I suppose something like:
routes_in = (
(r'/appname/model1/$id1/model2/$id2/model3/$id3',
'/appname/model3/index/$id1/$id2/$id3'),
(r'/appname/model1/$id1/model2/$id2',
'/appname/model2/index/$id1/$id2'),
(r'/appname/model1/$id1', '/appname/model1/index/$id1')
)
In that case, the values of $id1, $id2, and/or $id3 would be available to
the respective functions in request.args. Note, the pattern matching stops
at the first matched pattern, so you have to put the most specific pattern
first. Also, note that if there are more models/patterns to deal with, you
could write a function to generate the above list of tuples for you:
def generate_my_routes():
[code to generate tuples]
return route_tuples
routes_in = generate_my_routes()
Also, I wonder if you really need separate controllers for each model.
Perhaps instead you could have a single function in the default controller
that handles all the data requests from the models. In that case, in a URL
with model1/{model1_id}/model2/{model2_id}/model3/{model3_id}, model1,
model1_id, model2, etc. would all be in request.args[0:6]. Your function
could then parse the request.args and generate the appropriate query to
return the data.
Anthony
On Tuesday, March 20, 2012 2:06:26 PM UTC-4, rdodev wrote:
>
> Now more to our use case: a few days ago I wrote asking how could I
> accomplish and more logical-relational url scheme. One of the "hard"
> requirements from our customer is to have no action names in the URL
> and actions should be distinguished via REST. They also asked us to
> preserve relationships in the URL. So, the REST was easily
> accomplished via @request.restful() decorator on the index function of
> each controller. Now what I'm trying to accomplish is:
>
> Assume I have three models: model1 ... model3 (where model1 is the
> parent of model2 and model2 is parent of model3)
> And their corresponding controllers: cont1 ... cont3
>
> Ideally, a user (assuming already-authenticated), can GET request URL
> http://domain.com/AppName/model1/{model1_id}/model2/{model2_id} and
> that should be routed to the default function in cont2 controller
> passing {model2_is} as param
> A GET request to http://domain.com/AppName/model1/{model1_id}/model2
> should also route to the default function of cont2
> A GET request to
>
> http://domain.com/AppName/model1/{model1_id}/model2/{model2_id}/model3/{model3_id}
> should be routed to the default function of cont3
>
> ... and so forth.
>
> We already did a proof-of-concept with couple other frameworks, and
> would certainly like to present web2py as an option because it's
> otherwise a kickass framework, but the URLs are make-or-break for them
> and we really have no say in the matter. So any help in this specific
> case is much appreciated.
>
> On Tue, Mar 20, 2012 at 1:09 PM, Jonathan Lundell <jlund...@pobox.com>
> wrote:
> > On Mar 20, 2012, at 10:00 AM, Ruben Orduz wrote:
> >> Not sure why so defensive. All I'm suggesting is that there is
> >> precedent in simpler routing schemes: consider django's (or Rails'),
> >> for instance. In django you have a urls.py at the root, and then
> >> optionally one in your app folder. You have to explicitly tell the
> >> root urls.py which urls to use for your app. There's no guesswork and
> >> no contingent and thus no confusion. The routes auto-discovery and
> >> contingencies, in this case, and in my opinion, are not intuitive and
> >> I would say against the core python concept of explicit over implicit.
> >
> > The rule for the parametric router (as opposed to the pattern-based
> router) is that a) you can put an app-specific router in the base
> routes.py, but b) if an app has its own parametric router in its own
> routes.py, it overrides the base routers.
> >
> > The pattern-based router would need additional syntax to accomplish (a),
> though I supposed you could write your regexes cleverly and accomplish the
> same general thing.
> >
> > I think that's the proper behavior, though it's not an apples-to-apples
> comparison.
> >
> >
> >>
> >> On Tue, Mar 20, 2012 at 12:43 PM, Anthony <abasta...@gmail.com> wrote:
> >>>> I would suggest the following behavior though:
> >>>>
> >>>> routes_in and routes out in the _base_ routes.py should be completely
> >>>> ignored if routes_app is not commented out.
> >>>
> >>> No, it should not. You may have multiple applications, but perhaps
> only some
> >>> of them have app-specific routes (i.e., a routes.py file in the
> application
> >>> folder). In that case, you want routes_app to match routes that belong
> to
> >>> the apps with app-specific routes, but you want any routes that don't
> match
> >>> routes_app to fall back to the routes_in/routes_out in your base
> routes.py
> >>> file. We do not want to completely ignore routes_in/routes_out any time
> >>> routes_app is present -- we only want to ignore them for the specific
> apps
> >>> that are matched in routes_app.
> >>>
> >>>>
> >>>> routes_app should be always completely ignored in _app-specific_
> routes.py
> >>>
> >>> I believe this is already the case. The example file even says, "This
> entry
> >>> is meaningful only in the base routes.py."
> >>>
> >>> It might be easier if you describe the routing you're trying to do so
> we can
> >>> help you with your specific case.
> >>>
> >>> Anthony
> >
> >
>
>
