.. _guide.routing:

URI routing
===========
`URI routing` is the process of taking the requested URI and deciding which
application handler will handle the current request. For this, we initialize
the :class:`WSGIApplication` defining a list of `routes`: each `route`
analyses the current request and, if it matches certain criteria, returns
the handler and optional variables extracted from the URI.

webapp2 offers a powerful and extensible system to match and build URIs,
which is explained in details in this section.


Simple routes
-------------
The simplest form of URI route in webapp2 is a tuple ``(regex, handler)``,
where `regex` is a regular expression to match the requested URI path and
`handler` is a callable to handle the request. This routing mechanism is
fully compatible with App Engine's webapp framework.

This is how it works: a list of routes is registered in the
:ref:`WSGI application <guide.app>`. When the application receives a request,
it tries to match each one in order until one matches, and then call the
corresponding handler. Here, for example, we define three handlers and
register three routes that point to those handlers::

    class HomeHandler(webapp2.RequestHandler):
        def get(self):
            self.response.write('This is the HomeHandler.')

    class ProductListHandler(webapp2.RequestHandler):
        def get(self):
            self.response.write('This is the ProductListHandler.')

    class ProductHandler(webapp2.RequestHandler):
        def get(self, product_id):
            self.response.write('This is the ProductHandler. '
                'The product id is %s' % product_id)

    app = webapp2.WSGIApplication([
        (r'/', HomeHandler),
        (r'/products', ProductListHandler),
        (r'/products/(\d+)', ProductHandler),
    ])

When a request comes in, the application will match the request path to find
the corresponding handler. If no route matches, an ``HTTPException`` is raised
with status code 404, and the WSGI application can handle it accordingly (see
:ref:`guide.exceptions`).

The `regex` part is an ordinary regular expression (see the :py:mod:`re`
module) that can define groups inside parentheses. The matched group values are
passed to the handler as positional arguments. In the example above, the last
route defines a group, so the handler will receive the matched value when the
route matches (one or more digits in this case).

Important note: If the route includes a regex group, all handler methods that
receive requests from that route must include a parameter for the positional
arguments; otherwise the application will attempt to pass an argument to the
handler with no matching parameter and an exception will occur.

The `handler` part is a callable as explained in :ref:`guide.handlers`, and
can also be a string in dotted notation to be lazily imported when needed
(see explanation below in :ref:`Lazy Handlers <guide.routing.lazy-handlers>`).

Simple routes are easy to use and enough for a lot of cases but don't support
keyword arguments, URI building, domain and subdomain matching, automatic
redirection and other useful features. For this, webapp2 offers the extended
routing mechanism that we'll see next.


Extended routes
---------------
webapp2 introduces a routing mechanism that extends the webapp model to provide
additional features:

- **URI building:** the registered routes can be built when needed, avoiding
  hardcoded URIs in the app code and templates. If you change the route
  definition in a compatible way during development, all places that use that
  route will continue to point to the correct URI. This is less error prone and
  easier to maintain.
- **Keyword arguments:** handlers can receive keyword arguments from the
  matched URIs. This is easier to use and also more maintainable than positional
  arguments.
- **Nested routes:** routes can be extended to match more than the request
  path. We will see below a route class that can also match domains and
  subdomains.

And several other features and benefits.

The concept is similar to the simple routes we saw before, but instead of a
tuple ``(regex, handler)``, we define each route using the class
:class:`webapp2.Route`. Let's remake our previous routes using it::

    app = webapp2.WSGIApplication([
        webapp2.Route(r'/', handler=HomeHandler, name='home'),
        webapp2.Route(r'/products', handler=ProductListHandler, name='product-list'),
        webapp2.Route(r'/products/<product_id:\d+>', handler=ProductHandler, name='product'),
    ])

The first argument in the routes above is a
:ref:`URL template <guide.routing.the-url-template>`, the `handler`
argument is the :ref:`request handler <guide.handlers>` to be used, and the
`name` argument third is a name used to
:ref:`build a URI <guide.routing.building-uris>` for that route.

Check :meth:`webapp2.Route.__init__` in the API reference for the parameters
accepted by the ``Route`` constructor. We will explain some of them in details
below.

.. _guide.routing.the-url-template:

