> Read the latest version of this page
Edit me on GitHub

Combining Traversal and URL Dispatch

When you write most Pyramid applications, you’ll be using one or the other of two available resource location subsystems: traversal or URL dispatch. However, to solve a limited set of problems, it’s useful to use both traversal and URL dispatch together within the same application. Pyramid makes this possible via hybrid applications.

Warning

Reasoning about the behavior of a “hybrid” URL dispatch + traversal application can be challenging. To successfully reason about using URL dispatch and traversal together, you need to understand URL pattern matching, root factories, and the traversal algorithm, and the potential interactions between them. Therefore, we don’t recommend creating an application that relies on hybrid behavior unless you must.

A Review of Non-Hybrid Applications

When used according to the tutorials in its documentation Pyramid is a “dual-mode” framework: the tutorials explain how to create an application in terms of using either url dispatch or traversal. This chapter details how you might combine these two dispatch mechanisms, but we’ll review how they work in isolation before trying to combine them.

URL Dispatch Only

An application that uses url dispatch exclusively to map URLs to code will often have statements like this within your application startup configuration:

1
2
3
4
# config is an instance of pyramid.config.Configurator

config.add_route('foobar', '{foo}/{bar}', view='myproject.views.foobar')
config.add_route('bazbuz', '{baz}/{buz}', view='myproject.views.bazbuz')

Each route typically corresponds to a single view callable, and when that route is matched during a request, the view callable named by the view attribute is invoked.

Typically, an application that uses only URL dispatch won’t perform any calls to pyramid.config.Configurator.add_view() in its startup code.

Traversal Only

An application that uses only traversal will have view configuration declarations that look like this:

1
2
3
4
 # config is an instance of pyramid.config.Configurator

 config.add_view('mypackage.views.foobar', name='foobar')
 config.add_view('mypackage.views.bazbuz', name='bazbuz')

When the above configuration is applied to an application, the mypackage.views.foobar view callable above will be called when the URL /foobar is visited. Likewise, the view mypackage.views.bazbuz will be called when the URL /bazbuz is visited.

Typically, an application that uses traversal exclusively won’t perform any calls to pyramid.config.Configurator.add_route() in its startup code.

Hybrid Applications

Either traversal or url dispatch alone can be used to create a Pyramid application. However, it is also possible to combine the concepts of traversal and url dispatch when building an application: the result is a hybrid application. In a hybrid application, traversal is performed after a particular route has matched.

A hybrid application is a lot more like a “pure” traversal-based application than it is like a “pure” URL-dispatch based application. But unlike in a “pure” traversal-based application, in a hybrid application, traversal is performed during a request after a route has already matched. This means that the URL pattern that represents the pattern argument of a route must match the PATH_INFO of a request, and after the route pattern has matched, most of the “normal” rules of traversal with respect to resource location and view lookup apply.

There are only four real differences between a purely traversal-based application and a hybrid application:

  • In a purely traversal based application, no routes are defined; in a hybrid application, at least one route will be defined.
  • In a purely traversal based application, the root object used is global, implied by the root factory provided at startup time; in a hybrid application, the root object at which traversal begins may be varied on a per-route basis.
  • In a purely traversal-based application, the PATH_INFO of the underlying WSGI environment is used wholesale as a traversal path; in a hybrid application, the traversal path is not the entire PATH_INFO string, but a portion of the URL determined by a matching pattern in the matched route configuration’s pattern.
  • In a purely traversal based application, view configurations which do not mention a route_name argument are considered during view lookup; in a hybrid application, when a route is matched, only view configurations which mention that route’s name as a route_name are considered during view lookup.

More generally, a hybrid application is a traversal-based application except:

  • the traversal root is chosen based on the route configuration of the route that matched instead of from the root_factory supplied during application startup configuration.
  • the traversal path is chosen based on the route configuration of the route that matched rather than from the PATH_INFO of a request.
  • the set of views that may be chosen during view lookup when a route matches are limited to those which specifically name a route_name in their configuration that is the same as the matched route’s name.

To create a hybrid mode application, use a route configuration that implies a particular root factory and which also includes a pattern argument that contains a special dynamic part: either *traverse or *subpath.

The Root Object for a Route Match

A hybrid application implies that traversal is performed during a request after a route has matched. Traversal, by definition, must always begin at a root object. Therefore it’s important to know which root object will be traversed after a route has matched.

Figuring out which root object results from a particular route match is straightforward. When a route is matched:

  • If the route’s configuration has a factory argument which points to a root factory callable, that callable will be called to generate a root object.
  • If the route’s configuration does not have a factory argument, the global root factory will be called to generate a root object. The global root factory is the callable implied by the root_factory argument passed to the Configurator at application startup time.
  • If a root_factory argument is not provided to the Configurator at startup time, a default root factory is used. The default root factory is used to generate a root object.

Note

Root factories related to a route were explained previously within Route Factories. Both the global root factory and default root factory were explained previously within The Resource Tree.

