21 Performance Extensions

This chapter describes the Oracle performance extensions to the Java Database Connectivity (JDBC) standard.

This chapter covers the following topics:

Note:

Starting from Oracle Database 12c Release 1 (12.1), Oracle update batching is deprecated. Oracle recommends that you use standard JDBC batching instead of Oracle update batching.

Update Batching

You can reduce the number of round-trips to the database, thereby improving application performance, by grouping multiple UPDATE, DELETE, or INSERT statements into a single batch and having the whole batch sent to the database and processed in one trip. This is referred to as 'update batching'.

Note:

The JDBC 2.0 specification refers to 'update batching' as 'batch updates'.

This is especially useful with prepared statements, when you are repeating the same statement with different bind variables.

Oracle JDBC supports two distinct models for update batching:

  • The standard model, implementing the JDBC 2.0 specification, which is referred to as standard update batching

  • The Oracle-specific model, independent of the JDBC 2.0 specification, which is referred to as Oracle update batching

Note:

It is important to be aware that you cannot mix these models. In any single application, you can use one model or the other, but not both. Oracle JDBC driver will throw exceptions when you mix these.

This section covers the following topics:

Overview of Update Batching Models

This section compares and contrasts the general models and types of statements supported for standard update batching and Oracle update batching.

Note:

Starting from Oracle Database 12c Release 1 (12.1), Oracle update batching is deprecated. Oracle recommends that you use standard JDBC batching instead of Oracle update batching.

Oracle Model Versus Standard Model

Oracle update batching uses a batch value that typically results in implicit processing of a batch. The batch value is the number of operations you want to add to a batch for each trip to the database. As soon as that many operations have been added to the batch, the batch is processed. Note the following:

  • You can set a default batch for the connection object, which applies to any prepared statement run in that connection.

  • For any individual prepared statement object, you can set a statement batch value that overrides the connection batch value.

  • You can choose to explicitly process a batch at any time, overriding both the connection batch value and the statement batch value.

Standard update batching is a manual, explicit model. There is no batch value. You manually add operations to the batch, and then, explicitly choose when to process the batch.

Note:

  • Oracle recommends that you use JDBC standard features when possible. This recommendation applies to update batching as well. Oracle update batching is retained primarily for backwards compatibility.

  • For both standard update batching and Oracle update batching, Oracle recommends you to keep the batch sizes in the general range of 50 to 100. This is because though the drivers support larger batches, they in turn result in a large memory footprint with no corresponding increase in performance. Very large batches usually result in a decline in performance compared to smaller batches.

Types of Statements Supported

As implemented by Oracle, update batching is intended for use with prepared statements, when you are repeating the same statement with different bind variables. Be aware of the following:

  • Oracle update batching supports only prepared statement objects. For a callable statement, both the connection default batch value and the statement batch value are overridden with a value of 1. In an Oracle generic statement, there is no statement batch value, and the connection default batch value is overridden with a value of 1.

  • To adhere to the JDBC 2.0 standard, Oracle implementation of standard update batching supports callable statements, without OUT parameters, and generic statements, as well as prepared statements. You can migrate standard update batching into an Oracle JDBC application without difficulty.

  • You can batch only UPDATE, INSERT, or DELETE operations. Processing a batch that includes an operation that attempts to return a result set will cause an exception.

    Note:

    The Oracle implementation of standard update batching does not implement true batching for generic statements and callable statements. Although Oracle JDBC supports the use of standard batching syntax for Statement and CallableStatement objects, you will see performance improvement for only PreparedStatement objects.

Oracle Update Batching

The Oracle update batching feature associates a batch value with each prepared statement object. With Oracle update batching, instead of the JDBC driver running a prepared statement operation each time the executeUpdate method is called, the driver adds the statement to a batch of accumulated processing requests. The driver will pass all the operations to the database for processing once the batch value is reached. For example, if the batch value is 10, then each batch of 10 operations will be sent to the database and processed in one trip.

A method in the OracleConnection class enables you to set a default batch value for the Oracle connection as a whole, and this batch value applies to any Oracle prepared statement in the connection. For any particular Oracle prepared statement, a method in the OraclePreparedStatement class enables you to set a statement batch value that overrides the connection batch value. You can also override both batch values by choosing to manually process the pending batch.

Note:

  • Do not mix standard update batching with Oracle update batching in the same application. The JDBC driver will throw an exception when you mix these.

  • Disable auto-commit mode if you use update batching model. In case an error occurs while you are processing a batch, this provides you the option of committing or rolling back the operations that ran successfully prior to the error.

Oracle Update Batching Characteristics and Limitations

Note the following limitations and implementation details regarding Oracle update batching:

  • By default, there is no statement batch value and the connection batch value is 1.

  • Batch values between 5 and 30 tend to be the most effective. Setting a very high value might even have a negative effect. It is worth trying different values to verify the effectiveness for your particular application.

  • Regardless of the batch value in effect, if any of the bind variables of an Oracle prepared statement is of a data type which requires data to be streamed, for instance LOB, CLOB, then Oracle JDBC driver sets the batch value to 1 and sends any queued requests to the database for processing.

  • Oracle JDBC driver implicitly runs the sendBatch method of an Oracle prepared statement in any of the following circumstances:

    • The connection receives a COMMIT request as a result of calling the commit method.

    • The statement receives a close request.

    • The connection receives a close request.

