Your DAO is quite evolved, and I don't realy know Tapestry, but let me still try to say how I would do it:

First I would directly expose the Registry in the Global and the application gets the DAO directly from the Registry. The Registry should not be build up too often it's just too time-consuming (and you should call shut-down on the reigstry before you release it).

public Global{
 private Registry registry;
 public Global(){
  //build up the registry here
  //because I certainly need
  //and don't need synchronize later
 }

 public Registry getHiveMindRegistry(){
   return registry;
 }

 public void shutdownHiveMindRegistry() {
   registry.shutdown();
 }
}

Than I would give out the DOA directly as a service. Than as you also said (for your second question) I would implement a ConnectionPool service interface and a (or more) implementation(s) (one which just holds one connection - for your testing). Than I would change the DAO-implementation to take a ConnectionPool service as parameter and get the connection from there.

interface ConnectionPool {
 public Connection getConnection();
 public void returnConnection(Connection connection);
}

<module...>

<service-point id="DAO"
  interface="com.foobar.eci.dao.DAOFactory">
  <invoke-factory service-id="hivemind.BuilderFactory">
        <construct class="com.foobar.eci.dao.mysql.MySQLDAOFactory">
        <set-service id="MySqlConnectionPool" property".."/>
     </contruct>
  </invoke-factory>
</service>

<service-point id="ConnectionPool"
  interface="com.foobar.eci.dao.ConnectionPool">
  <invoke-factory service-id="hivemind.BuilderFactory">
        <construct class="com.foobar.eci.dao.mysql.MySQLSingelConnectionPool">
        <set ... the Connection properties/>
     (etc..)
     </construct>
  </invoke-factory>
</service>

Now you saved the factory and you have access to the single connection in your test-code. In running code you can still change to a real connection pool. With this suggestion you still have to change both the DAO service- definition and the ConnectionPool definition if you want to change the DAO implementation. However this can be cleaned up, see the Overriding-Services tutorial at the hivemind page.

Finally: I think (like you) you should if you are going to use a 'real' connection-pool take the connection out of the commands. I would give the commands the connection as a parameter each time they are called. The caller would get it from the ConnectionPool call the command with it and return it straight to the pool.



<invoke-factory service-id="hivemind.BuilderFactory">
<construct
                class="com.foobar.eci.dao.mysql.MySQLDAOFactory">
<set property="name" value="eci"/>
<set property="protocol" value="mysql"/>
<set property="driver"
                     value="org.gjt.mm.mysql.Driver"/>
<set property="host" value="localhost"/>
<set property="port" value="3906"/>
<set property="user" value="bogon"/>
<set property="password" value="foobar"/>
</construct>
</invoke-factory>
</service-point>
</module>
On Fri, 3 Oct 2003 06:51:56 -0500, Bill Lear <[EMAIL PROTECTED]> wrote:

I've been using hivemind to build a DAO factory.  I think I may be
using this in a rather crude way that undoes some of the good hivemind
has done, so I thought I'd ask for some advice.  I have a second question
regarding JDBC connections.

First, in Tapestry, I store my DAO object in the Global class, but
I seem to have reverted to the singleton pattern instead of
doing things the hivemind way:

public class Global {
private static DAO dao;

private static Object daoMutex = new Object();

public DAO getDAO() throws DAOException {
synchronized (daoMutex) {
if (dao == null) {
dao = createDAO();
}
}
return dao;
}

private DAO createDAO() throws DAOException {
DAOFactory f = (DAOFactory) Services.getService(DAOFactory.class);
return f.createDAO();
}
}

Services is a wrapper around Hivemind that gets a service based on a
class name:

public class Services {
public static Object getService(Class theClass) {
ClassResolver resolver = new DefaultClassResolver();
RegistryBuilder builder = new RegistryBuilder();

builder.processModules(resolver);

Registry registry = builder.constructRegistry(Locale.getDefault());

return registry.getService(theClass.getName(), theClass);
}
}

My hivemodule looks like this:

