values are always in bytes regardless of the character length semantics of the database:

The value returned in the alen, or the actual length field in binds and defines.
The value that appears in the length, prefixed in special datatypes like VARCHAR and LONG VARCHAR.
The value of the indicator variable in case of truncation.

The only exception to this rule is for string buffers in OCI_UTF16ID character set id; then the lengths are in UTF-16 units.

Note:

The buffer sizes in the bind and define calls and the piece sizes in the OCIGetPieceInfo() and OCISetPieceInfo() and the callbacks are always in bytes.

General Compatibility Issues for Character Length Semantics in OCI

For a release 9.0 or later client talking to an 8.1 or earlier server, OCI_ATTR_MAXCHAR_SIZE is not understood by the server, so this value will be ignored. If you specify only this value, OCI will derive the corresponding OCI_ATTR_MAXDATA_SIZE value based on the maximum bytes for each character for the client-side character set.
For an 8.1 or earlier client talking to a 9.0 or later server, the client will never be able to specify an OCI_ATTR_MAXCHAR_SIZE value, so the server will consider the client always expecting byte length semantics. This is similar to the situation when the client specifies only OCI_ATTR_MAXDATA_SIZE.

So in both cases, the server and client can exchange information in an appropriate manner.

Code Example for Inserting and Selecting Using OCI_ATTR_MAXCHAR_SIZE

When a column is created by specifying a number N of characters, the actual allocation in the data base will consider the worst scenario in the following table. The real bytes allocated will be a multiple of N, say M times N. Currently, M is three as the maximum bytes for each character in UTF-8.

For example, in the following table EMP, ENAME column is defined as 30 characters and ADDRESS is defined as 80 characters. Then the corresponding byte lengths in database are M*30 or 3*30=90, and M*80 or 3*80=240 respectively.

...
utext ename[31], address[81];
/* E' <= 30+ 1, D' <= 80+ 1, considering null-termination */
sb4 ename_max_chars = EC=20, address_max_chars = ED=60;
 /* EC <= (E' - 1), ED <= (D' - 1) */
sb4 ename_max_bytes = EB=80, address_max_bytes = DB=200;
 /* EB <= M * EC, DB <= M * DC */
text *insstmt = (text *)"INSERT INTO EMP(ENAME, ADDRESS) VALUES (:ENAME, \
:ADDRESS)";
text *selstmt = (text *)"SELECT ENAME, ADDRESS FROM EMP";
...
/* Inserting Column Data */
OCIStmtPrepare(stmthp1, errhp, insstmt, (ub4)strlen((char *)insstmt),
    (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
OCIBindByName(stmthp1, &bnd1p, errhp, (text *)":ENAME",
    (sb4)strlen((char *)":ENAME"),
    (void *)ename, sizeof(ename), SQLT_STR, (void *)&insname_ind,
    (ub2 *)alenp, (ub2 *)rcodep, (ub4)maxarr_len, (ub4 *)curelep, OCI_DEFAULT);
/* either */
OCIAttrSet((void *)bnd1p, (ub4)OCI_HTYPE_BIND, (void *)&ename_max_bytes,
    (ub4)0, (ub4)OCI_ATTR_MAXDATA_SIZE, errhp);
/* or */
OCIAttrSet((void *)bnd1p, (ub4)OCI_HTYPE_BIND, (void *)&ename_max_chars,
    (ub4)0, (ub4)OCI_ATTR_MAXCHAR_SIZE, errhp);
...
/* Retrieving Column Data */
OCIStmtPrepare(stmthp2, errhp, selstmt, strlen((char *)selstmt),
    (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
OCIDefineByPos(stmthp2, &dfn1p, errhp, (ub4)1, (void *)ename,
   (sb4)sizeof (ename),
   SQLT_STR, (void *)&selname_ind, (ub2 *)alenp, (ub2 *)rcodep,
   (ub4)OCI_DEFAULT);
/* if not called, byte semantics is by default */
OCIAttrSet((void *)dfn1p, (ub4)OCI_HTYPE_DEFINE, (void *)&ename_max_chars,
   (ub4)0,
   (ub4)OCI_ATTR_MAXCHAR_SIZE, errhp);
...

