.. _guide.app:

The WSGI application
====================
The WSGI application receives requests and dispatches the appropriate handler,
returning a response to the client. It stores the URI routes that the app will
accept, configuration variables and registered objects that can be shared
between requests. The WSGI app is also responsible for handling uncaught
exceptions, avoiding that stack traces "leak" to the client when in production.
Let's take an in depth look at it now.

.. note::
   If the WSGI word looks totally unfamiliar to you, read the
   `Another Do-It-Yourself Framework`_ tutorial by Ian Bicking. It is a very
   recommended introduction to WSGI and you should at least take a quick look
   at the concepts, but following the whole tutorial is really worth.

   A more advanced reading is the WSGI specification described in the
   `PEP 333 <http://www.python.org/dev/peps/pep-0333/>`_.


Initialization
--------------
The :class:`webapp2.WSGIApplication` class is initialized with three optional
arguments:

- ``routes``: a list of route definitions as described in :ref:`guide.routing`.
- ``debug``: a boolean flag that enables debug mode.
- ``config``: a dictionary of configuration values for the application.

Compared to webapp, only config was added; it is used as a standard way to
configure extra modules (sessions, internationalization, templates or your
own app configuration values).

Everything is pretty straighforward::

    import webapp2

    routes = [
        (r'/', 'handlers.HelloWorldHandler'),
    ]

    config = {}
    config['webapp2_extras.sessions'] = {
        'secret_key': 'something-very-very-secret',
    }

    app = webapp2.WSGIApplication(routes=routes, debug=True, config=config)


.. _guide.app.router:

Router
------
:ref:`guide.routing` is a central piece in webapp2, and its main component is
the :class:`webapp2.Router` object, available in the application as the
:attr:`webapp2.WSGIApplication.router` attribute.

The router object is responsible for everything related to mapping URIs to
handlers. The router:

- Stores registered "routes", which map URIs to the application handlers
  that will handle those requests.
- Matches the current request against the registered routes and returns the
  handler to be used for that request (or raises a ``HTTPNotFound`` exception
  if no handler was found).
- Dispatches the matched handler, i.e., calling it and returning a response
  to the ``WSGIApplication``.
- Builds URIs for the registered routes.

Using the ``router`` attribute you can, for example, add new routes to the
application after initialization using the ``add()`` method::

    import webapp2

    app = webapp2.WSGIApplication()
    app.router.add((r'/', 'handlers.HelloWorldHandler'))

The router has several methods to override how URIs are matched or built or how
handlers are adapted or dispatched without even requiring subclassing. For an
example of extending the default dispatching mechanism, see
:ref:`Request handlers: returned values <guide.handlers.returned_values>`.

Also check the :class:`Router API documentation <webapp2.Router>` for
a description of the methods :meth:`webapp2.Router.set_matcher`,
:meth:`webapp2.Router.set_dispatcher`, :meth:`webapp2.Router.set_adapter` and
:meth:`webapp2.Router.set_builder`.


.. _guide.app.config:

Config
------
When instantiating the app, you can pass a configuration dictionary which is
then accessible through the :attr:`webapp2.WSGIApplication.config` attribute.
A convention is to define configuration keys for each module, to avoid name
clashes, but you can define them as you wish, really, unless the module
requires a specific setup. First you define a configuration::

    import webapp2

    config = {'foo': 'bar'}

    app = webapp2.WSGIApplication(routes=[
        (r'/', 'handlers.MyHandler'),
    ], config=config)

Then access it as you need. Inside a ``RequestHandler``, for example::

    import webapp2

    class MyHandler(webapp2.RequestHandler):
        def get(self):
            foo = self.app.config.get('foo')
            self.response.write('foo value is %s' % foo)


.. _guide.app.registry:

Registry
--------
A simple dictionary is available in the application to register instances that
are shared between requests: it is the :attr:`webapp2.WSGIApplication.registry`
attribute. It can be used by anything that your app requires and the intention
is to avoid global variables in modules, so that you can have multiple app
instances using different configurations: each app has its own extra instances
for any kind of object that is shared between requests. A simple example that
registers a fictitious ``MyParser`` instance if it is not yet registered::

    import webapp2

    def get_parser():
        app = webapp2.get_app()
        # Check if the instance is already registered.
        my_parser = app.registry.get('my_parser')
        if not my_parser:
            # Import the class lazily.
            cls = webapp2.import_string('my.module.MyParser')
            # Instantiate the imported class.
            my_parser = cls()
            # Register the instance in the registry.
            app.registry['my_parser'] = my_parser

        return my_parser

The registry can be used to lazily instantiate objects when needed, and keep a
reference in the application to be reused.