<module id="com.foobar.eci.dao" version="1.0.0">
<service-point id="DAOFactory"
interface="com.foobar.eci.dao.DAOFactory">
<invoke-factory service-id="hivemind.BuilderFactory">
<construct
                class="com.foobar.eci.dao.mysql.MySQLDAOFactory">
<set property="name" value="eci"/>
<set property="protocol" value="mysql"/>
<set property="driver"
                     value="org.gjt.mm.mysql.Driver"/>
<set property="host" value="localhost"/>
<set property="port" value="3906"/>
<set property="user" value="bogon"/>
<set property="password" value="foobar"/>
</construct>
</invoke-factory>
</service-point>
</module>

I'm not quite sure how to have hivemind construct both the Global
object and DAO object, and set the DAO object in the Global object.


Second question: I use a command pattern to separate the implementation of
different methods in the DAO class. So, if I have a DAO that looks like
this:


public interface DAO {
Employee getEmployee(Integer employeeID) throws DAOException;

Employee[] getEmployees(String sortColumn) throws DAOException;

void insertEmployee(Employee employee) throws DAOException;
}

I implement this with a class to expose this interface, and three
classes to implement each of the commands.  Notice the MySQLDAO
class stores the JDBC connection object and passes it to each
of the command classes it constructs:

public class MySQLDAO implements DAO {
void init(Connection connection) throws DAOException {
try {
getEmployee = new GetEmployee(connection);
getEmployees = new GetEmployees(connection);
insertEmployee = new InsertEmployee(connection);
} catch (SQLException sex) {
LOG.error(sex.toString());
throw new DAOException(sex.toString());
}
}

private GetEmployee getEmployee;

public Employee getEmployee(Integer employeeID) throws DAOException {
return getEmployee.execute(employeeID);
}

private GetEmployees getEmployees;

public Employee[] getEmployees(String sortColumn) throws DAOException {
return getEmployees.execute(sortColumn);
}

private InsertEmployee insertEmployee;

public void insertEmployee(Employee employee) throws DAOException {
insertEmployee.execute(employee);
}
}

The factory is what I have Hivemind creating:

public class MySQLDAOFactory implements DAOFactory {
private String name;
private String protocol;
private String driver;
private String host;
private String port;
private String user;
private String password;

public DAO createDAO() throws DAOException {
MySQLDAO dao = new MySQLDAO();

try {
String connURL = "jdbc:" + protocol + "://";

if (host.length() == 0) {
connURL += name;
} else {
connURL += host + ":" + port + "/" + name;
}

// This ensures no timeouts for MySQL after default idle period
connURL += "?autoReconnect=true";

Class.forName(driver);
dao.init(DriverManager.getConnection(connURL, user, password));
} catch (SQLException sex) {
throw new DAOException(sex);
} catch (ClassNotFoundException ex) {
throw new DAOException(ex);
}

return dao;
}

// Get-Set methods for private variables elided ...
}

So, what I'm wondering is: I have embedded a JDBC connection object in
the each of the three command objects.  This seems too direct and
inflexible, but I don't know what other alternatives to use.  Also, I
would like to be able to unit-test each of the three command classes,
but I would prefer to be able to use hivemind to construct the JDBC
connection for me, in which case, I might re-factor things in the
MySQLDAO class itself to also take advantage of hivemind to get the
connection object (but then, what about pooling?  Can hivemind create
a JDBC connection?  Is this desirable?).

Finally, when testing the DAO, I need to be able to clear the database
before each test case method is run.  I would prefer to use the same
connection that the DAO has, but exposing that in the DAO interface
seems a bit lame (Connection getConnection()) and too concrete.

In any case, if anyone has comments or suggestions to share on managing
JDBC connections, I'd appreciate it.


Bill


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]




-- Using M2, Opera's revolutionary e-mail client: http://www.opera.com/m2/

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to