Code Example for UTF-16 Binding and Defining

The character set ID in bind and define of the CHAR or VARCHAR2, or in NCHAR or NVARCHAR variant handles can be set to assume that all data will be passed in UTF-16 (Unicode) encoding. To specify UTF-16, set OCI_ATTR_CHARSET_ID = OCI_UTF16ID.

See Also:

"Bind Handle Attributes"

OCI provides a typedef called utext to facilitate binding and defining of UTF-16 data. The internal representation of utext is a 16-bit unsigned integer, ub2. Operating systems where the encoding scheme of the wchar_t datatype conforms to UTF-16 can easily convert utext to the wchar_t datatype using cast operators.

Even for UTF-16 data, the buffer size in bind and define calls is assumed to be in bytes. Users should use the utext datatype as the buffer for input and output data.

The following pseudocode illustrates a bind and define for UTF-16 data:

...
OCIStmt  *stmthp1, *stmthp2;
OCIDefine *dfn1p, *dfn2p;
OCIBind *bnd1p, *bnd2p;
text *insstmt=
      (text *) "INSERT INTO EMP(ENAME, ADDRESS) VALUES (:ename, :address)"; \
text *selname =
      (text *) "SELECT ENAME, ADDRESS FROM EMP";
utext ename[21];   /* Name -    UTF-16 */
utext address[51]; /* Address - UTF-16 */
ub2 csid = OCI_UTF16ID;
sb4 ename_col_len = 20;
sb4 address_col_len = 50;
...
/* Inserting UTF-16 data */
OCIStmtPrepare (stmthp1, errhp, insstmt, (ub4)strlen ((char *)insstmt),
                (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
OCIBindByName (stmthp1, &bnd1p, errhp, (text*)":ENAME",
              (sb4)strlen((char *)":ENAME"),
              (void *) ename, sizeof(ename), SQLT_STR,
              (void *)&insname_ind, (ub2 *) 0, (ub2 *) 0, (ub4) 0,
              (ub4 *)0, OCI_DEFAULT);
OCIAttrSet ((void *) bnd1p, (ub4) OCI_HTYPE_BIND, (void *) &csid,
           (ub4) 0, (ub4)OCI_ATTR_CHARSET_ID, errhp);
OCIAttrSet((void *) bnd1p, (ub4) OCI_HTYPE_BIND, (void *) &ename_col_len,
           (ub4) 0, (ub4)OCI_ATTR_MAXDATA_SIZE, errhp);
...
/* Retrieving UTF-16 data */
OCIStmtPrepare (stmthp2, errhp, selname, strlen((char *) selname),
                (ub4)OCI_NTV_SYNTAX, (ub4)OCI_DEFAULT);
OCIDefineByPos (stmthp2, &dfn1p, errhp, (ub4)1, (void *)ename,
                (sb4)sizeof(ename), SQLT_STR,
                (void *)0, (ub2 *)0, (ub2 *)0, (ub4)OCI_DEFAULT);
OCIAttrSet ((void *) dfn1p, (ub4) OCI_HTYPE_DEFINE, (void *) &csid,
            (ub4) 0, (ub4)OCI_ATTR_CHARSET_ID, errhp);
...

PL/SQL REF CURSORs and Nested Tables in OCI

OCI provides the ability to bind and define PL/SQL REF CURSORs and nested tables. An application can use a statement handle to bind and define these types of variables. As an example, consider this PL/SQL block:

static const text *plsql_block = (text *)
  "begin \
     OPEN :cursor1 FOR SELECT employee_id, last_name, job_id, manager_id, \
             salary, department_id \
             FROM employees WHERE job_id=:job ORDER BY employee_id; \
     OPEN :cursor2 FOR SELECT * FROM departments ORDER BY department_id;
  end;";

An application allocates a statement handle for binding, by calling OCIHandleAlloc(), and then binds the :cursor1 placeholder to the statement handle, as in the following code, where :cursor1 is bound to stm2p.

status = OCIStmtPrepare (stm1p, errhp, (text *) plsql_block,
             strlen((char *)plsql_block), OCI_NTV_SYNTAX, OCI_DEFAULT);
...
status = OCIBindByName (stm1p, (OCIBind **) &bnd1p, errhp,
             (text *)":cursor1", (sb4)strlen((char *)":cursor1"),
             (void *)&stm2p, (sb4) 0,  SQLT_RSET, (void *)0,
               (ub2 *)0, (ub2 *)0, (ub4)0, (ub4 *)0,   (ub4)OCI_DEFAULT);

In this code, stm1p is the statement handle for the PL/SQL block, while stm2p is the statement handle which is bound as a REF CURSOR for later data retrieval. A value of SQLT_RSET is passed for the dty parameter.

