hibernate-devel

Re: [Hibernate] Ohhh this looks cool

[<phr...@im...>]

hi,

On Tue, 22 Oct 2002 04:34:54 +1000, "Gavin King" <ga...@ap...>
said:
> Can we use this to do lazy initialization of objects without the need for
> proxy interfaces?

wouldn't you have to do this at class loading time then? that is: how
would you make your generated look like the original one? with interfaces
it's doable at runtime, as long as the client refers to the generated
class through the interface...

    viktor
-- 
  
  phr...@im...

-- 
http://fastmail.fm - Access your email with Outlook or over the web

RE: [Hibernate] Ohhh this looks cool

[Ampie B. <amp...@mw...>]

hi,

Cglib is brilliant for simply wrapping method calls. As I understand its
main feature is generating classes dynamically  by merging two classes that
were not in any way related at compile time. The generated class looks like
a subclass of the original, so you don't need 1000 interfaces as with
proxies.

If access to the internal state of an object always goes through its
accessors, this tool could be ideal for lazy fetching of state just before a
getter or setter method call, maybe through fetch groups.

I think it is a good idea Gavin. In theory, it should be faster than
proxies, but one might just want to test that theory.

Ampie

-----Original Message-----
From: hib...@li...
[mailto:hib...@li...]On Behalf Of
phr...@im...
Sent: 22 October 2002 00:55
To: Gavin King; hib...@li...
Subject: Re: [Hibernate] Ohhh this looks cool


hi,

On Tue, 22 Oct 2002 04:34:54 +1000, "Gavin King" <ga...@ap...>
said:
> Can we use this to do lazy initialization of objects without the need for
> proxy interfaces?

wouldn't you have to do this at class loading time then? that is: how
would you make your generated look like the original one? with interfaces
it's doable at runtime, as long as the client refers to the generated
class through the interface...

    viktor
--

  phr...@im...

--
http://fastmail.fm - Access your email with Outlook or over the web


-------------------------------------------------------
This sf.net emial is sponsored by: Influence the future
of  Java(TM) technology. Join the Java Community
Process(SM) (JCP(SM)) program now.
http://ad.doubleclick.net/clk;4699841;7576298;k?http://www.sun.com/javavote
_______________________________________________
hibernate-devel mailing list
hib...@li...
https://lists.sourceforge.net/lists/listinfo/hibernate-devel

Re: [Hibernate] Ohhh this looks cool

[Viktor S. <phr...@im...>]

hi,

Ampie Barnard wrote:

>The generated class looks like a subclass of the original
>  
>
heh, but of course :) silly me...

>I think it is a good idea Gavin. In theory, it should be faster than
>proxies, but one might just want to test that theory.
>
i think the difference would be marginal, as dynamic proxies also use 
class generation behind the scenes...

    v.

Re: [Hibernate] Ohhh this looks cool

[Gavin K. <ga...@ap...>]

> i think the difference would be marginal, as dynamic proxies also use 
> class generation behind the scenes...

I have a suspicion it will be slightly slower with CGLIB, actually.

Re: [Hibernate] Ohhh this looks cool

[Gavin K. <ga...@ap...>]

I'm currently re-implementing the Proxy feature using CGLIB, as an
alternative
to JDK proxies. Its turning out really nicely. I have always been so
frustrated
by the inability to intercept calls to an object if the object doesn't
implement
an interface. CGLIB seems to solve that problem nicely.

