JAAS::Widget README
===================

** CVS mess-up: files in Widget should be in JAAS/Widget. The following files
should not exist:

Widget/Address.pm
Widget/Base.pm
Widget/Type/*
test1.pl

TESTING:

The test scripts in the test/ dir must be launched in sequence.

1) Start the Context server 'server.pl'. This is the brain of the operation.

2) In a separate process, run 'user-obj.pl'. It will create a user object
and a corresponding widget in Context.

3) Run either 'widget-set.pl' or 'widget-get.pl' for obvious effect. Both
dump (as in Data::Dumper) their results on screen.

4) To test the HTML::Mason components, request first 'form.mason'. This will
read the widget object (set in step #2) and display a form to fill out.
Clicking 'submit' will move to 'process.mason', which will check for
validation. If everything's OK, you're invited to click a button to go on to
'display.mason' which will show the data you entered.

You can always redo step #3 during this step to check out what's happenin'


PROBLEMS:

The Mason components, especially the JAAS::Widget::ClientHTML class, are
really a proof-of-concept that I wrote for fun and education. It is not
intended to be part of JAAS, altough it can inspire a few ideas.

mod_perl's code caching doesn't mix well with POE/IKC. In Mason, IKC clients
do not exit properly. Maybe my code is the problem, but the fact is right
now, you can't run a component twice. That means no page reloading and no
re-testing, unless you restart Apache first. See the proof-of-concept lame
excuse above.

For reasons currently unknown, the newer version of JAAS::Object::Lifetime
breaks things, so I use the previous one. I haven't got time to check it yet.
For a test, uncomment the followng line in 'user-obj.pl', method
'fill_widget':

#       print "$$: ", ref($w), "\n";

Both versions of Lifetime set $w as different objects.

I noticed at the last minute that the arguments to JAAS::Widget::Group->new()
didn't match well with Config/Factory syntax. It will be corrected very
soon.






 