As another example, consider the following:

static const text *nst_tab = (text *)
       "SELECT last_name, CURSOR(SELECT department_name, location_id \
        FROM  departments)  FROM employees WHERE last_name = 'FORD'";

The second position is a nested table, which an OCI application can define as a statement handle as follows:

status = OCIStmtPrepare (stm1p, errhp, (text *) nst_tab, 
         strlen((char *)nst_tab), OCI_NTV_SYNTAX, OCI_DEFAULT);
...
status = OCIDefineByPos (stm1p, (OCIDefine **) &dfn2p, errhp, (ub4)2, 
          (void *)&stm2p, (sb4)0, SQLT_RSET, (void *)0, (ub2 *)0,
                     (ub2 *)0, (ub4)OCI_DEFAULT);

After execution, when you fetch a row into stm2p it becomes a valid statement handle.

Note:

If you have retrieved multiple REF CURSORs, you must take care when fetching them into stm2p. If you fetch the first one, you can then perform fetches on it to retrieve its data. However, once you fetch the second REF CURSOR into stm2p, you no longer have access to the data from the first REF CURSOR.

OCI does not support PL/SQL REF CURSORs that were executed in scrollable mode.

OCI does not support scrollable REF CURSORs because you cannot scroll back to the rows already fetched by a REF CURSOR.

Runtime Data Allocation and Piecewise Operations in OCI

You can use OCI to perform piecewise inserts, updates, and fetches of data. You can also use OCI to provide data dynamically in case of array inserts or updates, instead of providing a static array of bind values. You can insert or retrieve a very large column as a series of chunks of smaller size, minimizing client-side memory requirements.

The size of individual pieces is determined at runtime by the application and can be uniform or not.

The piecewise functionality of OCI is particularly useful when performing operations on extremely large blocks of string or binary data, operations involving database columns that store CLOB, BLOB, LONG, RAW, or LONG RAW data.

The piecewise fetch is complete when the final OCIStmtFetch() call returns a value of OCI_SUCCESS.

In both the piecewise fetch and insert, it is important to understand the sequence of calls necessary for the operation to complete successfully. For a piecewise insert, you must call OCIStmtExecute() one time more than the number of pieces to be inserted (if callbacks are not used). This is because the first time OCIStmtExecute() is called, it merely returns a value indicating that the first piece to be inserted is required. As a result, if you are inserting n pieces, you must call OCIStmtExecute() a total of n+1 times.

Similarly, when performing a piecewise fetch, you must call OCIStmtFetch() once more than the number of pieces to be fetched.

Users who are binding to PL/SQL index-by tables can retrieve a pointer to the current index of the table during the OCIStmtGetPieceInfo() calls.

Valid Datatypes for Piecewise Operations

Only some datatypes can be manipulated in pieces. OCI applications can perform piecewise fetches, inserts, or updates of all the following datatypes:

VARCHAR2
STRING
LONG
LONG RAW
RAW
CLOB
BLOB

Another way of using this feature for all datatypes is to provide data dynamically for array inserts or updates. The callbacks should always specify OCI_ONE_PIECE for the piecep parameter of the callback for datatypes that do not support piecewise operations.

Types of Piecewise Operations

You can perform piecewise operations in two ways:

Use calls provided in the OCI library to execute piecewise operations under a polling paradigm.
Employ user-defined callback functions to provide the necessary information and data blocks.

When you set the mode parameter of an OCIBindByPos() or OCIBindByName() call to OCI_DATA_AT_EXEC, it indicates that an OCI application will be providing data for an INSERT or UPDATE dynamically at runtime.

Similarly, when you set the mode parameter of an OCIDefineByPos() call to OCI_DYNAMIC_FETCH, it indicates that an application will dynamically provide allocation space for receiving data at the time of the fetch.

