Spec-Zone .ru спецификации, руководства, описания, API

RowSet

10.1    RowSet Overview

A RowSet object differs from a ResultSet object in that it is a JavaBeans^TM component. Thus, it has a set of JavaBeans properties and follows the JavaBeans event model. A RowSet object's properties allow it to establish its own database connection and to execute its own query in order to fill itself with data. A rowset may be disconnected, that is, function without maintaining an open connection to a data source the whole time it is in use. In addition, a rowset can be serialized, which means that it can be sent to a remote object over a network.

In general, the JDBC API can be divided into two categories, the RowSet portion and the driver portion. RowSet and its supporting interfaces are intended to be implemented using the rest of the JDBC API. In other words, a class that implements the RowSet interface is a layer of software that is said to execute "on top" of a JDBC driver. Unlike other JDBC objects, a RowSet object contains within itself the means to operate without a driver and without being connected to a data source.

The release of J2SE 5.0 introduced a third category. In addition to the RowSet API and the driver, there are five standard implementations of the RowSet interface. These implementations provide a set of interfaces that extend the basic RowSet interface plus reference implementations for each of them. There is no requirement to use these implementations, but by using them, developers can be sure that their implementations follow the JDBC API in event handling, cursor manipulation, and other operations. The standard implementations are discussed more fully in "Standard Implementations," on page 808.

The RowSet interface provides a basic set of methods common to all rowsets, which this section describes. All RowSet objects are JavaBeans components; therefore, the RowSet interface has methods for adding and removing an event listener, and it has getter amd setter methods for all of its properties. Many of these properties support setting up a connection or executing a command. A rowset uses a connection with a data source in order to execute a query and produce a result set from which it will get its data. It may also use a connection to write modified data back to the data source. In addition, the RowSet interface has methods (one for each data type) for setting the values of input parameters, if any, in a RowSet object's command string. In the JDBC RowSet Implementations specification, these basic methods, defined in the RowSet interface, are provided in the BaseRowSet abstract class, which is discussed later.

Five other interfaces and one class work together with the RowSet interface behind the scenes. The class RowSetEvent and the interface RowSetListener support the JavaBeans event model. When a RowSet object's cursor moves or its data is modified, it will invoke the RowSetListener method corresponding to the event, providing it with a RowSetEvent object that identifies itself as the source of the event. Implementations are free to write extensions that add other RowSet events if they are needed.

A component that wants to be notified of the events that occur in a RowSet object will implement the RowSetListener interface and be registered with the RowSet object. Such a component, called a listener, is typically a GUI (graphical user interface) component, such as a table or bar chart, that is displaying the RowSet object's data. Because a listener is notified every time an event occurs in the rowset, it can keep its cursor position and data consistent with that of the rowset.

The interfaces RowSetInternal, RowSetReader, and RowSetWriter support the rowset reader/writer facility. A reader, an instance of a class that implements the RowSetReader interface, reads data and inserts it into a rowset. A writer, an instance of a class that implements the RowSetWriter interface, writes modified data back to the data source from which a rowset's data was retrieved.

The RowSetInternal interface provides additional methods for a reader or writer to use to manipulate the rowset's internal state. For example, a rowset can keep track of its original values, and RowSetInternal methods allow the writer to see if the corresponding data in the data source has been changed by someone else. In addition, RowSetInternal methods make it possible to retrieve the input parameters that were set for a rowset's command string and to retrieve the connection that was passed to it, if there is one. Finally, RowSetInternal methods allow a reader to set a new RowSetMetaData object, which describes to the rowset the rows that the reader will insert into it. The name of the interface is RowSetInternal for good reason. Its methods are used internally; an application does not call these methods directly.

Rowsets may be either connected or disconnected. A connected RowSet object maintains a connection to its data source the entire time it is in use, whereas a disconnected rowset is connected to its data source only while it is reading data from the data source or writing data to it. While the rowset is disconnected, it does not need a JDBC driver or the full implementation of the JDBC API. This makes it very lean and therefore an ideal container for sending a set of data to a thin client. The client can, if it chooses, make updates to the data and send the rowset back to the application server. On the server, the disconnected RowSet object uses its reader to make a connection to the data source and write data back to it. Exactly how this is done depends on how the reader is implemented. Typically, the reader delegates making a connection and reading data to the JDBC driver.

10.1.1     The Event Model for Rowsets

Three kinds of events can occur in a RowSet object: its cursor can move, one of its rows can change (be inserted, deleted, or updated), or its entire contents can be changed. The RowSetListener methods cursorMoved, rowChanged, and rowSetChanged correspond to these events. When an event occurs, the rowset will create a RowSetEvent object that identifies itself as the source of the event. The appropriate RowSetListener method will be invoked on each listener, with the RowSetEvent object being passed to the method. This will inform all of the rowset's listeners about the event.