    Note:

    A connection COMMIT request, statement close, or connection close has an effect on a pending batch only if you use Oracle update batching. However, if you use standard update batching, then it has no effect on a pending batch.
  • If the connection receives a ROLLBACK request before sendBatch has been called, then the pending batched operations are not removed. You must explicitly call clearBatch to do this.

Setting the Connection Batch Value

You can specify a default batch value for any Oracle prepared statement in your Oracle connection. To do this, use the setDefaultExecuteBatch method of the OracleConnection object. For example, the following code sets the default batch value to 20 for all prepared statement objects associated with the conn connection object:

((OracleConnection)conn).setDefaultExecuteBatch(20);

The connection batch value will apply to statement objects created after this batch value was set.

Note that instead of calling the setDefaultExecuteBatch method, you can set the defaultBatchValue Java property if you use a Java Properties object in establishing the connection.

Setting the Statement Batch Value

Use the following steps to set the statement batch value for a particular Oracle prepared statement. This will override any connection batch value set using the setDefaultExecuteBatch method of the OracleConnection instance for the connection in which the statement is processed.

  1. Write your prepared statement, and specify input values for the first row, as follows:

    PreparedStatement ps = conn.prepareStatement("INSERT INTO dept VALUES (?,?,?)");
    ps.setInt(1,12);
    ps.setString(2,"Oracle");
    ps.setString(3,"USA");
    
  2. Cast your prepared statement to OraclePreparedStatement, and apply the setExecuteBatch method. In this example, the batch size of the statement is set to 2.

    ((OraclePreparedStatement)ps).setExecuteBatch(2);
    

    If you wish, insert the getExecuteBatch method at any point in the program to check the default batch value for the statement, as follows:

    System.out.println(" Statement Execute Batch Value " +
                       ((OraclePreparedStatement)ps).getExecuteBatch());
    
  3. If you send an execute-update call to the database at this point, then no data will be sent to the database, and the call will return 0.

    // No data is sent to the database by this call to executeUpdate
    System.out.println("Number of rows inserted so far: " + ps.executeUpdate ());
    
  4. If you enter a set of input values for a second row and an execute-update, then the number of batch calls to executeUpdate will be equal to the batch value of 2. The data will be sent to the database, and both rows will be inserted in a single round-trip. Since this executeUpdate call is at the batch value boundary, and all the operations in the batch are executed on the server, it will return the total number of affected rows by the whole batch.

    ps.setInt(1, 11);
    ps.setString(2, "Applications");
    ps.setString(3, "Indonesia");
    
    int rows = ps.executeUpdate();
    System.out.println("Number of rows inserted now: " + rows);
    
    ps.close();
    

Note:

The batch value counter will reset internally if the data type for a column changes in between the batch. The total value of affected rows by the whole batch of previously batched items before the data type change, and the newly batched items will be returned at the new batch value boundary. For instance:

If the batch value is 8 after binding 4 integers to a varchar2 column, a string is bound to the column. This will trigger an internal bind value counter reset, and if the application continues to bind strings, then the batch will be executed on the 12th executeUpdate call, and will return the affected row count of all 12 operations.

Checking the Batch Value

To check the overall connection batch value of an Oracle connection instance, use the OracleConnection class getDefaultExecuteBatch method:

Integer batch_val = ((OracleConnection)conn).getDefaultExecuteBatch();

To check the particular statement batch value of an Oracle prepared statement, use the OraclePreparedStatement class getExecuteBatch method:

Integer batch_val = ((OraclePreparedStatement)ps).getExecuteBatch();

Note:

If no statement batch value has been set, then getExecuteBatch will return the connection batch value.

Overriding the Batch Value

If you want to process accumulated operations before the batch value in effect is reached, then use the sendBatch method of the OraclePreparedStatement object.

For this example, presume you set the connection batch value to 20. This sets the default batch value for all prepared statement objects associated with the connection to 20. You can accomplish this by casting your connection to OracleConnection and applying the setDefaultExecuteBatch method for the connection, as follows:

((OracleConnection)conn).setDefaultExecuteBatch (20);

Override the batch value as follows:

  1. Write your prepared statement, specify input values for the first row, and then process the statement, as follows:

    PreparedStatement ps =
       conn.prepareStatement("insert into dept values (?, ?, ?)");
        
    ps.setInt(1, 32);
    ps.setString(2, "Oracle");
    ps.setString(3, "USA");
    
    System.out.println (ps.executeUpdate ()); 
    

    The batch is not processed at this point. The ps.executeUpdate method returns 0.

  2. If you enter a set of input values for a second operation and call executeUpdate again, then the data will still not be sent to the database, because the batch value in effect for the statement is the connection batch value, which is 20.

    ps.setInt(1, 33);
    ps.setString(2, "Applications");
    ps.setString(3, "Indonesia");
    
    // this batch is still not executed at this point
    int rows = ps.executeUpdate();
        
    System.out.println("Number of rows updated before calling sendBatch: "  + rows);
    

    Note that the value of rows in the println statement is 0.

  3. If you apply the sendBatch method at this point, then the two previously batched operations will be sent to the database in a single round-trip. The sendBatch method also returns the total number of updated rows. This property of sendBatch is used by println to print the number of updated rows.