In each case, you can provide the run-time information for the INSERT, UPDATE, or FETCH in one of two ways: through callback functions, or by using piecewise operations. If callbacks are desired, an additional bind or define call is necessary to register the callbacks.

The following sections give specific information about run-time data allocation and piecewise operations for inserts, updates, and fetches.

Note:

Piecewise operations are also valid for SQL and PL/SQL blocks.

Providing INSERT or UPDATE Data at Runtime

When you specify the OCI_DATA_AT_EXEC mode in a call to OCIBindByPos() or OCIBindByName(), the value_sz parameter defines the total size of the data that can be provided at runtime. The application must be ready to provide to the OCI library the run-time IN data buffers on demand as many times as is necessary to complete the operation. When the allocated buffers are no longer required, they must be freed by the client.

Runtime data is provided in one of the two ways:

You can define a callback using the OCIBindDynamic() function, which when called at runtime returns either a piece or the whole data.
If no callbacks are defined, the call to OCIStmtExecute() to process the SQL statement returns the OCI_NEED_DATA error code. The client application then provides the IN/OUT data buffer or piece using the OCIStmtSetPieceInfo() call that specifies which bind and piece are being used.

Performing a Piecewise Insert or Update

Once the OCI environment has been initialized, and a database connection and session have been established, a piecewise insert begins with calls to prepare a SQL or PL/SQL statement and to bind input values. Piecewise operations using standard OCI calls rather than user-defined callbacks do not require a call to OCIBindDynamic().

Note:

Additional bind variables that are not part of piecewise operations may require additional bind calls, depending on their datatypes.

Following the statement preparation and bind, the application performs a series of calls to OCIStmtExecute(), OCIStmtGetPieceInfo() and OCIStmtSetPieceInfo() to complete the piecewise operation. Each call to OCIStmtExecute() returns a value that determines what action should be performed next. In general, the application retrieves a value indicating that the next piece needs to be inserted, populates a buffer with that piece, and then executes an insert. When the last piece has been inserted, the operation is complete.

Keep in mind that the insert buffer can be of arbitrary size and is provided at runtime. In addition, each inserted piece does not need to be of the same size. The size of each piece to be inserted is established by each OCIStmtSetPieceInfo() call.

Note:

If the same piece size is used for all inserts, and the size of the data being inserted is not evenly divisible by the piece size, the final inserted piece will be smaller. You must account for this by indicating the smaller size in the final OCIStmtSetPieceInfo() call.

The procedure is illustrated in Figure 5-3.

Figure 5-3 Performing Piecewise Insert

Description of "Figure 5-3 Performing Piecewise Insert"

Initialize the OCI environment, allocate the necessary handles, connect to a server, authorize a user, and prepare a statement request.
Bind a placeholder using OCIBindByName() or OCIBindByPos(). You do not need to specify the actual size of the pieces you will use, but you must provide the total size of the data that can be provided at runtime.
Call OCIStmtExecute() for the first time. No data is being inserted here, and the OCI_NEED_DATA error code is returned to the application. If any other value is returned, it indicates that an error occurred.
Call OCIStmtGetPieceInfo() to retrieve information about the piece that needs to be inserted. The parameters of OCIStmtGetPieceInfo() include a pointer to a value indicating if the required piece is the first piece, OCI_FIRST_PIECE, or a subsequent piece, OCI_NEXT_PIECE.
The application populates a buffer with the piece of data to be inserted and calls OCIStmtSetPieceInfo() with these parameters:
- a pointer to the piece
- a pointer to the length of the piece
- a value indicating whether this is the
1. first piece, OCI_FIRST_PIECE
2. an intermediate piece, OCI_NEXT_PIECE
3. the last piece, OCI_LAST_PIECE
Call OCIStmtExecute() again. If OCI_LAST_PIECE was indicated in step 5 and OCIStmtExecute() returns OCI_SUCCESS, all pieces were inserted successfully. If OCIStmtExecute() returns OCI_NEED_DATA, go back to Step 3 for the next insert. If OCIStmtExecute() returns any other value, an error occurred.

The piecewise operation is complete when the final piece has been successfully inserted. This is indicated by the OCI_SUCCESS return value from the final OCIStmtExecute() call.

