Mailing-List: contact commons-dev-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Jakarta Commons Developers List" <commons-dev@jakarta.apache.org>
Received-SPF: pass (asf.osuosl.org: domain of joerg.schaible@gmx.de designates
 213.165.64.20 as permitted sender)
From: =?iso-8859-1?q?J=F6rg_Schaible?= <joerg.schaible@gmx.de>
Organization: N/A
To: "James Carman" <james@carmanconsulting.com>
Subject: Re: [proxy] vs proxytoys
Date: Thu, 25 Aug 2005 00:49:26 +0200
User-Agent: KMail/1.8.1
Cc: "'Jakarta Commons Developers List'" <commons-dev@jakarta.apache.org>
References: <200508242132.j7OLWJSn023632@carmanconsulting.com>
In-Reply-To: <200508242132.j7OLWJSn023632@carmanconsulting.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline
Message-Id: <200508250049.26863.joerg.schaible@gmx.de>

On Wednesday 24 August 2005 23:32, James Carman wrote:
JC> To be quite honest, nobody has fully provided the JavassistProxyFactory
JC> implementation to commons-proxy yet, either. :-)  I started on it but
JC> couldn't get the intercepted proxies working. =20

:))

JC> One key difference, it seems, between proxytoys and commons-proxy is th=
at
JC> commons-proxy somewhat assumes that there will be a "target" object (so=
me
JC> actual object that implements the core functionality) to the invocation.
JC> ProxyToys doesn't seem to make that assumption (correct me if I'm wrong=
).

Right. Some view toy proxies can act without a target.

JC> One problem that I see with ProxyToys is in the implementation of the
JC> DelegatingInvoker and SimpleInvoker (what I would have to use if I did=
=20
have
JC> a target object) class.  No matter what ProxyFactory you're using,
JC> DelegatingInvoker and SimpleInvoker ultimately use JDK reflection to=20
invoke
JC> the method on the delegate/target.

Yes. In the end it the proxy uses always InvocationHandler.invoke() as=20
callback.

JC> Commons-proxy doesn't do that. =20

Must have a look at the code :)

JC> ProxyToys, ultimately, should be much
JC> slower, which is illustrated by the following code:
JC>=20
JC> public class CompareToProxyToys
JC> {
JC>     private static final int N =3D 100000000;
JC>=20
JC>     public static void main( String[] args )
JC>     {
JC>         final EchoImpl impl =3D new EchoImpl();
JC>         final Echo proxyToys =3D ( Echo )new
JC> com.thoughtworks.proxy.factory.CglibProxyFactory().createProxy( new=20
Class[]
JC> { Echo.class }, new SimpleInvoker( impl ) );
JC>         final Echo commonsProxy =3D ( Echo )new
JC> CglibProxyFactory().createProxy( new ConstantProvider<Echo>( impl ),
JC> Echo.class );
JC>         System.out.println( MessageFormat.format( "Average duration of
JC> ProxyToys is approximately {0,number,0.00#} times slower than
JC> commons-proxy.", new Double( averageDuration( proxyToys ) /
JC> averageDuration(commonsProxy ) ) ) );
JC>     }
JC>=20
JC>     private static double averageDuration( Echo echo )
JC>     {
JC>         final long before =3D System.nanoTime();
JC>         for( int i=3D 0; i < N; ++i )
JC>         {
JC>             echo.echoBack( "Hello, World!" );
JC>         }
JC>         final long after =3D System.nanoTime();
JC>         return ( ( after - before ) * 1.0 / N );
JC>     }
JC> }
JC>=20
JC> ProxyToys' average duration was approximately 7x slower than commons-pr=
oxy
JC> using this crude test (I'm not saying these are the best metrics in the
JC> world).

I'll investigate.

JC> Again, I allowed both implementations to use a CGLIB-based proxy.

Well, no <g>

The CglibProxyFactory of ProxyToys will use internally a JDK proxy if the=20
proxy implements exactly one interface. Don't ask me why, that is not my=20
code. But I'll try this without that "optimization".

=2D J=F6rg

---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org