The URL template
~~~~~~~~~~~~~~~~
The URL template defines the URL path to be matched. It can have regular
expressions for variables using the syntax ``<name:regex>``; everything
outside of ``<>`` is not interpreted as a regular expression to be matched.
Both name and regex are optional, like in the examples below:

=================  ==================================
Format             Example
=================  ==================================
``<name>``         ``'/blog/<year>/<month>'``
``<:regex>``       ``'/blog/<:\d{4}>/<:\d{2}>'``
``<name:regex>``   ``'/blog/<year:\d{4}>/<month:\d{2}>'``
=================  ==================================

The same template can mix parts with name, regular expression or both.

The name, if defined, is used to build URLs for the route. When it is set,
the value of the matched regular expression is passed as keyword argument to
the handler. Otherwise it is passed as positional argument.

If only the name is set, it will match anything except a slash. So these
routes are equivalent::

    Route('/<user_id>/settings', handler=SettingsHandler, name='user-settings')
    Route('/<user_id:[^/]+>/settings', handler=SettingsHandler, name='user-settings')

.. note::
   The handler only receives ``*args`` if no named variables are
   set. Otherwise, the handler only receives ``**kwargs``. This
   allows you to set regular expressions that are not captured:
   just mix named and unnamed variables and the handler will
   only receive the named ones.

.. _guide.routing.lazy-handlers:

Lazy handlers
~~~~~~~~~~~~~
One additional feature compared to webapp is that the handler can also be
defined as a string in dotted notation to be lazily imported when needed.

This is useful to avoid loading all modules when the app is initialized: we
can define handlers in different modules without needing to import all of them
to initialize the app. This is not only convenient but also speeds up the
application startup.

The string must contain the package or module name and the name of the handler
(a class or function name). Our previous example could be rewritten using
strings instead of handler classes and splitting our handlers in two files,
``handlers.py`` and ``products.py``::

    app = webapp2.WSGIApplication([
        (r'/', 'handlers.HomeHandler'),
        (r'/products', 'products.ProductListHandler'),
        (r'/products/(\d+)', 'products.ProductHandler'),
    ])

In the first time that one of these routes matches, the handlers will be
automatically imported by the routing system.

.. _guide.routing.custom-methods:

Custom methods
~~~~~~~~~~~~~~
A parameter ``handler_method`` can define the method of the handler that will
be called, if handler is a class. If not defined, the default behavior is to
translate the HTTP method to a handler method, as explained in
:ref:`guide.handlers`. For example::

    webapp2.Route(r'/products', handler='handlers.ProductsHandler', name='products-list', handler_method='list_products')

Alternatively, the handler method can be defined in the handler string,
separated by a colon. This is equivalent to the previous example::

    webapp2.Route(r'/products', handler='handlers.ProductsHandler:list_products', name='products-list')

.. _guide.routing.restricting-http-methods:

Restricting HTTP methods
~~~~~~~~~~~~~~~~~~~~~~~~
If needed, the route can define a sequence of allowed HTTP methods. Only if the
request method is in that list or tuple the route will match. If the method is
not allowed, an ``HTTPMethodNotAllowed`` exception is raised with status code
405. For example::

    webapp2.Route(r'/products', handler='handlers.ProductsHandler', name='products-list', methods=['GET'])

This is useful when using functions as handlers, or alternative handlers that
don't translate the HTTP method to the handler method like the default
:class:`webapp2.RequestHandler` does.

.. _guide.routing.restricting-uri-schemes:

Restricting URI schemes
~~~~~~~~~~~~~~~~~~~~~~~
Like with HTTP methods, you can specify the URI schemes allowed for a route,
if needed. This is useful if some URIs must be accessed using 'http' or 'https'
only. For this, set the ``schemes`` parameter when defining a route::

    webapp2.Route(r'/products', handler='handlers.ProductsHandler', name='products-list', schemes=['https'])

The above route will only match if the URI scheme is 'https'.


.. _guide.routing.domain-and-subdomain-routing:

Domain and subdomain routing
----------------------------
The routing system can also handle domain and subdomain matching. This is done
using a special route class provided in the :mod:`webapp2_extras.routes`
module: the :class:`webapp2_extras.routes.DomainRoute`. It is initialized with
a pattern to match the current server name and a list of nested
:class:`webapp2.Route` instances that will only be tested if the domain or
subdomain matches.

For example, to restrict routes to a subdomain of the appspot domain::

    import webapp2
    from webapp2_extras import routes

    app = webapp2.WSGIApplication([
        routes.DomainRoute('<subdomain>.app-id.appspot.com', [
            webapp2.Route('/', handler=SubdomainHomeHandler, name='subdomain-home'),
        ]),
        webapp2.Route('/', handler=HomeHandler, name='home'),
    ])

In the example above, we define a template ``'<subdomain>.app-id.appspot.com'``
for the domain matching. When a request comes in, only if the request server
name matches that pattern, the nested route will be tested. Otherwise the
routing system will test the next route until one matches. So the first route
with path ``/`` will only match when a subdomain of the ``app-id.appspot.com``
domain is accessed. Otherwise the second route with path ``/`` will be used.

The template follows the same syntax used by :class:`webapp2.Route` and
must define named groups if any value must be added to the match results.
In the example above, an extra `subdomain` keyword is passed to the handler,
but if the regex didn't define any named groups, nothing would be added.

Matching only www, or anything except www
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
A common need is to set some routes for the main subdomain (``www``) and
different routes for other subdomains. The webapp2 routing system can handle
this easily.

To match only the ``www`` subdomain, simple set the domain template to a fixed
value::

    routes.DomainRoute('www.mydomain.com', [
        webapp2.Route('/', handler=HomeHandler, name='home'),
    ])

To match any subdomain except the ``www`` subdomain, set a regular expression
that excludes ``www``::

    routes.DomainRoute(r'<subdomain:(?!www\.)[^.]+>.mydomain.com', [
        webapp2.Route('/', handler=HomeHandler, name='home'),
    ])

Any subdomain that matches and is not ``www`` will be passed as a parameter
``subdomain`` to the handler.

Similarly, you can restrict matches to the main ``appspot`` domain **or**
a ``www`` domain from a custom domain::

    routes.DomainRoute(r'<:(app-id\.appspot\.com|www\.mydomain\.com)>', [
        webapp2.Route('/', handler=HomeHandler, name='home'),
    ])

And then have a route that matches subdomains of the main ``appspot`` domain
**or** from a custom domain, except ``www``::

    routes.DomainRoute(r'<subdomain:(?!www)[^.]+>.<:(app-id\.appspot\.com|mydomain\.com)>', [
        webapp2.Route('/', handler=HomeHandler, name='home'),
    ])


.. _guide.routing.path-prefix-routes:

Path prefix routes
------------------
The :mod:`webapp2_extras.routes` provides a class to wrap routes that start
with a common path: the :mod:`webapp2_extras.routes.PathPrefixRoute`.
The intention is to avoid repetition when defining routes.

For example, imagine we have these routes::

    app = WSGIApplication([
        Route('/users/<user:\w+>/', UserOverviewHandler, 'user-overview'),
        Route('/users/<user:\w+>/profile', UserProfileHandler, 'user-profile'),
        Route('/users/<user:\w+>/projects', UserProjectsHandler, 'user-projects'),
    ])

We could refactor them to reuse the common path prefix::

    import webapp2
    from webapp2_extras import routes

    app = WSGIApplication([
        routes.PathPrefixRoute('/users/<user:\w+>', [
            webapp2.Route('/', UserOverviewHandler, 'user-overview'),
            webapp2.Route('/profile', UserProfileHandler, 'user-profile'),
            webapp2.Route('/projects', UserProjectsHandler, 'user-projects'),
        ]),
    ])

This is not only convenient, but also performs better: the nested routes
will only be tested if the path prefix matches.


.. _guide.routing.other-prefix-routes:

Other prefix routes
-------------------
The :mod:`webapp2_extras.routes` has other convenience classes that accept
nested routes with a common attribute prefix:

- :mod:`webapp2_extras.routes.HandlerPrefixRoute`: receives a handler module
  prefix in dotted notation and a list of routes that use that module.
- :mod:`webapp2_extras.routes.NamePrefixRoute`: receives a handler name
  prefix and a list of routes that start with that name.


.. _guide.routing.building-uris:

Building URIs
-------------
Because our routes have a ``name``, we can use the routing system to build
URIs whenever we need to reference those resources inside the application.
This is done using the function :func:`webapp2.uri_for` or the method
:meth:`webapp2.RequestHandler.uri_for` inside a handler, or calling
:meth:`webapp2.Router.build` directly (a ``Router`` instance is set as an
attribute ``router`` in the WSGI application).

For example, if you have these routes defined for the application::

    app = webapp2.WSGIApplication([
        webapp2.Route('/', handler='handlers.HomeHandler', name='home'),
        webapp2.Route('/wiki', handler=WikiHandler, name='wiki'),
        webapp2.Route('/wiki/<page>', handler=WikiHandler, name='wiki-page'),
    ])

Here are some examples of how to generate URIs for them::

    # /
    uri = uri_for('home')
    # http://localhost:8080/
    uri = uri_for('home', _full=True)
    # /wiki
    uri = uri_for('wiki')
    # http://localhost:8080/wiki
    uri = uri_for('wiki', _full=True)
    # http://localhost:8080/wiki#my-heading
    uri = uri_for('wiki', _full=True, _fragment='my-heading')
    # /wiki/my-first-page
    uri = uri_for('wiki-page', page='my-first-page')
    # /wiki/my-first-page?format=atom
    uri = uri_for('wiki-page', page='my-first-page', format='atom')

Variables are passed as positional or keyword arguments and are required if
the route defines them. Keyword arguments that are not present in the route
are added to the URI as a query string.

Also, when calling ``uri_for()``, a few keywords have special meaning:

_full
  If True, builds an absolute URI.
_scheme
  URI scheme, e.g., `http` or `https`. If defined, an absolute URI is always
  returned.
_netloc
  Network location, e.g., `www.google.com`. If defined, an absolute URI is
  always returned.
_fragment
  If set, appends a fragment (or "anchor") to the generated URI.

Check :meth:`webapp2.Router.build` in the API reference for a complete
explanation of the parameters used to build URIs.


Routing attributes in the request object
----------------------------------------
The parameters from the matched route are set as attributes of the request
object when a route matches. They are ``request.route_args``, for positional
arguments, and ``request.route_kwargs``, for keyword arguments.

The matched route object is also available as ``request.route``.