> Cglib is brilliant for simply wrapping method calls. As I understand its
> main feature is generating classes dynamically  by merging two classes
that
> were not in any way related at compile time. The generated class looks
like
> a subclass of the original, so you don't need 1000 interfaces as with
> proxies.
>
> If access to the internal state of an object always goes through its
> accessors, this tool could be ideal for lazy fetching of state just before
a
> getter or setter method call, maybe through fetch groups.
>
> I think it is a good idea Gavin. In theory, it should be faster than
> proxies, but one might just want to test that theory.
>
> Ampie
>
> -----Original Message-----
> From: hib...@li...
> [mailto:hib...@li...]On Behalf Of
> phr...@im...
> Sent: 22 October 2002 00:55
> To: Gavin King; hib...@li...
> Subject: Re: [Hibernate] Ohhh this looks cool
>
>
> hi,
>
> On Tue, 22 Oct 2002 04:34:54 +1000, "Gavin King" <ga...@ap...>
> said:
> > Can we use this to do lazy initialization of objects without the need
for
> > proxy interfaces?
>
> wouldn't you have to do this at class loading time then? that is: how
> would you make your generated look like the original one? with interfaces
> it's doable at runtime, as long as the client refers to the generated
> class through the interface...
>
>     viktor
> --
>
>   phr...@im...
>
> --
> http://fastmail.fm - Access your email with Outlook or over the web
>
>
> -------------------------------------------------------
> This sf.net emial is sponsored by: Influence the future
> of  Java(TM) technology. Join the Java Community
> Process(SM) (JCP(SM)) program now.
>
http://ad.doubleclick.net/clk;4699841;7576298;k?http://www.sun.com/javavote
> _______________________________________________
> hibernate-devel mailing list
> hib...@li...
> https://lists.sourceforge.net/lists/listinfo/hibernate-devel
>
>
>
> -------------------------------------------------------
> This sf.net emial is sponsored by: Influence the future
> of  Java(TM) technology. Join the Java Community
> Process(SM) (JCP(SM)) program now.
>
http://ad.doubleclick.net/clk;4699841;7576298;k?http://www.sun.com/javavote
> _______________________________________________
> hibernate-devel mailing list
> hib...@li...
> https://lists.sourceforge.net/lists/listinfo/hibernate-devel

Re: [Hibernate] Ohhh this looks cool

[Gavin K. <ga...@ap...>]

> Cglib is brilliant for simply wrapping method calls. As I understand its
> main feature is generating classes dynamically  by merging two classes
that
> were not in any way related at compile time. The generated class looks
like
> a subclass of the original, so you don't need 1000 interfaces as with
> proxies.

And another advantage is that you don't need to break identity equals
the way you do with java.lang.reflect.Proxy. (The Proxy is always a
different object.)

Very, very nice. I wish I knew about this before!

Re: [Hibernate] Ohhh this looks cool

[Gavin K. <ga...@ap...>]

Ah. I ran into a limitation. CGLIB is great for our proxies, as long as the
persistent classes are not polymorphic. But in the case of a class with
subclasses, we don't know the concrete class at proxy-creation time. And
theres no way for the proxy to be an instance of both Bar and Baz if Bar and
Baz are two possible subclasses of foo. Thats a Java language limitation. So
we end up needing the proxy interfaces after all, in that case. Oh well, it
still helps for non-polymorphic associations!

RE: [Hibernate] Ohhh this looks cool

[Ampie B. <amp...@mw...>]

I did some superficial tests under JDK1.4. With invocation handling that
just delegates to the actual implementation, it seems that CG is slower than
Proxies initially. But after several thousand invocations CG gets
substantially (20+%) faster than Proxies. My theory is that CG somehow
benefits more from hot spot compilation. I might be missing something, so if
you want to see my tests I'll send it to the list.

P.S. sorry for barging in on the list like this. I only realised it is
actually the devel list after it was too late.

-----Original Message-----
From: hib...@li...
[mailto:hib...@li...]On Behalf Of Gavin
King
Sent: 22 October 2002 06:41
To: Viktor Szathmary
Cc: hib...@li...
Subject: Re: [Hibernate] Ohhh this looks cool



> i think the difference would be marginal, as dynamic proxies also use
> class generation behind the scenes...

I have a suspicion it will be slightly slower with CGLIB, actually.



-------------------------------------------------------
This sf.net emial is sponsored by: Influence the future of
Java(TM) technology. Join the Java Community Process(SM) (JCP(SM))
program now. http://ad.doubleclick.net/clk;4699841;7576301;v?
http://www.sun.com/javavote
_______________________________________________
hibernate-devel mailing list
hib...@li...
https://lists.sourceforge.net/lists/listinfo/hibernate-devel