Using *traverse In a Route Pattern

A hybrid application most often implies the inclusion of a route configuration that contains the special token *traverse at the end of a route’s pattern:

1
config.add_route('home', '{foo}/{bar}/*traverse')

A *traverse token at the end of the pattern in a route’s configuration implies a “remainder” capture value. When it is used, it will match the remainder of the path segments of the URL. This remainder becomes the path used to perform traversal.

Note

The *remainder route pattern syntax is explained in more detail within Route Pattern Syntax.

Note that unlike the examples provided within URL Dispatch, the add_route configuration statement named previously does not pass a view argument. This is because a hybrid mode application relies on traversal to do resource location and view lookup instead of invariably invoking a specific view callable named directly within the matched route’s configuration.

Because the pattern of the above route ends with *traverse, when this route configuration is matched during a request, Pyramid will attempt to use traversal against the root object implied by the root factory that is implied by the route’s configuration. Since no root_factory argument is explicitly specified for this route, this will either be the global root factory for the application, or the default root factory. Once traversal has found a context resource, view lookup will be invoked in almost exactly the same way it would have been invoked in a “pure” traversal-based application.

Let’s assume there is no global root factory configured in this application. The default root factory cannot be traversed: it has no useful __getitem__ method. So we’ll need to associate this route configuration with a custom root factory in order to create a useful hybrid application. To that end, let’s imagine that we’ve created a root factory that looks like so in a module named routes.py:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
class Resource(object):
    def __init__(self, subobjects):
       self.subobjects = subobjects

    def __getitem__(self, name):
       return self.subobjects[name]

root = Traversable(
        {'a':Resource({'b':Resource({'c':Resource({})})})}
       )

def root_factory(request):
    return root

Above, we’ve defined a (bogus) resource tree that can be traversed, and a root_factory function that can be used as part of a particular route configuration statement:

1
2
config.add_route('home', '{foo}/{bar}/*traverse',
                 factory='mypackage.routes.root_factory')

The factory above points at the function we’ve defined. It will return an instance of the Traversable class as a root object whenever this route is matched. Instances of the``Resource`` class can be used for tree traversal because they have a __getitem__ method that does something nominally useful. Since traversal uses __getitem__ to walk the resources of a resource tree, using traversal against the root resource implied by our route statement is a reasonable thing to do.

Note

We could have also used our root_factory callable as the root_factory argument of the Configurator constructor, instead of associating it with a particular route inside the route’s configuration. Every hybrid route configuration that is matched but which does not name a factory attribute will use the use global root_factory function to generate a root object.

When the route configuration named home above is matched during a request, the matchdict generated will be based on its pattern: {foo}/{bar}/*traverse. The “capture value” implied by the *traverse element in the pattern will be used to traverse the resource tree in order to find a context resource, starting from the root object returned from the root factory. In the above example, the root object found will be the instance named root in routes.py.

If the URL that matched a route with the pattern {foo}/{bar}/*traverse, is http://example.com/one/two/a/b/c, the traversal path used against the root object will be a/b/c. As a result, Pyramid will attempt to traverse through the edges a, b, and c, beginning at the root object.

In our above example, this particular set of traversal steps will mean that the context resource of the view would be the Traversable object we’ve named c in our bogus resource tree and the view name resulting from traversal will be the empty string; if you need a refresher about why this outcome is presumed, see The Traversal Algorithm.

At this point, a suitable view callable will be found and invoked using view lookup as described in View Configuration, but with a caveat: in order for view lookup to work, we need to define a view configuration that will match when view lookup is invoked after a route matches:

1
2
3
config.add_route('home', '{foo}/{bar}/*traverse',
                 factory='mypackage.routes.root_factory')
config.add_view('mypackage.views.myview', route_name='home')

Note that the above call to add_view() includes a route_name argument. View configurations that include a route_name argument are meant to associate a particular view declaration with a route, using the route’s name, in order to indicate that the view should only be invoked when the route matches.

Calls to add_view() may pass a route_name attribute, which refers to the value of an existing route’s name argument. In the above example, the route name is home, referring to the name of the route defined above it.

The above mypackage.views.myview view callable will be invoked when:

  • the route named “home” is matched
  • the view name resulting from traversal is the empty string.
  • the context resource is any object.

It is also possible to declare alternate views that may be invoked when a hybrid route is matched:

1
2
3
4
5
config.add_route('home', '{foo}/{bar}/*traverse',
                 factory='mypackage.routes.root_factory')
config.add_view('mypackage.views.myview', name='home')
config.add_view('mypackage.views.another_view', name='another',
                route_name='home')

The add_view call for mypackage.views.another_view above names a different view and, more importantly, a different view name. The above mypackage.views.another_view view will be invoked when:

  • the route named “home” is matched
  • the view name resulting from traversal is another.
  • the context resource is any object.