For example, if a pie chart component, pieChart, wants to display the data in the RowSet object rset, pieChart must implement the RowSetListener methods cursorMoved, rowChanged, and rowSetChanged. The implementations of these methods specify what pieChart will do in response to an event on rset. After implementing these methods, pieChart can be registered with rset. When pieChart is added as a listener to rset, it will be notified when an event occurs on rset by having the appropriate method invoked with a RowSetEvent object as its parameter. The listener pieChart can then update itself to reflect the current data and cursor position of rset. If pieChart does not need to reflect one of the events on rset, it can implement the RowSetListener method for that event so that it does nothing. For instance, if pieChart does not need to show the current cursor position in rset, it can have the method cursorMoved do nothing.

Any number of components may be listeners for a given RowSet object. If, for example, the bar graph component barGraph is also displaying the data in rset, it can become a listener by implementing the RowSetListener methods and then being registered with rset. The following lines of code register the two components pieChart and barGraph as listeners with rset.

    rset.addRowSetListener(pieChart);
    rset.addRowSetListener(barGraph);

    rset.removeRowSetListener(pieChart);
    rset.removeRowSetListener(barGraph);

In the following code fragment, pieChart and barGraph are registered as listeners with the RowSet object rset. After rset fills itself with new data by calling the method execute, event notification takes place behind the scenes. (The method execute will be explained in detail in the section "Executing a Command ," on page 805.) As the first step in the event notification process, the RowSetEvent object rsetEvent is created and initialized with rset. Next the pieChart and barGraph versions of the method rowSetChanged are called with rsetEvent as their arguments. This tells pieChart and barGraph that all the data in rset has changed, and each listener will carry out its own implementation of the method rowSetChanged.

rset.addRowSetListener(pieChart);
rset.addRowSetListener(barGraph);
. . .
rset.execute();
  
// The following methods will be invoked behind the scenes:
RowSetEvent rsetEvent = new RowSetEvent(this);
pieChart.rowSetChanged(rsetEvent);
barGraph.rowSetChanged(rsetEvent);

10.1.2     Properties for a Rowset

The following list gives the getter and setter methods defined on the RowSet interface for retrieving and setting a RowSet object's properties. The two exceptions are the methods getConcurrency and getType, which are inherited from the ResultSet interface rather than being defined in the RowSet interface.

getCommand	setCommand
ResultSet.getConcurrency	setConcurrency
getDataSourceName	setDataSourceName
getEscapeProcessing	setEscapeProcessing
getMaxFieldSize	setMaxFieldSize
getMaxRows	setMaxRows
getPassword	setPassword
getQueryTimeout	setQueryTimeout
getTransactionIsolation	setTransactionIsolation
ResultSet.getType	setType
getTypeMap	setTypeMap
getUrl	setUrl
getUsername	setUsername

In addition, the standard JDBC RowSet implementations may have various properties that are specific to them.

The following code fragment, in which rset is a RowSet object, sets properties that are typically required for establishing a connection with a data source using a DataSource object.

    rset.setDataSourceName("jdbc/logicalDataSourceName");
    rset.setUsername("cervantes");
    rset.setPassword("secret");

A RowSet object also has methods for setting properties that affect command execution. For example, as a sampling of these methods, the following code fragment sets twenty seconds as the longest a driver will wait for a statement to execute, sets 1024 as the largest number of rows rset may contain, and specifies that rset will be allowed to read only data from committed transactions.

    rset.setQueryTimeout(20);
    rset.setMaxRows(1024);
    rset.setTransactionIsolation(Connection.TRANSACTION_READ_COMMITTED);

    rset.setType(ResultSet.TYPE_SCROLL_INSENSITIVE);
    rset.setConcurrency(ResultSet.CONCUR_UPDATABLE);

Note that a RowSet object may be scrollable even if it uses a driver that does not support scrollable result sets. In fact, a rowset may often be used in place of a regular result set as a way of getting a scrollable result set. This is discussed in more detail in the section "Traversing a RowSet Object," on page 804.

An application fills a RowSet instance with data by executing the RowSet object's command string. This string must be a query that, when executed, will produce a result set from which the RowSet object will get its data. The method setCommand sets the String object supplied to it as the command that will be executed when the method execute is invoked. For example, the following line of code sets the command string to a query that selects the name and salary from every row in the table EMPLOYEES.

    rset.setCommand("SELECT NAME, SALARY FROM EMPLOYEES");

10.1.3     Setting Parameters for the Command String

A newly created RowSet object exists but has no data until it is populated with a call to the method execute or populate. In order to use the method execute, the necessary properties must be set, and values for any placeholder parameters must be set. Parameter values can be set at run time, which, for example, allows an application to be interactive and accept user input.