    // Execution of both previously batched executes will happen
    // at this point. The number of rows updated will be
    // returned by sendBatch.
    rows = ((OraclePreparedStatement)ps).sendBatch();
    
    System.out.println("Number of rows updated by calling sendBatch: " + rows);
    ps.close();
    

Committing the Changes in Oracle Batching

After you process the batch, you must still commit the changes, presuming auto-commit is disabled as recommended.

Calling the commit method on the connection object in Oracle batching not only commits operations in batches that have been processed, but also issues an implicit sendBatch call to process all pending batches. So, the commit method effectively commits changes for all operations that have been added to a batch.

Update Counts in Oracle Batching

In a nonbatching situation, the executeUpdate method of an OraclePreparedStatement object returns the number of database rows affected by the operation.

In an Oracle batching situation, this method returns the number of rows affected at the time the method is invoked, as follows:

  • If an executeUpdate call results in the operation being added to the batch, then the method returns a value of 0, because nothing was written to the database yet.

  • If an executeUpdate call results in the batch value being reached and the batch being processed, then the method returns the total number of rows affected by all operations in the batch.

Similarly, the sendBatch method of an OraclePreparedStatement object returns the total number of rows affected by all operations in the batch.

Example 21-1 illustrates the use of Oracle update batching.

Example 21-1 Oracle Update Batching

The following example illustrates how you use the Oracle update batching feature. It assumes you have imported the oracle.driver.* interfaces.

...
OracleDataSource ods = new OracleDataSource();
ods.setURL("jdbc:oracle:oci);
ods.setUser("HR");
ods.setPassword("hr");

Connection conn = ods.getConnection();
conn.setAutoCommit(false);

PreparedStatement ps = 
  conn.prepareStatement("insert into dept values (?, ?, ?)"); 
     
//Change batch size for this statement to 3 
((OraclePreparedStatement)ps).setExecuteBatch (3);
 
ps.setInt(1, 23); 
ps.setString(2, "Sales"); 
ps.setString(3, "USA"); 
ps.executeUpdate(); //JDBC queues this for later execution 
 
ps.setInt(1, 24); 
ps.setString(2, "Blue Sky"); 
ps.setString(3, "Montana"); 
ps.executeUpdate(); //JDBC queues this for later execution 
 
ps.setInt(1, 25); 
ps.setString(2, "Applications"); 
ps.setString(3, "India"); 
ps.executeUpdate(); //The queue size equals the batch value of 3 
                    //JDBC sends the requests to the database

ps.setInt(1, 26); 
ps.setString(2, "HR"); 
ps.setString(3, "Mongolia"); 
ps.executeUpdate(); //JDBC queues this for later execution
 
((OraclePreparedStatement)ps).sendBatch(); // JDBC sends the queued request
conn.commit();

ps.close();
...

Note:

Updates deferred through batching can affect the results of other queries. In the following example, if the first query is deferred due to batching, then the second will return unexpected results:
UPDATE emp SET name = "Sue" WHERE name = "Bob";
SELECT name FROM emp WHERE name = "Sue";

Error Reporting in Oracle Update Batching

If any one of the batched operations fails to complete successfully or attempts to return a result set during an executeBatch call, then the processing stops and a java.sql.BatchUpdateException is generated.

If the exception is raised, you can call the getUpdateCounts method on the BatchUpdateException object to retrieve the update count. This method returns an int array of update counts, just as the executeBatch method does.

Prior to Oracle Database 11g Release 1, the integer array returned contains n Statement.EXECUTE_FAILED entries, where n is the size of the batch. However, this does not indicate where in the batch the error occurred. The only option you have is to rollback the transaction.

Starting from Oracle Database 11g Release 1, the integer array returned contains n Statement.SUCCESS_NO_INFO entries, where n is the number of elements in the batch that have been successfully executed.

Note:

The execution of the batch always stops with the first element of the batch that generates an error.

Standard Update Batching

JDBC standard update batching, unlike the Oracle update batching model, depends on explicitly adding statements to the batch using an addBatch method and explicitly processing the batch using an executeBatch method. In the Oracle model, you call executeUpdate as in a nonbatching situation, but whether an operation is added to the batch or the whole batch is processed is typically determined implicitly, depending on whether or not a predetermined batch value is reached.

Note:

  • Do not mix standard update batching with Oracle update batching in the same application. Oracle JDBC driver will throw exceptions when these are mixed.

  • Disable auto-commit mode if you use either update batching model. In case an error occurs while you are processing a batch, this provides you the option of committing or rolling back the operations that ran successfully prior to the error.

Limitations in the Oracle Implementation of Standard Batching

This section discusses the limitations and implementation details regarding the Oracle implementation of standard update batching.

In Oracle JDBC applications, update batching is intended for use with prepared statements that are being processed repeatedly with different sets of bind values.

The Oracle implementation of standard update batching does not implement true batching for generic statements and callable statements. Even though Oracle JDBC supports the use of standard batching for Statement and CallableStatement objects, you are unlikely to see performance improvement.

Adding Operations to the Batch

When any statement object is first created, its statement batch is empty. Use the standard addBatch method to add an operation to the statement batch. This method is specified in the standard java.sql.Statement, PreparedStatement, and CallableStatement interfaces, which are implemented by the oracle.jdbc.OracleStatement, OraclePreparedStatement, and OracleCallableStatement interfaces, respectively.

