Using CachedRowSetObjects
Trail: JDBC(TM) Database Access
Lesson: JDBC Basics

Using CachedRowSetObjects

A CachedRowSet object is special in that it can operate without being connected to its data source, that is, it is a disconnected RowSet object. It gets its name from the fact that it stores (caches) its data in memory so that it can operate on its own data rather than on the data stored in a database.

The CachedRowSet interface is the superinterface for all disconnected RowSet objects, so everything demonstrated here also applies to WebRowSet, JoinRowSet, and FilteredRowSet objects.

Note that although the data source for a CachedRowSet object (and the RowSet objects derived from it) is almost always a relational database, a CachedRowSet object is capable of getting data from any data source that stores its data in a tabular format. For example, a flat file or spreadsheet could be the source of data. This is true when the RowSetReader object for a disconnected RowSet object is implemented to read data from such a data source. The reference implementation of the CachedRowSet interface has a RowSetReader object that reads data from a relational database, so in this tutorial, the data source is always a database.

The following topics are covered:

Setting Up CachedRowSet Objects

Setting up a CachedRowSet object involves the following:

Creating CachedRowSet Objects

You can create a new CachedRowSet object in the different ways:

Note: Alternatively, you can use the constructor from the CachedRowSet implementation of your JDBC driver. However, implementations of the RowSet interface will differ from the reference implementation. These implementations will have different names and constructors. For example, the Oracle JDBC driver's implementation of the CachedRowSet interface is named oracle.jdbc.rowset.OracleCachedRowSet.

Using the Default Constructor

One of the ways you can create a CachedRowSet object is by calling the default constructor defined in the reference implementation, as is done in the following line of code:

CachedRowSet crs = new CachedRowSetImpl();

The object crs has the same default values for its properties that a JdbcRowSet object has when it is first created. In addition, it has been assigned an instance of the default SyncProvider implementation, RIOptimisticProvider.

A SyncProvider object supplies a RowSetReader object (a reader) and a RowSetWriter object (a writer), which a disconnected RowSet object needs in order to read data from its data source or to write data back to its data source. What a reader and writer do is explained later in the sections What Reader Does and What Writer Does. One thing to keep in mind is that readers and writers work entirely in the background, so the explanation of how they work is for your information only. Having some background on readers and writers should help you understand what some of the methods defined in the CachedRowSet interface do in the background.

Setting CachedRowSet Properties

Generally, the default values for properties are fine as they are, but you may change the value of a property by calling the appropriate setter method. There are some properties without default values that you must set yourself.

In order to get data, a disconnected RowSet object must be able to connect to a data source and have some means of selecting the data it is to hold. The following properties hold information necessary to obtain a connection to a database.

Which of these properties you must set depends on how you are going to make a connection. The preferred way is to use a DataSource object, but it may not be practical for you to register a DataSource object with a JNDI naming service, which is generally done by a system administrator. Therefore, the code examples all use the DriverManager mechanism to obtain a connection, for which you use the url property and not the datasourceName property.

The following lines of code set the username, password, and url properties so that a connection can be obtained using the DriverManager class. (You will find the JDBC URL to set as the value for the url property in the documentation for your JDBC driver.)

public void setConnectionProperties(
    String username, String password) {
    // ...

Another property that you must set is the command property. In the reference implementation, data is read into a RowSet object from a ResultSet object. The query that produces that ResultSet object is the value for the command property. For example, the following line of code sets the command property with a query that produces a ResultSet object containing all the data in the table MERCH_INVENTORY:

crs.setCommand("select * from MERCH_INVENTORY");

Setting Key Columns

If you are going make any updates to the crs object and want those updates saved in the database, you must set one more piece of information: the key columns. Key columns are essentially the same as a primary key because they indicate one or more columns that uniquely identify a row. The difference is that a primary key is set on a table in the database, whereas key columns are set on a particular RowSet object. The following lines of code set the key columns for crs to the first column:

int [] keys = {1};

The first column in the table MERCH_INVENTORY is ITEM_ID. It can serve as the key column because every item identifier is different and therefore uniquely identifies one row and only one row in the table MERCH_INVENTORY. In addition, this column is specified as a primary key in the definition of the MERCH_INVENTORY table. The method setKeyColumns takes an array to allow for the fact that it may take two or more columns to identify a row uniquely.

As a point of interest, the method setKeyColumns does not set a value for a property. In this case, it sets the value for the field keyCols. Key columns are used internally, so after setting them, you do nothing more with them. You will see how and when key columns are used in the section Using SyncResolver Objects.

Populating CachedRowSet Objects

Populating a disconnected RowSet object involves more work than populating a connected RowSet object. Fortunately, the extra work is done in the background. After you have done the preliminary work to set up the CachedRowSet object crs, the following line of code populates crs:


The data in crs is the data in the ResultSet object produced by executing the query in the command property.