The RowSet interface, like the PreparedStatement interface, has setter methods for setting the value of input parameters. There is a setter method for each data type, including the SQL99 data types. The following code fragment sets the command string and then sets its two input parameters with values. Assuming that the column DEPT stores values of type VARCHAR, the method setString is used to set both parameters because that is the appropriate method for setting VARCHAR values.

    rset.setCommand(
      "SELECT NAME, SALARY FROM EMPLOYEES WHERE DEPT = ? OR DEPT = ?");
    rset.setString(1, "SALES");
    rset.setString(2, "MARKETING");

Any parameters in a RowSet object's command string must be set with values before the command is executed. When a rowset is disconnected, the parameters that have been set are used by the reader's readData method, which is invoked internally by the method execute. A rowset stores these parameter values in an internal hashtable, and the readData method retrieves them with a call to the rowset's RowSetInternal.getParams method.

10.1.4     Traversing a RowSet Object

Even though the methods for moving a RowSet object's cursor are identical to those of a ResultSet object from the user's point of view, a RowSet object's implementation of these methods is different. A RowSet object needs to let the listeners registered with it know about each movement of its cursor. Consequently, the cursor movement methods for a RowSet object are implemented to trigger the internal event notification process. For example, when the method next is called, its implementation will create a RowSetEvent object and call each listener's cursorMoved method, supplying the RowSetEvent object as the parameter. The listener uses the RowSetEvent object to find out in which RowSet object the cursor has moved and then invokes its implementation of the method cursorMoved. The implementation could do nothing, or it could, for example, call the method ResultSet.getRow to get the cursor's current position and update the listener's display of that row's data.

To demonstrate that cursor movements are the same in RowSet and ResultSet, the following code fragment iterates forward through the RowSet object rset and prints out the two values retrieved from each row.

    rset.beforeFirst();
    while (rset.next()) {
      System.out.println(rset.getString(1) + " " + rset.getFloat(2));
    }

10.1.5     Executing a Command

A disconnected rowset needs a reader (an object that implements the RowSetReader interface) and a writer (an object that implements the RowSetWriter interface) . In the JDBC RowSet Implementations, a reader and writer are encapsulated in the the SyncProvider class, which is part of the javax.sql.rowset.spi package. This package, known as the SPI (Service Provider Interface), is explained later. The SPI makes it much easier to implement and deploy a reader and writer. The two reference implementations, RIOptimisticProvider and RIXmlProvider, provide immediate implementations.

A disconnected rowset must also implement the RowSetInternal interface to make additional access to its internal state available to the reader and writer. With its reader/writer framework in place, a rowset's execute method is able to delegate tasks to its reader and writer components.

With the addition of the standard RowSet implementations, developers will find it much easier to implement the reader/writer facilities. They can leverage the reader/writer facilities that are already included in these standard implementations by simply incorporating them into their own implementations.

In a typical implementation, a disconnected rowset's execute method will invoke the reader's readData method to accomplish the job of reading new data into the rowset. Generally, after clearing the rowset of its current contents, the readData method will get the properties it needs and establish a connection with the data source. If there are any parameters to be set, readData retrieves them from the rowset and sets them appropriately in the rowset's command string. Then the readData method executes the command string to produce a result set. Finally, readData populates the rowset with the data from the result set and sets that data as the original values.

The reader's readData method may also be implemented to set the rowset's metadata. One of the many possible implementations is to have the readData method create a RowSetMetaData object and set it with information about the columns in the data source that is about to be read. The readData method next sets the new RowSetMetaData object to be the one associated with the rowset. The rowset can then use the RowSetMetaData object to see the format for the data that will be read into it.

When the rowset's command string is executed, all of a rowset's listeners need to be notified so that they can take the appropriate action. The execute method will invoke the rowSetChanged method on each listener, supplying rowSetChanged with a newly created RowSetEvent object that identifies the RowSet object in which the event occurred. As is true of most of what the method execute does, the notification of listeners is invisible to the application programmer using a rowset.

One more task for the execute method is that of setting the "original" values maintained by the rowset. These are the values, returned by the method RowSetInternal.getOriginal, that existed immediately before the most recent update. A RowSet object's original values may be the values it got from the data source, but this is not necessarily true. The first time a RowSet object's data is synchronized with its data source, its original values will be the same as the values it originally got from the data source. However, if some of a RowSet object's values are modified a second time, the original values will be the values that existed after the first modifications, not those from the data source. The point of keeping track of a RowSet object's original values is to be able to check whether the corresponding values in the data source have been changed.

If a SyncProvider object is implemented to check for conflicts, it compares the original values with the ones in the data source. If the rowset's original values (pre-modification values) and the underlying data source's values are not the same, there will be a conflict if the RowSet object's modified values are written to the data source. If the values are the same, however, there is no conflict. Depending on the synchronization model being used, a SyncProvider object may or may not write the new rowset values to the data source when there is a conflict.