For a Statement object, the addBatch method takes a Java String with a SQL operation as input. For example:

...
Statement stmt = conn.createStatement();

stmt.addBatch("INSERT INTO emp VALUES(1000, 'Joe Jones')");
stmt.addBatch("INSERT INTO dept VALUES(260, 'Sales')");
stmt.addBatch("INSERT INTO emp_dept VALUES(1000, 260)");
...

At this point, three operations are in the batch.

Note:

Remember, however, that in the Oracle implementation of standard update batching, you will probably see no performance improvement in batching generic statements.

For prepared statements, update batching is used to batch multiple runs of the same statement with different sets of bind parameters. For a PreparedStatement or OraclePreparedStatement object, the addBatch method takes no input. It simply adds the operation to the batch using the bind parameters last set by the appropriate setXXX methods. This is also true for CallableStatement or OracleCallableStatement objects, but remember that in the Oracle implementation of standard update batching, you will probably see no performance improvement in batching callable statements.

For example:

...
PreparedStatement pstmt = 
          conn.prepareStatement("INSERT INTO employees VALUES(?, ?)");

pstmt.setInt(1, 2000);
pstmt.setString(2, "Milo Mumford");
pstmt.addBatch();

pstmt.setInt(1, 3000);
pstmt.setString(2, "Sulu Simpson");
pstmt.addBatch();
...

At this point, two operations are in the batch.

Because a batch is associated with a single prepared statement object, you can batch only repeated runs of a single prepared statement, as in this example.

Processing the Batch

To process the current batch of operations, use the executeBatch method of the statement object. This method is specified in the standard Statement interface, which is extended by the standard PreparedStatement and CallableStatement interfaces.

Note:

If you add too many operations to a batch by calling the addBatch method several times and create a very large batch (for example, with more than or equal to 100,000 rows), then while calling the executeBatch method on the whole batch, you may face severe performance problems in terms of memory. To avoid this issue, the JDBC driver transparently breaks up the large batches into smaller internal batches and makes a roundtrip to the server for each internal batch. This makes your application slightly slower because of each round-trip overhead, but optimizes memory significantly. However, if each bound row is very large in size (for example, more than about 1MB each or so), then this process can impact the overall performance negatively because in such a case, the performance gained in terms of memory will be less than the performance lost in terms of time.

Following is an example that repeats the prepared statement addBatch calls shown previously and then processes the batch:

...
PreparedStatement pstmt = 
          conn.prepareStatement("INSERT INTO employees VALUES(?, ?)");

pstmt.setInt(1, 2000);
pstmt.setString(2, "Milo Mumford");
pstmt.addBatch();

pstmt.setInt(1, 3000);
pstmt.setString(2, "Sulu Simpson");
pstmt.addBatch();

int[] updateCounts = pstmt.executeBatch();
...

Row Count per Iteration for Array DMLs

Starting from Oracle Database 12c Release 1 (12.1), the executeBatch method has been improved so that it returns an int array of size that is the same as the number of records in the batch and each item in the return array is the number of database table rows affected by the corresponding record of the batch. For example, if the batch size is 5, then the executeBatch method returns an array of size 5. In case of an error in between execution of the batch, the executeBatch method cannot return a value, instead it throws a BatchUpdateException. In this case, the exception itself carries an int array of size n as its data, where n is the number of successful record executions. For example, if the batch is of size 5 and the error occurs at the 4th record, then the BatchUpdateException has an array of size 3 (3 records executed successfully) and each item in the array represents how many rows were affected by each of them.

Committing the Changes in the Oracle Implementation of Standard Batching

After you process the batch, you must still commit the changes, presuming auto-commit is disabled as recommended.

Calling commit, commits nonbatched operations and batched operations for statement batches that have been processed, but for the Oracle implementation of standard batching, has no effect on pending statement batches that have not been processed.

Clearing the Batch

To clear the current batch of operations instead of processing it, use the clearBatch method of the statement object. This method is specified in the standard Statement interface, which is extended by the standard PreparedStatement and CallableStatement interfaces.

Keep the following things in mind:

  • When a batch is processed, operations are performed in the order in which they were batched.

  • After calling addBatch, you must call either executeBatch or clearBatch before a call to executeUpdate, otherwise there will be a SQL exception.

  • A clearBatch or executeBatch call resets the statement batch to empty.

  • The statement batch is not reset to empty if the connection receives a ROLLBACK request. You must explicitly call clearBatch to reset it.

    Note:

    • Oracle recommends not to use the clearBatch method with Oracle update batching. With Oracle update batching, binds can be implicitly flushed to the server at any point of time due to various conditions. This makes the behavior of the clearBatch method vary with each case and unpredictable at run time because the method cannot clear any bound data that is already flushed to the server. However, it is OK to use the clearBatch method after a rollback to clear any traces.

      Also, the clearBatch method displays this unpredictable behavior with only Oracle update batching and works correctly with standard update batching.

    • If you are using Oracle update batching in Oracle Database 12c Release 1 (12.1), then you do not have to clear your batches explicitly in the code after a rollback. However, if you are using Oracle update batching in an earlier release, then you have to invoke the clearBatch method to clear your batches explicitly after a rollback.

  • Invoking clearBatch method after a rollback works for all releases.

  • An executeBatch call closes the current result set of the statement object, if one exists.

  • Nothing is returned by the clearBatch method.