What is different is that the CachedRowSet implementation for the execute method does a lot more than the JdbcRowSet implementation. Or more correctly, the CachedRowSet object's reader, to which the method execute delegates its tasks, does a lot more.

Every disconnected RowSet object has a SyncProvider object assigned to it, and this SyncProvider object is what provides the RowSet object's reader (a RowSetReader object). When the crs object was created, it was used as the default CachedRowSetImpl constructor, which, in addition to setting default values for properties, assigns an instance of the RIOptimisticProvider implementation as the default SyncProvider object.

What Reader Does

When an application calls the method execute, a disconnected RowSet object's reader works behind the scenes to populate the RowSet object with data. A newly created CachedRowSet object is not connected to a data source and therefore must obtain a connection to that data source in order to get data from it. The reference implementation of the default SyncProvider object (RIOptimisticProvider) provides a reader that obtains a connection by using the values set for the user name, password, and either the JDBC URL or the data source name, whichever was set more recently. Then the reader executes the query set for the command. It reads the data in the ResultSet object produced by the query, populating the CachedRowSet object with that data. Finally, the reader closes the connection.

Updating CachedRowSet Object

In the Coffee Break scenario, the owner wants to streamline operations. The owner decides to have employees at the warehouse enter inventory directly into a PDA (personal digital assistant), thereby avoiding the error-prone process of having a second person do the data entry. A CachedRowSet object is ideal in this situation because it is lightweight, serializable, and can be updated without a connection to the data source.

The owner will have the application development team create a GUI tool for the PDA that warehouse employees will use for entering inventory data. Headquarters will create a CachedRowSet object populated with the table showing the current inventory and send it using the Internet to the PDAs. When a warehouse employee enters data using the GUI tool, the tool adds each entry to an array, which the CachedRowSet object will use to perform the updates in the background. Upon completion of the inventory, the PDAs send their new data back to headquarters, where the data is uploaded to the main server.

This section covers the following topics:

Updating Column Values

Updating data in a CachedRowSet object is just the same as updating data in a JdbcRowSet object. For example, the following code fragment from increments the value in the column QUAN by 1 in the row whose ITEM_ID column has an item identifier of 12345:

while ( {
        "Found item " + crs.getInt("ITEM_ID") +
        ": " + crs.getString("ITEM_NAME"));
    if (crs.getInt("ITEM_ID") == 1235) {
        int currentQuantity = crs.getInt("QUAN") + 1;
        System.out.println("Updating quantity to " +
        crs.updateInt("QUAN", currentQuantity + 1);
        // Synchronizing the row
        // back to the DB

Inserting and Deleting Rows

Just as with updating a column value, the code for inserting and deleting rows in a CachedRowSet object is the same as for a JdbcRowSet object.

The following excerpt from inserts a new row into the CachedRowSet object crs:

crs.updateInt("ITEM_ID", newItemId);
crs.updateString("ITEM_NAME", "TableCloth");
crs.updateInt("SUP_ID", 927);
crs.updateInt("QUAN", 14);
Calendar timeStamp;
timeStamp = new GregorianCalendar();
timeStamp.set(2006, 4, 1);
    new Timestamp(timeStamp.getTimeInMillis()));

If headquarters has decided to stop stocking a particular item, it would probably remove the row for that coffee itself. However, in the scenario, a warehouse employee using a PDA also has the capability of removing it. The following code fragment finds the row where the value in the ITEM_ID column is 12345 and deletes it from the CachedRowSet crs:

while ( {
    if (crs.getInt("ITEM_ID") == 12345) {

Updating Data Sources

There is a major difference between making changes to a JdbcRowSet object and making changes to a CachedRowSet object. Because a JdbcRowSet object is connected to its data source, the methods updateRow, insertRow, and deleteRow can update both the JdbcRowSet object and the data source. In the case of a disconnected RowSet object, however, these methods update the data stored in the CachedRowSet object's memory but cannot affect the data source. A disconnected RowSet object must call the method acceptChanges in order to save its changes to the data source. In the inventory scenario, back at headquarters, an application will call the method acceptChanges to update the database with the new values for the column QUAN.


What Writer Does

Like the method execute, the method acceptChanges does its work invisibly. Whereas the method execute delegates its work to the RowSet object's reader, the method acceptChanges delegates its tasks to the RowSet object's writer. In the background, the writer opens a connection to the database, updates the database with the changes made to the RowSet object, and then closes the connection.

Using Default Implementation

The difficulty is that a conflict can arise. A conflict is a situation in which another party has updated a value in the database that corresponds to a value that was updated in a RowSet object. Which value should persist in the database? What the writer does when there is a conflict depends on how it is implemented, and there are many possibilities. At one end of the spectrum, the writer does not even check for conflicts and just writes all changes to the database. This is the case with the RIXMLProvider implementation, which is used by a WebRowSet object. At the other end, the writer ensures that there are no conflicts by setting database locks that prevent others from making changes.