If the SyncProvider implementation does not write new values to the underlying data source, the execute method will reset the original values back to the values currently in the rowset. Then the next time execute is called, which changes all of the values in the rowset, the SyncProvider object can retrieve these values to compare with those in the underlying data source to see if there is a conflict.

In summary, when an application invokes the execute method, many operations take place behind the scenes. When the rowset is disconnected, the following take place: the contents of the rowset are replaced with new data, any listeners using the rowset's data are notified, the rowset's metadata is updated, and the rowset's original values are set to the current data values. For a connected rowset, the execute method generally just populates the rowset with new data and notifies the listeners of the event. There is no possibility of a conflict because changes made to a connected RowSet object are also made to the data source.

10.1.6     Using a RowSet Object's Metadata

    RowSetMetaData rsetmd = (RowSetMetaData)rset.getMetaData();
    int columnCount = rsetmd.getColumnCount();

The interface RowSetMetaData defines setter methods corresponding to each of the getter methods defined in ResultSetMetaData (except that there are no methods for setting the class name for a column or for setting whether the column is read-only, possibly writable, or definitely writable). The RowSetMetaData setter methods are called by a reader after it has read new data into a rowset and created a new RowSetMetaData object for describing the RowSet object's columns. The following code shows what a reader might do behind the scenes. It creates the new RowSetMetaData object rowsetmd for the RowSet object rowset, sets the information for the columns, and finally calls the RowSetInternal method setMetaData to set rowsetmd as the metadata for rowset.

    rowset.execute();
    // ... as part of its implementation, execute calls readData
    reader.readData((RowSetInternal)this);
    
    // ... as part of the implementation of readData, the reader for
    // the rowset would do something like the following to update the
    // metadata for the rowset
    RowSetMetaData rowsetmd = new ...; // create an instance of a class
    // that implements RowSetMetaData
    rowsetmd.setColumnCount(3);
    rowsetmd.setColumnType(1, Types.INTEGER);
    rowsetmd.setColumnType(2, Types.VARCHAR);
    rowsetmd.setColumnType(3, Types.BLOB);
    // ... set other column information
    rowset.setMetaData(rowsetmd);

10.2    Standard Implementations

The standard implementations consist of two parts, the interfaces and the reference implementations. The interfaces are in the javax.sql.rowset package; the implementations are in the com.sun.rowset package. They were developed with input from experts in the database field through the Java Community Process as JSR 114. The goal was to standardize key rowset functionality so that developers can leverage it in their own implementations. The standard RowSet implementations are:

• JdbcRowSet
  
  
• CachedRowSet
  
  
• WebRowSet
  
  
• FilteredRowSet
  
  
• JoinRowSet
  
  

10.2.1     Implementation Basics

• Properties--fields for storing property values and methods for getting and setting property values
  
  
• Event notification mechanism--methods for adding and deleting listeners and for notifying listeners of changes
  
  
• Setter methods--a complete set of methods for setting the placeholder parameters in a RowSet object's command property
  
  
• Default values--all RowSet implementations have the following default values:
  
  
  • Is scrollable and does not show changes made by others
    
    
  • Is updatable
    
    
  • Does not show rows that have been deleted
    
    
  • Has no time limit for how long a driver may take to execute the RowSet object's command
    
    
  • Has no limit for the number of bytes that columns of type BINARY, VARBINARY, LONGVARBINARY, CHAR, VARCHAR, or LONGVARCHAR may contain
    
    
  • Does not see uncommitted data (make "dirty" reads)
    
    
  • Has escape processing is turned on
    
    
  • Has its connection's type map is set to null
    
    
  • Has an empty Hashtable object for storing the values set for the placeholder parameters in the RowSet object's command
    
    

As stated previously, RowSet objects may be connected or disconnected. The RowSet implementations fall into the following categories:

• Connected (maintains a connection to the data source using a JDBC driver)
  
  
  • JdbcRowSet
    
    
• Disconnected (not connected to the data source except when reading data from it or writing data to it)
  
  
  • CachedRowSet
    
    
  • JoinRowSet
    
    
  • FilteredRowSet
    
    
  • WebRowSet
    
    

The CachedRowSet interface provides the methods that a disconnected RowSet object needs. In the standard implementations, a disconnected RowSet extends the BaseRowSet class and implements the CachedRowSet interface. In addition, the JoinRowSet implementation implements the JoinRowSet interface, the FilteredRowSet implementation implements the FilteredRowSet interface, and the WebRowSet implementation implements the WebRowSet interface.

