Hi folks; I'm trying to figure out how best to integrate SQLAlchemy mappers with Trellis components, so that Trellis applications can use SQLAlchemy in a relatively-transparent way.

The catch is that Trellis components use custom descriptors (like the property builtin, only completely different) so that the Trellis system can see when attributes are read or written, so that inter-component dependencies can be automatically detected, and dependencies automatically updated when values change.

I've been digging through the docs and the sqlalchemy.orm package trying to figure out a clean way to integrate the two, but it appears that the only documented ways to deal with something like this are to either 1) create alternate attributes like '_foo' to be mapped, and then delegate to those attributes from the descriptor, or 2) create an dummy class to which the mapping is applied, and then create linkages between the mapped object and the "real" object.

...and all that's just for the non-collection attributes. I'd really like to use Trellis collections (which also detect and auto-notify their dependencies), too.

Before diving even deeper into this, and possibly end up relying on undocumented aspects of the ORM package for all this, I thought it might be a good idea to see if anybody has any pointers on accomplishing this, especially if there are standard interfaces or things I can subclass to add this stuff in. My ideal outcome would be if I could set things up so that when somebody defines mappings on a Trellis component class, it "just works", even if that means I have to have a whole bunch of logic behind the scenes plugging into SQLAlchemy-supplied hooks.

If that isn't possible, however, I could provide an alternative mapping interface, but I'd still like to be able to have SQLAlchemy returning trellis components, rather than dummy objects that then have to be re-wrapped. (Of course, there's yet another layer of workaround possible, but...)

Anyway, the documentation for the mapping framework seems a bit slim, apart from a few bits about doing collections. I thought I'd check to make sure I'm looking in the right place(s), and check in to see if any of the SQLAlchemy developers have any interest in this sort of integration.

The primary use case for this Trellis-SQLAlchemy integration is database-backed GUI applications that use the Trellis to automatically update displays from model changes, and to update model objects from user input. The Trellis makes this kind of code very simple because it doesn't require explicit publish-subscribe operations; inter-object dependencies are automatically detected, and dependent operations are automatically re-run when their dependencies change.

The specific application that will be used for first, is the experimental re-architecture branch of the Chandler open-source PIM. Right now, there are some simple Trellis-based GUI and model components, that we'd like to hook up to a database backend.

Anyway, if anyone has any pointers, ideas, etc., it would be most helpful. Thanks.