[GNUe-dev] geas2gsd

[chafar]

Hello.

I've made my first incursion into python/gnue whith the help of J.Cater
(with his gnue-common dev guide). I think the result could be
interesting for more people, so I post here, attached, for you, core
developers, dispose as convenient.

What it does is to export user selected geas classes data to schema
format. I've tried against appserver and (postgres) backend and seem to
work, though just with a few examples, so it's easy to have bugs.

It doesn't use as much gnue code as it should, so I'm not sure you'll
want to changing geas2gsd to gnue-geas2gsd ;). I signed a copyright
assignment to FSF related to gnue, so this shouldn't be a problem. If,
instead, you want it as contrib code I could send you a copy with the
proper license requirements (along as I could figure out what they
are;).

I've seen Sacha's interes in a thing like that in irc logs, and don't
know if he'w subscribed to this list, so I'll try to remember to put
his addres into Cc.

Finally, a word about introspection (recently discussed). Not just gcds
preview has shown that defining data structures could be as easy with
them as with schemas (that is, the existence of gnue module in the db
can be transparent if you don't want to know about it) but, as my code
shows, is the better and easier font for introspection and, even more,
it should be extremely more easy using clients against appserver than
against backend.

What I want to say is that using gnue clients against backend
datasources should be considered obsolete in a medium (short, better;)
term, and in such scenario, efforts towards backend introspection may
be considered a luxury(?... well, if you don't want to be able to take a
pre-existent database and build a geas arquitecture for it, just now).

That is, in my opinion, geas itself should be the unique mandatory
authority about data structure of geas applications, and non-geas gnue
apps should be considered nearly-obsolete.

greetings
-- 
José Esteban
Granada - Spain
http://alcances.net