10.2.2     Overview of the JdbcRowSet Implementation

A JdbcRowSet object has many uses. Probably the most common use is to make a ResultSet object scrollable and thereby make better use of legacy drivers that do not support scrolling. A JdbcRowSet object's rows of data (and those in any RowSet object) are identical to those in the ResultSet object that is the result of executing the rowset's command. Therefore, if the rowset is scrollable, it is the equivalent of having a scrollable ResultSet object even if the ResultSet object itself is not scrollable.

Another common use is to make the driver or a ResultSet object a JavaBeans component. Like all RowSet objects, a JdbcRowSet object is a JavaBeans component. By being continuously connected to a driver, it serves as a wrapper for the driver, which effectively makes the driver a JavaBeans component. This means that a driver presented as a JdbcRowSet object can be one of the Beans that a tool makes available for composing an application. Being continuously connected also means that a JdbcRowSet object is able to serve as a wrapper for its ResultSet object. It can take calls invoked on it and, in turn, call them on its ResultSet object. As a consequence, the ResultSet object can be, for example, a component in a GUI application that uses Swing technology.

The following code fragment illustrates creating a JdbcRowSet object, setting its properties, and executing the command string in its command property. The JdbcRowSet implementation provides a default constructor, but being a JavaBeans component, a JdbcRowSet implementation will probably most often be created by a visual JavaBeans development tool.

    JdbcRowSet jrs = new JdbcRowSetImpl();
    jrs.setCommand("SELECT * FROM TITLES);
    jrs.setURL("jdbc:myDriver:myAttribute");
    jrs.setUsername("cervantes");
    jrs.setPassword("sancho");
    jrs.execute();

From this point on, the code can simply use ResultSet methods because it is effectively operating on a ResultSet object. It can navigate the rows in jrs, retrieve column values, update column values, insert new rows, and so on. For example, the next two lines of code go to the second row and retrieve the value in the first column using ResultSet methods.

    jrs.absolute(2);
    String title = jrs.getString(1);

10.2.3     Overview of the CachedRowSet Implementation

A CachedRowSet object can populate itself with data from a tabular data source, and because it is updatable, it can also modify its data. As with all RowSet objects, in addition to getting data in, it can get data out, propagating its modifications back to the underlying data source.

10.2.4     Uses for a CachedRowSet Object

One of the major uses for a CachedRowSet instance is to pass tabular data between different components of a distributed application, such as EnterpriseJavaBeans ^TM (EJB^TM) components running in an application server. The server can use the JDBC API to retrieve a set of rows from a database and then use a CachedRowSet object to send the data over the network to, for example, a thin client running in a web browser.

Another use for a CachedRowSet object is to provide scrolling and updating capabilities to a ResultSet object that does not itself have this functionality. For example, as with a JdbcRowSet object, an application can create a CachedRowSet object initialized with the data from a ResultSet object and then operate on the rowset instead of the result set. With the rowset set to be scrollable and updatable, the application can move the cursor, make updates, and then propagate the updates back to the data source.

The following code fragment populates the CachedRowSet object crset with the data from the ResultSet object rs.

    ResultSet rs = stmt.executeQuery("SELECT * FROM AUTHORS");
    CachedRowSet crset = new CachedRowSetImpl();
    crset.populate(rs);

The CachedRowSet implementation uses the base API from the BaseRowSet abstract class and the CachedRowSet interface, which adds methods for the additional functionality it needs. For example, it needs a way to connect to a data source and read data from it. If a CachedRowSet object's data is modified while it is disconnected, it also needs a way to connect to the data source and write the new data back to it. The reader and writer facilities, based on the RowSetReader and RowSetWriter interfaces and encapsulated in a SyncProvider implementation, provide these capabilities.

A writer implementation supplies a mechanism for updating the data source with any changes it has made while it was disconnected, thus making the modified data persistent. This can get complicated if there is a conflict, that is, if another user has already modified the same data in the data source. The SyncProvider class includes mechanisms that afford varying degrees of control over how accessible data in the database is to others, and consequently, it has some control over the number of conflicts that may occur. Even more important, a SyncProvider implementation determines the level of care to be taken in synchronization.

The JDBC RowSet Implementations specification includes two different reference implementations for synchronization using two different levels of concurrency. The RIXmlProvider does not check for conflicts and simply writes its modified data to the data source. The writer defined in the RIOptimisticProvider implementation checks for conflicts, and if a conflict exists, allows an application to use a SyncResolver object to decide whether or not to write modified data on a case by case basis. Third parties can write SyncProvider implementations offering different degrees of synchronization, and a disconnected rowset may use the one that best fits its needs. The SyncProvider class and the SyncResolver interface are covered in the section "Using the javax.sql.rowset.spi Package".