The writer for the crs object is the one provided by the default SyncProvider implementation, RIOptimisticProvider. The RIOPtimisticProvider implementation gets its name from the fact that it uses an optimistic concurrency model. This model assumes that there will be few, if any, conflicts and therefore sets no database locks. The writer checks to see if there are any conflicts, and if there is none, it writes the changes made to the crs object to the database, and those changes become persistent. If there are any conflicts, the default is not to write the new RowSet values to the database.

In the scenario, the default behavior works very well. Because no one at headquarters is likely to change the value in the QUAN column of COF_INVENTORY, there will be no conflicts. As a result, the values entered into the crs object at the warehouse will be written to the database and thus will be persistent, which is the desired outcome.

Using SyncResolver Objects

In other situations, however, it is possible for conflicts to exist. To accommodate these situations, the RIOPtimisticProvider implementation provides an option that lets you look at the values in conflict and decide which ones should be persistent. This option is the use of a SyncResolver object.

When the writer has finished looking for conflicts and has found one or more, it creates a SyncResolver object containing the database values that caused the conflicts. Next, the method acceptChanges throws a SyncProviderException object, which an application may catch and use to retrieve the SyncResolver object. The following lines of code retrieve the SyncResolver object resolver:

try {
} catch (SyncProviderException spe) {
    SyncResolver resolver = spe.getSyncResolver();

The object resolver is a RowSet object that replicates the crs object except that it contains only the values in the database that caused a conflict. All other column values are null.

With the resolver object, you can iterate through its rows to locate the values that are not null and are therefore values that caused a conflict. Then you can locate the value at the same position in the crs object and compare them. The following code fragment retrieves resolver and uses the SyncResolver method nextConflict to iterate through the rows that have conflicting values. The object resolver gets the status of each conflicting value, and if it is UPDATE_ROW_CONFLICT, meaning that the crs was attempting an update when the conflict occurred, the resolver object gets the row number of that value. Then the code moves the cursor for the crs object to the same row. Next, the code finds the column in that row of the resolver object that contains a conflicting value, which will be a value that is not null. After retrieving the value in that column from both the resolver and crs objects, you can compare the two and decide which one you want to become persistent. Finally, the code sets that value in both the crs object and the database using the method setResolvedValue, as shown in the following code:

try {
} catch (SyncProviderException spe) {
    SyncResolver resolver = spe.getSyncResolver();
    // value in crs
    Object crsValue;
    // value in the SyncResolver object
    Object resolverValue; 
    // value to be persistent
    Object resolvedValue; 

    while (resolver.nextConflict()) {
        if (resolver.getStatus() ==
            SyncResolver.UPDATE_ROW_CONFLICT) {
            int row = resolver.getRow();
            int colCount =
            for (int j = 1; j <= colCount; j++) {
                if (resolver.getConflictValue(j)
                    != null) {
                    crsValue = crs.getObject(j);
                    resolverValue = 

                    // ...
                    // compare crsValue and
                    // resolverValue to
                    // determine the value to be
                    // persistent

                    resolvedValue = crsValue;
                        j, resolvedValue);

Notifying Listeners

Being a JavaBeans component means that a RowSet object can notify other components when certain things happen to it. For example, if data in a RowSet object changes, the RowSet object can notify interested parties of that change. The nice thing about this notification mechanism is that, as an application programmer, all you have to do is add or remove the components that will be notified.

This section covers the following topics:

Setting Up Listeners

A listener for a RowSet object is a component that implements the following methods from the RowSetListener interface:

An example of a component that might want to be a listener is a BarGraph object that graphs the data in a RowSet object. As the data changes, the BarGraph object can update itself to reflect the new data.

As an application programmer, the only thing you must do to take advantage of the notification mechanism is to add or remove listeners. The following line of code means that every time the cursor for the crs objects moves, values in crs are changed, or crs as a whole gets new data, the BarGraph object bar will be notified:


You can also stop notifications by removing a listener, as is done in the following line of code:


Using the Coffee Break scenario, assume that headquarters checks with the database periodically to get the latest price list for the coffees it sells online. In this case, the listener is the PriceList object priceList at the Coffee Break web site, which must implement the RowSetListener methods cursorMoved, rowChanged, and rowSetChanged. The implementation of the cursorMoved method could be to do nothing because the position of the cursor does not affect the priceList object. The implementations for the rowChanged and rowSetChanged methods, on the other hand, must ascertain what changes have been made and update priceList accordingly.

How Notification Works

In the reference implementation, methods that cause any of the RowSet events automatically notify all registered listeners. For example, any method that moves the cursor also calls the method cursorMoved on each of the listeners. Similarly, the method execute calls the method rowSetChanged on all listeners, and acceptChanges calls rowChanged on all listeners.

Sending Large Amounts of Data

The sample code CachedRowSetSample.testCachedRowSet demonstrates how data can be sent in smaller pieces.

Previous page: Using JdbcRowSet Objects
Next page: Using JoinRowSet Objects