Unfortunately I did an accidental post of this
in the RI devs mailinglist
speaking of being chaotic and early morning.
:-(
Werner
Manfred Geiler schrieb:
Cool thing, Werner. Congrats!
--Manfred
On Tue, May 13, 2008 at 7:45 AM, Werner Punz <[EMAIL PROTECTED]> wrote:
Hello everyone
I just wanted to give notification that I took a small break from my
components project which is still on track and that I am working on myfaces
groovy support.
I got the first artefacts already reloading, and a first code will
be made available by the end of the week, early next week.
(For the Sun guys reading, thanks a lot you gave me the final push to do
it, although I did many things differently than you did)
Ok what will be done:
First of all I replaced all the factories with my own ones
to enable the entire system, I also had to write my own context listener
to handle the classloader issues.
(We really need a change in the spec here to enable scripting properly -
Ed?)
Secondly once the factories were in place I added proxy generation code
wherever needed to enable reloading proxies.
Third, a classloader which forks into the groovy system to load the groovy
code.
The groovy code also has its own reloading weaving code added to enable
reloading of groovy files on the groovy side of things (the woven aop
reloading code is lost on the java side however if you just deliver classes
instead of objects, my first approach was a try to enable everything on the
java side)
So what do we get
Reloading for most artefacts (probably on method level if things work out
the way I intend them to be (For certain artefacts
there are contracts you have to program in on the groovy side to enable
this - aka expose the private properties some artefacts have otherwise
a on method reloading will not be possible).
Maybe and this is a big maybe, if I can get it up and running (I want to
replace code on the fly) reloading of methods on groovy classes loaded by
groovy over the new classloader.
Again this is a big if, I have not prototyped this fully yet, but it should
be possible.
The idea is, once you load an in groovy object over the classloader
it should be possible to change its methods on the fly via the meta
programming capabilities of groovy.
Ok first code around friday or early next week. After that I will start
further discussions.
And again, thanks to Ryan and Ed for finally pushing me towards it
(indirectly by doing it).
I also have to admit I have had a look at some parts of the code to check
how you guys solved some problems I have been facing - especially the
dreaded classloader issues and weaving issues.
(I did most of the stuff differently though due to the different approach I
am doing, of a mixed groovy/java infrastructure to enable some things not
reachable from the java side that easily, also I did not want to change the
core code, I wanted to have it more as an extension).
If you want to have a look at the code upfront
before next week, send me a private mail, I just do not want to post
it yet because it still is not done enough for a public post.
Especially the init code I am still very unhappy with.
Werner