The implementation of the CachedRowSet interface can be used as a basis for implementing other disconnected rowsets. For example, the implementations of the WebRowSet, FilteredRowSet, and JoinRowSet interfaces are all based on the implementation of the CachedRowSet interface. Therefore, the following discussion of the CachedRowSet interface goes into some detail about the methods it defines and what they do. The sections on the other disconnected rowset implementations discuss the capabilities they have beyond those of the CachedRowSet implementation.

10.2.5     Creating a CachedRowSet Object

    CachedRowSet crset = new CachedRowSetImpl();

    crset.setCommand("SELECT * FROM EMPLOYEES");
    crset.setEscapeProcessing(false);

10.2.6     Setting a SyncProvider Object

If the fully qualified class name of a SyncProvider implementation is supplied to the constructor, the SyncFactory will initialize the RowSet object with an instance of that SyncProvider implementation if it has been registered with the SyncFactory. The following line of code creates a CachedRowSet object whose synchronization provider is an instance of the given SyncProvider implementation.

    CachedRowSet crset = new CachedRowSetImpl("com.supersoftware.providers.HighAvailabilityProvider");

    crset.setSyncProvider("com.supersoftware.providers.HighAvailabilityProvider");

10.2.7     Populating a CachedRowSet Object

The version of execute inherited from the RowSet interface has to establish a connection behind the scenes before it can execute the rowset's command string. The version of execute defined in the CachedRowSet interface is passed a Connection object, so it does not need to establish its own connection. The method populate does not need to establish its own connection or execute a query because it is passed the ResultSet object from which it will get its data. Thus, the method populate does not require that the properties affecting a connection be set before it is called, nor does it require that the command string be set. The following code fragment shows how the method populate might be used.

    ResultSet rs = stmt.executeQuery("SELECT NAME, SALARY FROM EMPLOYEES");

    // The code that generates rs
    . . .
    CachedRowSet crset = new CachedRowSetImpl();
    crset.populate(rs);

The implementation details of the method populate may vary, of course, but because it is passed a ResultSet object, populate is in some ways quite different from the method execute. As stated previously, the populate method does not need to establish a connection. Nor does it use a reader because it can get its data directly from the given result set. By using ResultSetMetaData methods, it can get information about the format of the result set so that it is able to use the appropriate getter methods in the ResultSet interface to retrieve the ResultSet object's data.

The methods execute and populate do have some similarities. The main similarity is that both methods change the contents of the entire rowset. As a consequence, they both cause the following: notification of listeners, setting the original values equal to the current rowset values, and updating the rowset's metadata information.

The following code fragments give examples of using the two versions of the method execute. Both versions execute a query, so both require that a command string be set. When no connection has been passed to the method execute, the appropriate connection properties must be set before the command can be called. The execute method will call the rowset's reader, which will use the necessary properties to establish a connection with the data source. Once the connection is established, the reader can call the method executeQuery to execute the rowset's command string and get a result set from which to retrieve data.

    CachedRowSet crset1 = new CachedRowSetImpl();
    crset1.setCommand("SELECT NAME, SALARY FROM EMPLOYEES");
    crset1.setDataSourceName("jdbc/myDataSource");
    crset1.setUsername("paz");
    crset1.setPassword("p38c3");
    crset1.setTransactionIsolation(Connection.TRANSACTION_READ_COMMITTED);

    crset1.execute();

    CachedRowSet crset2 = new CachedRowSetImpl();
    crset2.setCommand("SELECT NAME, SALARY FROM EMPLOYEES");
    crset2.execute(con);

10.2.8     Accessing Data

    crset.last();
    String note = crset.getString(1);

10.2.9     Modifying Data

An updater method modifies the value in the specified column in the current row, assigning it the value passed to it (usually as the second parameter). An updater method changes only the value in the rowset, which is cached in memory; it does not affect the value in the underlying data source. Also, it does not affect the value that the rowset keeps track of as the original value.

When an application has made all its updates to a row, it calls the method updateRow. In a CachedRowSet implementation, this method signals that the updates for the current row are complete, but, like the updater methods, it does not affect the values in the underlying data source. The updateRow method likewise does not affect the values stored as original values. After calling the updateRow method on all the rows being updated, an application needs to call the CachedRowSet method acceptChanges. This method invokes a writer component internally to propagate changes to the data source backing the rowset, and for each column value that was changed, it sets the original value to the current value.

Once it has called updateRow, an application may no longer call the method cancelRowUpdates to undo updates to the current row. If it has not yet invoked the method acceptChanges, however, it can call the method restoreOriginal, which undoes the updates to all rows by replacing the updated values in the rowset with the original values. The restoreOriginal method does not need to interact with the underlying data source.