Following is an example that repeats the prepared statement addBatch calls shown previously but then clears the batch under certain circumstances:

...
PreparedStatement pstmt = 
          conn.prepareStatement("INSERT INTO employees VALUES(?, ?)");

pstmt.setInt(1, 2000);
pstmt.setString(2, "Milo Mumford");
pstmt.addBatch();

pstmt.setInt(1, 3000);
pstmt.setString(2, "Sulu Simpson");
pstmt.addBatch();

if (...condition...)
{
   int[] updateCounts = pstmt.executeBatch();
   ...
}
else
{
   pstmt.clearBatch();
   ...
}

Update Counts in the Oracle Implementation of Standard Batching

If a statement batch is processed successfully, then the integer array, or update counts array, returned by the statement executeBatch call will always have one element for each operation in the batch. In the Oracle implementation of standard update batching, the values of the array elements are as follows:

  • For a prepared statement batch, the array contains the actual update counts indicating the number of rows affected by each operation.

  • For a generic statement batch, the array contains the actual update counts indicating the number of rows affected by each operation. The actual update counts can be provided only in the case of generic statements in the Oracle implementation of standard batching.

  • For a callable statement batch, the array contains the actual update counts indicating the number of rows affected by each operation.

In your code, upon successful processing of a batch, you should be prepared to handle either -2, 1, or true update counts in the array elements. For a successful batch processing, the array contains either all -2, 1, or all positive integers.

Example 21-2 illustrates the use of standard update batching.

Example 21-2 Standard Update Batching

This example combines the sample fragments in the previous sections, accomplishing the following steps:

  1. Disabling auto-commit mode, which you should always do when using either update batching model

  2. Creating a prepared statement object

  3. Adding operations to the batch associated with the prepared statement object

  4. Processing the batch

  5. Committing the operations from the batch

conn.setAutoCommit(false);

PreparedStatement pstmt = 
          conn.prepareStatement("INSERT INTO employees VALUES(?, ?)");

pstmt.setInt(1, 2000);
pstmt.setString(2, "Milo Mumford");
pstmt.addBatch();

pstmt.setInt(1, 3000);
pstmt.setString(2, "Sulu Simpson");
pstmt.addBatch();

int[] updateCounts = pstmt.executeBatch();

conn.commit();

pstmt.close();
...

You can process the update counts array to determine if the batch processed successfully.

Error Handling in the Oracle Implementation of Standard Batching

If any one of the batched operations fails to complete successfully or attempts to return a result set during an executeBatch call, then the processing stops and a java.sql.BatchUpdateException is generated.

After a batch exception, the update counts array can be retrieved using the getUpdateCounts method of the BatchUpdateException object. This returns an int array of update counts, just as the executeBatch method does. In the Oracle implementation of standard update batching, contents of the update counts array are as follows, after a batch is processed:

  • For a prepared statement batch, in case of an error in between execution of the batch, the executeBatch method cannot return a value, instead it throws a BatchUpdateException. In this case, the exception itself carries an int array of size n as its data, where n is the number of successful record executions. For example, if the batch is of size 5 and the error occurs at the 4th record, then the BatchUpdateException has an array of size 3 (3 records executed successfully) and each item in the array represents how many rows were affected by each of them.

  • For a generic statement batch or callable statement batch, the update counts array is only a partial array containing the actual update counts up to the point of the error. The actual update counts can be provided because Oracle JDBC cannot use true batching for generic and callable statements in the Oracle implementation of standard update batching.

    For example, if there were 20 operations in the batch, the first 13 succeeded, and the 14th generated an exception, then the update counts array will have 13 elements, containing actual update counts of the successful operations.

    You can either commit or roll back the successful operations in this situation, as you prefer.

In your code, upon failed processing of a batch, you should be prepared to handle either -3 or true update counts in the array elements when an exception occurs. For a failed batch processing, you will have either a full array of -3 or a partial array of positive integers.

Intermixing Batched Statements and Nonbatched Statements

You cannot call executeUpdate for regular, nonbatched processing of an operation if the statement object has a pending batch of operations.

However, you can intermix batched operations and nonbatched operations in a single statement object if you process nonbatched operations either prior to adding any operations to the statement batch or after processing the batch. Essentially, you can call executeUpdate for a statement object only when its update batch is empty. If the batch is non-empty, then an exception will be generated.

For example, it is valid to have a sequence, such as the following:

...
PreparedStatement pstmt = 
          conn.prepareStatement("INSERT INTO employees VALUES(?, ?)");

pstmt.setInt(1, 2000);
pstmt.setString(2, "Milo Mumford");

int scount = pstmt.executeUpdate();   // OK; no operations in pstmt batch

pstmt.setInt(1, 3000);
pstmt.setString(2, "Sulu Simpson");
pstmt.addBatch();                    // Now start a batch

pstmt.setInt(1, 4000);
pstmt.setString(2, "Stan Leland");
pstmt.addBatch();

int[] bcounts = pstmt.executeBatch();

pstmt.setInt(1, 5000);
pstmt.setString(2, "Amy Feiner");

int scount = pstmt.executeUpdate();   // OK; pstmt batch was executed
...

Intermixing nonbatched operations on one statement object and batched operations on another statement object within your code is permissible. Different statement objects are independent of each other with regard to update batching operations. A COMMIT request will affect all nonbatched operations and all successful operations in processed batches, but will not affect any pending batches.

