Repository for the RowSet reference implementations of the SyncProvider abstract class. These implementations provide a disconnected RowSet object with the ability to synchronize the data in the underlying data source with its data. These implementations are provided as the default SyncProvider implementations and are accessible via the SyncProvider SPI managed by the SyncFactory.

1.0 SyncProvider Reference Implementations

The main job of a SyncProvider implementation is to manage the reader and writer mechanisms. The SyncProvider SPI, as specified in the javax.sql.rowset.spi package, provides a pluggable mechanism by which javax.sql.RowSetReader and javax.sql.RowSetWriter implementations can be supplied to a disconnected RowSet object.

A reader, a javax.sql.RowSetReader object, does the work necessary to populate a RowSet object with data. A writer, a javax.sql.RowSetWriter object, does the work necessary for synchronizing a RowSet object's data with the data in the originating source of data. Put another way, a writer writes a RowSet object's data back to the data source.

Generally speaking, the course of events is this. The reader makes a connection to the data source and reads the data from a ResultSet object into its RowSet object. Then it closes the connection. While the RowSet object is disconnected, an application makes some modifications to the data and calls the method acceptChanges. At this point, the writer is called to write the changes back to the database table or view from which the original data came. This is called synchronization.

If the data in the originating data source has not changed, there is no problem with just writing the RowSet object's new data to the data source. If it has changed, however, there is a conflict that needs to be resolved. One way to solve the problem is not to let the data in the data source be changed in the first place, which can be done by setting locks on a row, a table, or the whole data source. Setting locks is a way to avoid conflicts, but it can be very expensive. Another approach, which is at the other end of the spectrum, is simply to assume that no conflicts will occur and thus do nothing to avoid conflicts. Different SyncProvider implementations may handle synchronization in any of these ways, varying from doing no checking for conflicts, to doing various levels of checking, to guaranteeing that there are no conflicts.

The SyncProvider class offers methods to help a RowSet object discover and manage how a provider handles synchronization. The method getProviderGrade returns the grade of synchronization a provider offers. An application can direct the provider to use a particular level of locking by calling the method setDataSourceLock and specifying the level of locking desired. If a RowSet object's data came from an SQL VIEW, an application may call the method supportsUpdatableView to find out whether the VIEW can be updated.

Synchronization is done completely behind the scenes, so it is third party vendors of synchronization provider implementations who have to take care of this complex task. Application programmers can decide which provider to use and the level of locking to be done, but they are free from having to worry about the implementation details.

The JDBC RowSet Implementations reference implementation provides two implementations of the SyncProvider class:

2.0 Basics in RowSet Population & Synchronization

A rowset's first task is to populate itself with rows of column values. Generally, these rows will come from a relational database, so a rowset has properties that supply what is necessary for making a connection to a database and executing a query. A rowset that does not need to establish a connection and execute a command, such as one that gets its data from a tabular file instead of a relational database, does not need to have these properties set. The vast majority of RowSets, however, do need to set these properties. The general rule is that a RowSet is required to set only the properties that it uses.

The command property contains the query that determines what data a RowSet will contain. Rowsets have methods for setting a query's parameter(s), which means that a query can be executed multiple times with different parameters to produce different result sets. Or the query can be changed to something completely new to get a new result set.

Once a rowset contains the rows from a ResultSet object or some other data source, its column values can be updated, and its rows can be inserted or deleted. Any method that causes a change in the rowset's values or cursor position also notifies any object that has been registered as a listener with the rowset. So, for example, a table that displays the rowset's data in an applet can can be notified of changes and make updates as they occur.

The changes made to a rowset can be propagated back to the original data source to keep the rowset and its data source synchronized. Although this involves many operations behind the scenes, it is completely transparent to the application programmer and remains the concern of the RowSet provider developer. All an application has to do is invoke the method acceptChanges, and the data source backing the rowset will be updated to match the current values in the rowset.

A disconnected rowset, such as a CachedRowSet or WebRowSet object, establishes a connection to populate itself with data from a database and then closes the connection. The RowSet object will remain disconnected until it wants to propagate changes back to its database table, which is optional. To write its changes back to the database (synchronize with the database), the rowset establishes a connection, write the changes, and then once again disconnects itself.

3.0 Other Possible Implementations

There are many other possible implementations of the SyncProvider abstract class. One possibility is to employ a more robust synchronization model, which would give a RowSet object increased trust in the provider's ability to get any updates back to the original data source. Another possibility is a more formal synchronization mechanism such as SyncML (http://www.syncml.org/)