The following code fragment updates the first two rows in the CachedRowSet object crset.

    crset.execute();
    // crset is initialized with its original and current values
    
    crset.first();
    crset.updateString(1, "Jane_Austen");
    crset.updateFloat(2, 150000f);
    crset.updateRow();
    // the current value of the first row has been updated
    
    crset.relative(1);
    crset.updateString(1, "Toni_Morrison");
    crset.updateFloat(2, 120000f);
    crset.updateRow();
    // the current value of the second row has been updated
    
    crset.acceptChanges();
    // the original value has been set to the current value and the 
    // database has been updated

10.2.10     Customizing Readers and Writers

The CachedRowSet interface defines the method setSyncProvider, which provides a CachedRowSet object with its implementation of a reader and writer. It also defines a getSyncProvider method for retrieving a rowset's SyncProvider object. A rowset's reader and writer operate completely behind the scenes, performing any number of tasks that can be customized to provide additional functionality.

The RowSetReader interface has one public method, readData, which can be customized in various ways. For example, a reader can be implemented so that it reads data straight from a file or from some other non-SQL data source rather than from a database using a JDBC driver. Such a reader might use the method RowSet.insertRow to insert new rows into the rowset. When invoked by a reader, this method could also be implemented so that it updates the original values stored by the rowset.

The RowSetWriter interface has one public method, writeData, which writes data that has been modified back to the underlying data source. This method can likewise be customized in a variety of ways. The writer establishes a connection with the data source, just as the reader does, and depending on how it is implemented, may or may not check for conflicts. The RowSetInternal methods getOriginal and getOriginalRow supply the values that existed before the current modifications, so the writer can compare them with the values read from the data source to see if the data source has been modified. How the writer decides whether or not to write data when there is a conflict again depends on how it is implemented.

A CachedRowSet object that has been configured with a custom reader and/or writer can be made available as a normal JavaBeans component. This means that developers writing applications do not have to worry about customizing readers and writers and can concentrate on the more important aspects of using rowsets effectively.

10.2.11     Other Methods

10.2.12     Using the javax.sql.rowset.spi Package

• SyncFactory
  
  
• SyncFactoryException
  
  
• SyncProvider
  
  
• SyncProviderException
  
  
• SyncResolver
  
  
• XmlReader
  
  
• XmlWriter
  
  
• TransactionalWriter
  
  

• RowSetReader
  
  
• RowSetWriter
  
  

    SyncFactory.registerProvider("com.supersoftware.providers.HAProvider");

A SyncProvider implementation may also be registered by adding it to the system properties. This can be done at the command line at execution time as follows:

    -Drowset.provider.classname=com.supersoftware.providers.HAProvider

When a RowSet object requests a particular SyncProvider implementation, the SyncFactory will search for it and create an instance of it and return it to the RowSet object. For example, in the following line of code, the CachedRowSet object crset requests the com.supersoftware.providers.HAProvider.

    crset.setSyncProvider("com.supersoftware.providers.HAProvider");

A CachedRowSet object or any of its subclasses can discover which SyncProvider implementations have been registered with the SyncFactory and are thus available for its use by calling the following line of code.

    java.util.Enumeration providers = SyncFactory.getRegisteredProviders();

The TransactionalWriter interface allows a CachedRowSet object that is participating in a global or local transaction to get fine-grained control of the transactional boundaries. To get this functionality, a CachedRowSet object needs to use a SyncProvider object that implements the TransactionalWriter interface.

The reference implementation provides a means by which an application can choose to resolve conflicts on a case by case basis. After the writer finds all the conflicts, it throws a SyncProviderException object. An application can catch that exception and use it to create a SyncResolver object, a specialized kind of RowSet object. A SyncResolver object mirrors the RowSet object experiencing the conflicts, having the same number of rows and columns; however, it contains only the data that is in conflict.

The SyncResolver object retrieves each conflict value in a row, comparing the value in the data source with the value in the RowSet object. After a decision is made as to which value should persist for each conflict in the row, the SyncResolver object sets those values with the method setResolvedValue. The SyncResolver object then goes to the next conflict and repeats the process until there are no more conflicts.

10.2.13     CachedRowSet Summary

10.2.14     WebRowSet Implementation

The default constructor for a WebRowSet requests an RIXmlProvider implementation so that the WebRowSet object will be able to read and write a rowset as an XML document. It is also possible to set a WebRowSet object with a third-party SyncProvider implementation that provides XML capabilities.

A WebRowSet object is designed to work well in a distributed client/server application. It is similar to a CachedRowSet implementation in that both connect a thin client to an application server. Thus, they are both good for providing data to a thin client. The difference is that they use different protocols. Because a CachedRowSet object is serializable, it can be sent to another component using RMI/IIOP (Remote Method Invocation/Internet Interoperability Protocol). A WebRowSet object uses HTTP/XML (Hypertext Transfer Protocol/eXtensible Markup Language) to communicate with the middle tier, so that, for example, Web clients can talk to Java servlets that provide data access.