Premature Batch Flush

Premature batch flush happens due to a change in cached metadata. Cached metadata can be changed due to various reasons, such as the following:

  • The initial bind was null and the following bind is not null.

  • A scalar type is initially bound as string and then bound as scalar type or the reverse.

The premature batch flush count is summed to the return value of the next executeUpdate or sendBatch method.

The old functionality lost all these batch flush values which can be obtained now. To switch back to the old functionality, you can set the AccumulateBatchResult property to false, as follows:

java.util.Properties info = new java.util.Properties(); 
info.setProperty("user", "HR"); 
info.setProperty("passwd", "hr"); 
// other properties 
... 

// property: batch flush type 
info.setProperty("AccumulateBatchResult", "false");

OracleDataSource ods = new OracleDataSource();
ods.setConnectionProperties(info);
ods.setURL("jdbc:oracle:oci:@"");
Connection conn = ods.getConnection(); 

Note:

The AccumulateBatchResult property is set to true by default.

Example 21-3 illustrates premature batch flushing.

Example 21-3 Premature Batch Flushing

((OraclePreparedStatement)pstmt).setExecuteBatch (2); 

pstmt.setNull(1, OracleTypes.NUMBER); 
pstmt.setString(2, "test11"); 
int count = pstmt.executeUpdate(); // returns 0 

/* 
* Premature batch flush happens here. 
*/ 
pstmt.setInt(1, 22);
pstmt.setString(2, "test22"); 
int count = pstmt.executeUpdate(); // returns 0 

pstmt.setInt(1, 33);
pstmt.setString(2, "test33");
/* 
*  returns 3 with the new batching scheme where as, 
*  returns 2 with the old batching scheme. 
*/ 
int count = pstmt.executeUpdate();

Additional Oracle Performance Extensions

In addition to update batching, Oracle JDBC drivers support the following extensions that improve performance by reducing round-trips to the database:

  • Prefetching rows

    This reduces round-trips to the database by fetching multiple rows of data each time data is fetched. The extra data is stored in client-side buffers for later access by the client. The number of rows to prefetch can be set as desired.

  • Specifying column types

    This avoids an inefficiency in the standard JDBC protocol for performing and returning the results of queries.

  • Suppressing database metadata TABLE_REMARKS columns

    This avoids an expensive outer join operation.

Oracle provides several extensions to connection properties objects to support these performance extensions. These extensions enable you to set the remarksReporting flag and default values for row prefetching and update batching.

This section covers the following topics:

Prefetching LOB Data

For the JDBC drivers prior to Oracle Database 11g Release 2 JDBC drivers, if you want to retrieve LOB data in one round trip, then you have to fetch the data as VARCHAR2 type, that is, you have to use OracleTypes.VARCHAR or OracleTypes.LONGVARCHAR with the JDBC defineColumnType method. The limitation of this approach is that when LOB data is fetched as CHAR type, the locator cannot be fetched along with the data. So, if the application wants to get the LOB data at a later point of time, or if the application wants to perform other LOB operations, then one more round trip is required to get the LOB locator, as LOB locator is not available to the application.

Note:

Array operations on LOB locators are not supported in the JDBC APIs.

Starting from Oracle Database 11g Release 2 JDBC drivers, the number of round trips is reduced by prefetching frequently used metadata, such as the LOB length and the chunk size as well as the beginning of the LOB data along with the locator during regular fetch operations. For small LOBs, the data may be totally prefetched in one single round trip, that is, the select parse, execution, and fetch occurs in one round trip, and performance is improved greatly. For large LOBs that are larger than 5 times the prefetch size, the performance improvement is not very significant as only the round trip for retrieving the chunk size is not needed.

defaultLobPrefetchSize Connection Property

Starting from Oracle Database 11g Release 2, there is a new connection property oracle.jdbc.defaultLobPrefetchSize that can be used to set the default LOB prefetch size for the connection. This connection property is defined as the following constant: OracleConnection.CONNECTION_PROPERTY_DEFAULT_LOB_PREFETCH_SIZE. The value of this property is used as the default LOB prefetch size for the current connection. The default value of this connection property is 4000. If you want to change the default value at the statement level, then use the setLobPrefetchSize method defined in oracle.jdbc.OracleStatement interface. You can change the default value to:

  • -1 to disable LOB prefetch for the current connection

  • 0 to enable LOB prefetch for metadata only

  • Any value greater than 0 to specify the number of bytes for BLOBs and the number of characters for CLOBs to be prefetched along with the locator during fetch operations

Use getLobPrefetchSize method defined in oracle.jdbc.OracleStatement interface to retrieve the LOB prefetch size.

You can also set the value of LOB prefetch size at the column level by using the defineColumnType method. The column-level value overrides any value that is set at the connection or statement level.

See Also:

The Javadoc for more information

Note:

If LOB prefetch is not disabled at the connection level or statement level, it cannot be disabled at the column level.

Oracle Row-Prefetching Limitations

There is no maximum prefetch setting. The default value is 10. Larger or smaller values may be appropriate depending on the number of rows and columns expected from the query. You can set the default connection row-prefetch value using a Properties object.

When a statement object is created, it receives the default row-prefetch setting from the associated connection. Subsequent changes to the default connection row-prefetch setting will have no effect on the statement row-prefetch setting.