For instance, if the URL http://example.com/one/two/a/another is provided to an application that uses the previously mentioned resource tree, the mypackage.views.another view callable will be called instead of the mypackage.views.myview view callable because the view name will be another instead of the empty string.

More complicated matching can be composed. All arguments to route configuration statements and view configuration statements are supported in hybrid applications (such as predicate arguments).

Using the traverse Argument In a Route Definition

Rather than using the *traverse remainder marker in a pattern, you can use the traverse argument to the add_route() method.

When you use the *traverse remainder marker, the traversal path is limited to being the remainder segments of a request URL when a route matches. However, when you use the traverse argument or attribute, you have more control over how to compose a traversal path.

Here’s a use of the traverse pattern in a call to add_route():

1
2
config.add_route('abc', '/articles/{article}/edit',
                 traverse='/articles/{article}')

The syntax of the traverse argument is the same as it is for pattern.

If, as above, the pattern provided is articles/{article}/edit, and the traverse argument provided is /{article}, when a request comes in that causes the route to match in such a way that the article match value is 1 (when the request URI is /articles/1/edit), the traversal path will be generated as /1. This means that the root object’s __getitem__ will be called with the name 1 during the traversal phase. If the 1 object exists, it will become the context of the request. The Traversal chapter has more information about traversal.

If the traversal path contains segment marker names which are not present in the pattern argument, a runtime error will occur. The traverse pattern should not contain segment markers that do not exist in the path.

Note that the traverse argument is ignored when attached to a route that has a *traverse remainder marker in its pattern.

Traversal will begin at the root object implied by this route (either the global root, or the object returned by the factory associated with this route).

Making Global Views Match

By default, only view configurations that mention a route_name will be found during view lookup when a route that has a *traverse in its pattern matches. You can allow views without a route_name attribute to match a route by adding the use_global_views flag to the route definition. For example, the myproject.views.bazbuz view below will be found if the route named abc below is matched and the PATH_INFO is /abc/bazbuz, even though the view configuration statement does not have the route_name="abc" attribute.

1
2
config.add_route('abc', '/abc/*traverse', use_global_views=True)
config.add_view('myproject.views.bazbuz', name='bazbuz')

Using *subpath in a Route Pattern

There are certain extremely rare cases when you’d like to influence the traversal subpath when a route matches without actually performing traversal. For instance, the pyramid.wsgi.wsgiapp2() decorator and the pyramid.view.static helper attempt to compute PATH_INFO from the request’s subpath, so it’s useful to be able to influence this value.

When *subpath exists in a pattern, no path is actually traversed, but the traversal algorithm will return a subpath list implied by the capture value of *subpath. You’ll see this pattern most commonly in route declarations that look like this:

1
2
config.add_route('static', '/static/*subpath',
                 view='mypackage.views.static_view')

Where mypackage.views.static_view is an instance of pyramid.view.static. This effectively tells the static helper to traverse everything in the subpath as a filename.

Corner Cases

A number of corner case “gotchas” exist when using a hybrid application. We’ll detail them here.

Registering a Default View for a Route That Has a view Attribute

It is an error to provide both a view argument to a route configuration and a view configuration which names a route_name that has no name value or the empty name value. For example, this pair of declarations will generate a “conflict” error at startup time.

1
2
3
config.add_route('home', '{foo}/{bar}/*traverse',
                 view='myproject.views.home')
config.add_view('myproject.views.another', route_name='home')

This is because the view argument to the add_route() above is an implicit default view when that route matches. add_route calls don’t need to supply a view attribute. For example, this add_route call:

1
2
config.add_route('home', '{foo}/{bar}/*traverse',
                 view='myproject.views.home')

Can also be spelled like so:

1
2
config.add_route('home', '{foo}/{bar}/*traverse')
config.add_view('myproject.views.home', route_name='home')

The two spellings are logically equivalent. In fact, the former is just a syntactical shortcut for the latter.

Binding Extra Views Against a Route Configuration that Doesn’t Have a *traverse Element In Its Pattern

Here’s another corner case that just makes no sense:

1
2
3
config.add_route('abc', '/abc', view='myproject.views.abc')
config.add_view('myproject.views.bazbuz', name='bazbuz',
                route_name='abc')

The above view declaration is useless, because it will never be matched when the route it references has matched. Only the view associated with the route itself (myproject.views.abc) will ever be invoked when the route matches, because the default view is always invoked when a route matches and when no post-match traversal is performed.

To make the above view declaration useful, the special *traverse token must end the route’s pattern. For example:

1
2
3
config.add_route('abc', '/abc/*traverse', view='myproject.views.abc')
config.add_view('myproject.views.bazbuz', name='bazbuz',
                route_name='abc')

With the above configuration, the myproject.views.bazbuz view will be invoked when the request URI is /abc/bazbuz, assuming there is no object contained by the root object with the key bazbuz. A different request URI, such as /abc/foo/bar, would invoke the default myproject.views.abc view.