Hi,
Because i want Database Objects to control them self completely (like Enity
beans)
I must have a way to get Transaction support and make sure that in on thread
the same connection is being used. For example:
class Person
{
private String name;
private Country country;
public static Person findByPrimaryKey(int i)
{
Connection connection = datasource.getConnection();
// Select
Person person = new Person();
person.setName(rs.getString(1));
person.setCountry(Countr.findByPrimaryKey(rs.getInt(2)));
connection.close();
}
}
class Country
{
privatge String name
public static Country findByPrimaryKey(int i)
{
// Now another connection is being taken from the datasource
// if max == 2 then another call in this thread to the datasource
// will present a deadlock!!
Connection connection = datasource.getConnection();
// Select
Country country = new Country ();
/// Filling
connection.close()
return country;
}
}
So i want a datasource to return a connection that is already attached to
the
current thread.
Same goes for updating or inserting of data for example a new Person with a
new Country object that i want to insert into the database
person clas:
{
public void store()
{
Connection connection = datasource.getConnection();
// Before storing the country id the country must first be stored:
country.store();
// if all goes well:
// Store the person (insert into person values(?,?,?,?)
ps.setInt(2,country.getPrimaryKey());
// if not all goes well the connection.rollback must be called.
// but one problem the country was stored under a different connection
// so it will not be rolledback!!
}
}
country class
{
public void store()
{
// Another connection is taken so if it goes bad then you can't
rollback
// everything!!
Connection connection = datasource.getConnection();
// Store the country (insert into person values(?,?,?,?)
}
}
To get around those above problems i am building a Transaction class.
This wraps around the Datasource
it has a static method:
public static Transaction startTransaction() // Only starts a transaction
onces for the current thread.
and the Transaction object has methods:
public Connection getConnection()
public void setRollbackOnly() // Error occured somewhere so only a rollback
allowed.
public int getStatus() // Must rollback, Healty.
public void rollback()
public void commit() // Can only do this if mustRollback == false else throw
exception
public void stopTransaction()
In the above examples i don't do
datasource.getConnection()
but
Transaction transaction = Transaction.startTransaction();
Connection connection = transaction.getConnection();
// Cannot call connection.close() , connection.rollback(),
connection.commit() directly!!!
transaction.stopTransaction();
If anyone is interrested or has some comments i am happy to hear it.
Johan
> -----Original Message-----
> From: Michael McCallister [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, March 08, 2001 11:19 PM
> To: [EMAIL PROTECTED]
> Subject: Business Object Bean Persistence Strategies
>
>
> What strategies or patterns do people use to manage persistence
> of business
> object data in a Struts application when there is no EJB layer
> and there is
> a desire to keep the business objects as independent of the web
> portion of
> the application as possible? Do you use the Struts DataSource and pass
> either it or a Connection as a parameter to bean methods that take
> responsibility for managing persistence? Do you follow the J2EE
> blueprint
> and create separate Data Access Objects to support persistence? Are you
> using an open-source framework to manage persistence? This seems like a
> common problem, but I haven't seen much talk about common solutions.
>
>
> Mike
>
>