If a column of a result set is of data type LONG, LONG RAW or LOBs returned through the data interface, that is, the streaming types, then JDBC changes the statement row-prefetch setting to 1, even if you never actually read a value of either of these types.

Setting the prefetch size can affect the performance of an application. Increasing the prefetch size will reduce the number of round-trips required to get all the data, but will increase memory usage. This will depend on the number and size of the columns in the query and the number of rows expected to be returned. It will also depend on the memory and CPU loading of the JDBC client machine. The optimum for a standalone client application will be different from a heavily loaded application server. The speed and latency of the network connection should also be considered.

Note:

Starting from Oracle Database 11g Release 1, the Thin driver can fetch the first prefetch_size number of rows from the server in the very first round-trip. This saves one round-trip in SELECT statements.

If you are migrating an application from earlier releases of Oracle JDBC drivers to 10g Release 1 (10.1) or later releases of Oracle JDBC drivers, then you should revisit the optimizations that you had done earlier, because the memory usage and performance characteristics may have changed substantially.

A common situation that you may encounter is, say, you have a query that selects a unique key. The query will return only zero or one row. Setting the prefetch size to 1 will decrease memory and CPU usage and cannot increase round-trips. However, you must be careful to avoid the error of requesting an extra fetch by writing while(rs.next()) instead of if(rs.next()).

If you are using the JDBC Thin driver, then use the useFetchSizeWithLongColumn connection property, because it will perform PARSE, EXECUTE, and FETCH in a single round-trip.

Tuning of the prefetch size should be done along with tuning of memory management in your JVM under realistic loads of the actual application.

Note:

  • Do not mix the JDBC 2.0 fetch size application programming interface (API) and the Oracle row-prefetching API in your application. You can use one or the other, but not both.

  • Be aware that setting the Oracle fetch size value can affect not only queries, but also explicitly refetching rows in a result set through the result set refreshRow method, which is relevant for scroll-sensitive/read-only, scroll-sensitive/updatable, and scroll-insensitive/updatable result sets, and the window size of a scroll-sensitive result set, affecting how often automatic refetches are performed. However, the Oracle fetch size value will be overridden by any setting of the fetch size.

Defining Column Types

Note:

Starting from Oracle Database 12c Release 1 (12.1), the defineColumnType method is deprecated. For more information, refer to "Deprecated Features".

The implementation of defineColumnType changed significantly since Oracle Database 10g. Previously, defineColumnType was used both as a performance optimization and to force data type conversion. In previous releases, all of the drivers benefited from calls to defineColumnType. Starting from Oracle Database 10g, the JDBC Thin driver no longer needs the information provided. The JDBC Thin driver achieves maximum performance without calls to defineColumnType. The JDBC Oracle Call Interface (OCI) and server-side internal drivers still get better performance when the application uses defineColumnType.

If your code is used with both the JDBC Thin and OCI drivers, you can disable the defineColumnType method when using the Thin driver by setting the connection property disableDefineColumnType to true. Doing this makes defineColumnType have no effect. Do not set this connection property to true when using the JDBC OCI or server-side internal drivers.

You can also use defineColumnType to control how much memory the client-side allocates or to limit the size of variable-length data.

Follow these general steps to define column types for a query:

  1. If necessary, cast your statement object to OracleStatement, OraclePreparedStatement, or OracleCallableStatement, as applicable.

  2. If necessary, use the clearDefines method of your Statement object to clear any previous column definitions for this Statement object.

  3. On each column, call the defineColumnType method of your Statement object, passing it these parameters:

    • Column index (integer)

    • Type code (integer)

      Use the static constants of the java.sql.Types class or oracle.jdbc.OracleTypes class, such as Types.INTEGER, Types.FLOAT, Types.VARCHAR, OracleTypes.VARCHAR, and OracleTypes.ROWID. Type codes for standard types are identical in these two classes.

    • Type name (string)

      For structured objects, object references, and arrays, you must also specify the type name. For example, Employee, EmployeeRef, or EmployeeArray.

    • Maximum field size (integer)

      Optionally specify a maximum data length for this column.

      You cannot specify a maximum field size parameter if you are defining the column type for a structured object, object reference, or array. If you try to include this parameter, it will be ignored.

    • Form of use (short)

      Optionally specify a form of use for the column. This can be OraclePreparedStatement.FORM_CHAR to use the database character set or OraclePreparedStatement.FORM_NCHAR to use the national character set. If this parameter is omitted, the default is FORM_CHAR.

    For example, assuming stmt is an Oracle statement, use:

    stmt.defineColumnType(column_index, typeCode);
    

    If the column is VARCHAR or equivalent and you know the length limit:

    stmt.defineColumnType(column_index, typeCode, max_size);
    

    For an NVARCHAR column where the original maximum length is desired and conversion to the database character set is requested:

    stmt.defineColumnType(column_index, typeCode, 0,
       OraclePreparedStatement.FORM_CHAR );
    

    For structured object, object reference, and array columns:

    stmt.defineColumnType(column_index, typeCode, typeName);
    

    Set a maximum field size if you do not want to receive the full default length of the data. Calling the setMaxFieldSize method of the standard JDBC Statement class sets a restriction on the amount of data returned. Specifically, the size of the data returned will be the minimum of the following:

    • The maximum field size set in defineColumnType

    • The maximum field size set in setMaxFieldSize

    • The natural maximum size of the data type

After you complete these steps, use the executeQuery method of the statement to perform the query.