XML has become more and more important for Web services because of the portability of data it provides. The JDBC RowSet Implementations specification includes a standard WebRowSet XML Schema, available at http://java.sun.com/xml/ns/jdbc/webrowset.xsd, to which a standard WebRowSet object adheres. This means that if two parties have the XML schema for a WebRowSet object, they can use it to exchange rowsets in a common format even though they may store their data internally in entirely different formats. This makes a WebRowSet object a powerful tool for data exchange.

Because the Java platform provides portability of code and XML provides portability of data, they are the ideal combination for Web services. Technologies like Java^TM API for XML-based RPC, SOAP with Attachments API for Java^TM, Java^TM Architecture for XML Binding, and Java^TM API for XML Registries make developing Web services easier and easier. Plus the infrastructure that the J2EE platform provides saves developers from having to program their own "plumbing" for the management of distributed transactions, connection pooling, and security.

Being able to use a WebRowSet object for sending data adds even more to the value of using the Java platform for Web services. For example, because it uses the standard XML schema, a WebRowSet object can be part of a Web services message (generally using SOAP with Attachments API for Java technology).

The following code fragment is a simple example that creates the WebRowSet object wrs, populates it with the data of the ResultSet object rs, and then updates a column value. The final two lines output wrs in XML format.

    WebRowSet wrs = new WebRowSetImpl();
    wrs.populate(rs);
    
    //perform updates
    wrs.absolute(2)
    wrs.updateString(1, "newString");
    
    FileWriter fWriter = new FileWriter("/share/net/output.xml");
    wrs.writeXml(fWriter);

    WebRowSet wrs = new WebRowsetImpl();
    
    FileReader fReader = new FileReader("/share/net/output.xml");
    wrs.readXml(fReader);
    
    wrs.absolute(2);
    String str = wrs.getString(1);

10.2.15     FilteredRowSet Implementation

Suppose you have a FilteredRowSet object frs that contains all of the employees in Company XYZ. You have figured the average salary for the company, including everyone in the average. Now you want to get information about only the employees whose names range from Aaronson to Lee. The following code fragment causes frs to make available only the rows where the values in the column NAMES are in the range of Aaronson to Lee.

In the following code fragment, the FilteredRowSet object frs is populated with data from the ResultSet object rs. Then the code creates a Range object, specifying that the last names Aaronson through Lee, which are in the column NAME, make up the range of names that the method next can return.

    FilteredRowSet frs = new FilteredRowSetImpl();
    frs.populate(rs);
    
    Range names = new Range("Aaronson", "Lee", findColumn("NAME"));
    frs.setFilter(names);
    
    while (frs.next()) {
      String name = frs.getString("NAME");
      . . . // add each name to, for example, a mailing list
      // only names from "Aaronson" to "Lee" will be returned
    }

10.2.16     JoinRowSet Implementation

The following code fragment demonstrates how the data from two CachedRowSet objects is joined into one JoinRowSet object. In this scenario, data from the table EMPLOYEES is joined with data from the table BONUS_PLAN, thereby giving information about each employee and his or her bonus plan all in one rowset. In both of the original rowsets, the first column is the employee identification number, so that is the column used to match the data from one table with data from the other. After the join, each row in the JoinRowSet object will contain the columns from both rowsets that pertain to the same employee ID.

The first line of code creates the JoinRowSet object jrs. Next, all the columns from the table EMPLOYEES are used to populate the new CachedRowSet object empl, the first column of empl is set as the match column, and empl is added to jrs. Then, in similar fashion, all the columns from the table BONUS_PLAN are used to populate the CachedRowSet object bonus, the first column of bonus is set as the match column, and bonus is added to jrs. In both empl and bonus, the first column is EMPLOYEE_ID, which is the primary key. It is the column both rowsets have in common, and it can be used to match the information from empl about an employee with the information in bonus about the same employee. The last lines of code navigate jrs and retrieve a column value.

    JoinRowSet jrs = new JoinRowSetImpl();
    
    ResultSet rs1 = stmt.executeQuery("SELECT * FROM EMPLOYEES");
    CachedRowSet empl = new CachedRowSetImpl();
    empl.populate(rs1);
    empl.setMatchColumn(1); // The first column is EMPLOYEE_ID 
    jrs.addRowSet(empl);
    
    ResultSet rs2 = stmt.executeQuery("SELECT * FROM BONUS_PLAN");
    CachedRowSet bonus = new CachedRowSetImpl();
    bonus.populate(rs2);
    bonus.setMatchColumn(1); // The first column is EMPLOYEE_ID 
    jrs.addRowSet(bonus);
    
      // The jrs instance now joins the two rowsets. The application
      // can browse the combined data as if it were browsing one single
      // RowSet object.
    
    jrs.first();
    int employeeID = jrs.getInt(1);
    String employeeName = jrs.getString(2);