Mailing-List: contact dev-help@geronimo.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@geronimo.apache.org
Mime-Version: 1.0 (Apple Message framework v618)
In-Reply-To: <58537.220.247.207.122.1090561380.squirrel@220.247.207.122>
References: <55991.220.247.207.122.1090412828.squirrel@220.247.207.122>
 <58315.220.247.207.122.1090462925.squirrel@220.247.207.122>
 <32409D56-DC09-11D8-BD30-000393DB559A@coredevelopers.net>
 <58537.220.247.207.122.1090561380.squirrel@220.247.207.122>
Content-Type: text/plain; charset=US-ASCII; format=flowed
Message-Id: <20C7E6A8-DCC9-11D8-BD30-000393DB559A@coredevelopers.net>
Content-Transfer-Encoding: 7bit
From: Dain Sundstrom <dain@coredevelopers.net>
Subject: Re: [code generation]webservice deployment in geronimo
Date: Fri, 23 Jul 2004 09:55:40 -0700
To: dev@geronimo.apache.org

On Jul 22, 2004, at 10:43 PM, Srinath Perera wrote:

>> If we need to generate Java source code and compile it, then I would
>> rather have the deployment process simply do that.  What I was really
>> getting at is can we simply generate the byte code directly and skip
>> the Java source generation and compile steps?
>
> Have heard of generating the byte code directly but never done it my 
> self
> before. My only concern is in that way will the debuging become
> nightmare. (as we are not good @ reading byte codes as computers.)
>
> plus there would be quite bit of code changing in the generation code..
> but that is not impossible if it worth it.

Normally, you write the standard code reused across all beans in normal 
java code that you compile an pack in a library jar. Then you only 
generate a tiny bit of byte code to implement an interface and call 
your library java code.  This is how we implement EJB.  We generate an 
implementation of the EJB interfaces and all that implementations does 
is call our EJB library code, which does stuff like check permissions 
and start transactions.

Anyway, it is not important for now, but something I'd like to do in 
the future.

-dain