Note:

It is no longer necessary to specify a data type for each column of the expected result set.

Example 21-4 illustrates the use of this feature. It assumes you have imported the oracle.jdbc.* interfaces.

Example 21-4 Defining Column Types

OracleDataSource ods = new OracleDataSource();
ods.setURL("jdbc:oracle:thin:@localhost:5221:orcl");
ods.setUser("HR");
ods.setPassword("hr");
Connection conn = ods.getConnection();

Statement stmt = conn.createStatement();
// Allocate only 2 chars for this column (truncation will happen)
((OracleStatement)stmt).defineColumnType(1, Types.VARCHAR, 2);
ResultSet rset = stmt.executeQuery("select ename from emp");
while(rset.next() )
  System.out.println(rset.getString(1));
stmt.close();

As this example shows, you must cast the Statement object, stmt, to OracleStatement in the invocation of the defineColumnType method. The createStatement method of the connection returns an object of type java.sql.Statement, which does not have the defineColumnType and clearDefines methods. These methods are provided only in the OracleStatement implementation.

The define-extensions use JDBC types to specify the desired types. The allowed define types for columns depend on the internal Oracle type of the column.

All columns can be defined to their natural JDBC types. In most cases, they can be defined to the Types.CHAR or Types.VARCHAR type code.

Table 21-1 lists the valid column definition arguments you can use in the defineColumnType method.

Table 21-1 Valid Column Type Specifications

If the column has Oracle
SQL type:
You can use defineColumnType
to define it as:

NUMBER, VARNUM

BIGINT, TINYINT, SMALLINT, INTEGER, FLOAT, REAL, DOUBLE, NUMERIC, DECIMAL, CHAR, VARCHAR

CHAR, VARCHAR2

CHAR, VARCHAR

LONG

CHAR, VARCHAR, LONGVARCHAR

LONGRAW

LONGVARBINARY, VARBINARY, BINARY

RAW

VARBINARY, BINARY

DATE

DATE, TIME, TIMESTAMP, CHAR, VARCHAR

ROWID

ROWID

BLOB

VARBINARY, BINARY

CLOB

LONG, CHAR, VARCHAR


It is always valid to use defineColumnType with the original data type of the column.

Reporting DatabaseMetaData TABLE_REMARKS

The getColumns, getProcedureColumns, getProcedures, and getTables methods of the database metadata classes are slow if they must report TABLE_REMARKS columns, because this necessitates an expensive outer join. For this reason, the JDBC driver does not report TABLE_REMARKS columns by default.

You can enable TABLE_REMARKS reporting by passing a true argument to the setRemarksReporting method of an OracleConnection object.

Equivalently, instead of calling setRemarksReporting, you can set the remarksReporting Java property if you use a Java Properties object in establishing the connection.

If you are using a standard java.sql.Connection object, you must cast it to OracleConnection to use setRemarksReporting.

Example 21-5 illustrates how to enable TABLE_REMARKS reporting.

Example 21-5 TABLE_REMARKS Reporting

Assuming conn is the name of your standard Connection object, the following statement enables TABLE_REMARKS reporting:

((oracle.jdbc.OracleConnection)conn ).setRemarksReporting(true);

Considerations for getColumns

By default, the getColumns method does not retrieve information about the columns if a synonym is specified. To enable the retrieval of information if a synonym is specified, you must call the setIncludeSynonyms method on the connection as follows:

((oracle.jdbc.OracleConnection)conn ).setIncludeSynonyms(true)

This will cause all subsequent getColumns method calls on the connection to include synonyms. This is similar to setRemarksReporting. Alternatively, you can set the includeSynonyms connection property. This is similar to the remarksReporting connection property.

However, bear in mind that if includeSynonyms is true, then the name of the object returned in the table_name column will be the synonym name, if a synonym exists. This is true even if you pass the table name to getColumns.

Considerations for getProcedures and getProcedureColumns Methods

According to JDBC versions 1.1 and 1.2, the methods getProcedures and getProcedureColumns treat the catalog, schemaPattern, columnNamePattern, and procedureNamePattern parameters in the same way. In the Oracle definition of these methods, the parameters are treated differently:

  • catalog

    Oracle does not have multiple catalogs, but it does have packages. Consequently, the catalog parameter is treated as the package name. This applies both on input, which is the catalog parameter, and the output, which is the catalog column in the returned ResultSet. On input, the construct " ", which is an empty string, retrieves procedures and arguments without a package, that is, standalone objects. A null value means to drop from the selection criteria, that is, return information about both standalone and packaged objects. That is, it has the same effect as passing in the percent sign (%). Otherwise, the catalog parameter should be a package name pattern, with SQL wild cards, if desired.

  • schemaPattern

    All objects within Oracle database must have a schema, so it does not make sense to return information for those objects without one. Thus, the construct " ", which is an empty string, is interpreted on input to mean the objects in the current schema, that is, the one to which you are currently connected. To be consistent with the behavior of the catalog parameter, null is interpreted to drop the schema from the selection criteria. That is, it has the same effect as passing in %. It can also be used as a pattern with SQL wild cards.

  • procedureNamePattern and columnNamePattern

    The empty string (" ") does not make sense for either parameter, because all procedures and arguments must have names. Thus, the construct " " will raise an exception. To be consistent with the behavior of other parameters, null has the same effect as passing in percent sign (%).