[jifty-devel] Improving our plugin system to make things like
jpeacock at rowman.com
Mon Dec 11 09:26:44 EST 2006
Jesse Vincent wrote:
> Jifty ends up handing the actions for the plugin off to:
The actions /should/ be (at least how I tried to make them work):
which are just trivial wrappers around the corresponding base classes.
My testing seems to suggest that this part is working (at least it seems
like it from the "...dispatching..." debug statements).
The problem is that the Login base classes are currently hard-coded for
the fields in the base model class, so either we expose a method that
the child class can override to validate the input fields or we make the
validator method use the correct database schema to build the validation
on the fly (the latter being my current preferred method). This would
mean we could still use the normal Perl inheritance model (but see below).
Sadly, I'm in the midst of a somewhat failed system upgrade (my root
partition got corrupted), so I'm unable to do anything constructive
until I have that sorted out. Computers suck sometimes... :(
> An alternate vision is brewing in my head. I'm thinking of doing
> something like allowing plugins to manipulate the inheritance tree of
> application classes.
But how does that inheritance happen in practice? I'm not sure I see
how this would be any more flexible than the on-the-fly subclassing that
the Jifty::Plugin::ClassLoader already does.
> Does this sound like a sane design? I think it would somewhat simplify
> our conceptual model and be a lot more flexible. But thoughts are much
I *think* we can already get this as long as we can get the field
validation stuff worked out.
Director of Information Research and Technology
Rowman & Littlefield Publishing Group
4501 Forbes Boulevard
Lanham, MD 20706
More information about the jifty-devel