Piecewise updates are performed in a similar manner. In a piecewise update operation the insert buffer is populated with data that is being updated and OCIStmtExecute() is called to execute the update.

Piecewise Operations with PL/SQL

An OCI application can perform piecewise operations with PL/SQL for IN, OUT, and IN/OUT bind variables in a method similar to that outlined previously. Keep in mind that all placeholders in PL/SQL statements are bound, rather than defined. The call to OCIBindDynamic() specifies the appropriate callbacks for OUT or IN/OUT parameters.

Providing FETCH Information at Runtime

When a call is made to OCIDefineByPos() with the mode parameter set to OCI_DYNAMIC_FETCH, an application can specify information about the data buffer at the time of fetch. You may also need to call OCIDefineDynamic() to set callback function that will be invoked to get information about your data buffer.

Run-time data is provided in one of the two ways:

You can define a callback using the OCIDefineDynamic(). The value_sz parameter defines the maximum size of the data that will be provided at runtime. When the client library needs a buffer to return the fetched data, the callback will be invoked to provide a run-time buffer into which a either piece or the whole data will be returned.
If no callbacks are defined, the OCI_NEED_DATA error code is returned and the OUT data buffer or piece can then be provided by the client application using OCIStmtSetPieceInfo(). The OCIStmtGetPieceInfo() call provides Information about which define and which piece are involved.

Performing a Piecewise Fetch

The fetch buffer can be of arbitrary size. In addition, each fetched piece does not need to be of the same size. The only requirement is that the size of the final fetch must be exactly the size of the last remaining piece. The size of each piece to be fetched is established by each OCIStmtSetPieceInfo() call. This process is illustrated in Figure 5-4.

Figure 5-4 Performing Piecewise Fetch

Description of "Figure 5-4 Performing Piecewise Fetch"

Initialize the OCI environment, allocate necessary handles, connect to a database, authorize a user, prepare a statement, and execute the statement.
Define an output variable using OCIDefineByPos(), with mode set to OCI_DYNAMIC_FETCH. At this point you do not need to specify the actual size of the pieces you will use, but you must provide the total size of the data that will be fetched at runtime.
Call OCIStmtFetch() for the first time. No data is retrieved, and the OCI_NEED_DATA error code is returned to the application. If any other value is returned, an error occurred.
Call OCIStmtGetPieceInfo() to obtain information about the piece to be fetched. The piecep parameter indicates whether it is the first piece, OCI_FIRST_PIECE, a subsequent piece, OCI_NEXT_PIECE, or the last piece, OCI_LAST_PIECE.
Call OCIStmtSetPieceInfo() to specify the fetch buffer.
Call OCIStmtFetch() again to retrieve the actual piece. If OCIStmtFetch() returns OCI_SUCCESS, all the pieces have been fetched successfully. If OCIStmtFetch() returns OCI_NEED_DATA, return to Step 4 to process the next piece. If any other value is returned, an error occurred.

Piecewise Binds and Defines for LOBs

There are two ways of doing piecewise binds and defines for LOBs:

Using the data interface

You can bind or define character data for CLOB columns using SQLT_CHR (VARCHAR2) or SQLT_LNG (LONG) as the input datatype for the following functions. You can also bind or define raw data for BLOB columns using SQLT_LBI (LONG RAW), and SQLT_BIN (RAW) as the input datatype for these functions:
- OCIDefineByPos()
- OCIBindByName()
- OCIBindByPos(}
  See Also:
  - "Binding LOB Data" for usage and examples for both INSERT and UPDATE statements
  - "Defining LOB Data" for usage and examples of SELECT statements
All the piecewise operations described later are supported for CLOB and BLOB columns in this case.
Using the LOB locator

You can bind or define a LOB locator for CLOB and BLOB columns using SQLT_CLOB (CLOB) or SQLT_BLOB (BLOB) as the input datatype for the following functions.
- OCIDefineByPos()
- OCIBindByName()
- OCIBindByPos(}
You must then call OCILob* functions to read and manipulate the data. OCILobRead2() and OCILobWrite2() support piecewise and callback modes.

See Also:

"OCILobRead2()"
"OCILobWrite2()"
"LOB Read and Write Callbacks" for information about streaming using callbacks with OCILobWrite2() and OCILobRead2().