A registry dictionary is also available in the
:ref:`request object <guide.request.registry>`, to store shared objects
used during a single request.


Error handlers
--------------
As described in :ref:`guide.exceptions`, a dictionary is available in the app
to register error handlers as the :attr:`webapp2.WSGIApplication.error_handlers`
attribute. They will be used as a last resource if exceptions are not caught
by handlers. It is a good idea to set at least error handlers for 404 and 500
status codes::

    import logging

    import webapp2

    def handle_404(request, response, exception):
        logging.exception(exception)
        response.write('Oops! I could swear this page was here!')
        response.set_status(404)

    def handle_500(request, response, exception):
        logging.exception(exception)
        response.write('A server error occurred!')
        response.set_status(500)

    app = webapp2.WSGIApplication([
        webapp2.Route(r'/', handler='handlers.HomeHandler', name='home')
    ])
    app.error_handlers[404] = handle_404
    app.error_handlers[500] = handle_500


Debug flag
----------
A debug flag is passed to the WSGI application on instantiation and is
available as the :attr:`webapp2.WSGIApplication.debug` attribute. When in
debug mode, any exception that is now caught is raised and the stack trace is
displayed to the client, which helps debugging. When not in debug mode, a
'500 Internal Server Error' is displayed instead.

You can use that flag to set special behaviors for the application during
development.

For App Engine, it is possible to detect if the code is running using the SDK
or in production checking the 'SERVER_SOFTWARE' environ variable::

    import os

    import webapp2

    debug = os.environ.get('SERVER_SOFTWARE', '').startswith('Dev')

    app = webapp2.WSGIApplication(routes=[
        (r'/', 'handlers.HelloWorldHandler'),
    ], debug=debug)


Thread-safe application
-----------------------
By default, webapp2 is thread-safe when the module
:class:`webapp2_extras.local` is available. This means that it can be used
outside of App Engine or in the upcoming App Engine Python 2.7 runtime.
This also works in non-threaded environments such as App Engine Python 2.5.

See in the :ref:`tutorials.quickstart.nogae` tutorial an explanation on how
to use webapp2 outside of App Engine.


Running the app
---------------
The application is executed in a CGI environment using the method
:meth:`webapp2.WSGIApplication.run`. When using App Engine, it uses
the functions ``run_bare_wsgi_app`` or ``run_wsgi_app`` from
``google.appengine.ext.webapp.util``. Outside of App Engine, it uses the
:py:mod:`wsgiref.handlers` module. Here's the simplest example::

    import webapp2

    class HelloWebapp2(webapp2.RequestHandler):
        def get(self):
            self.response.write('Hello, webapp2!')

    app = webapp2.WSGIApplication([
        ('/', HelloWebapp2),
    ], debug=True)

    def main():
        app.run()

    if __name__ == '__main__':
        main()


Unit testing
------------
As described in :ref:`guide.testing`, the application has a convenience method
to test handlers: :meth:`webapp2.WSGIApplication.get_response`. It
receives the same parameters as ``Request.blank()`` to build a request and call
the application, returning the resulting response from a handler::

    class HelloHandler(webapp2.RequestHandler):
        def get(self):
            self.response.write('Hello, world!')

    app = webapp2.WSGIapplication([('/', HelloHandler)])

    # Test the app, passing parameters to build a request.
    response = app.get_response('/')
    assert response.status_int == 200
    assert response.body == 'Hello, world!'


Getting the current app
-----------------------
The active ``WSGIApplication`` instance can be accessed at any place of your
app using the function :func:`webapp2.get_app`. This is useful, for example, to
access the app registry or configuration values::

    import webapp2

    app = webapp2.get_app()
    config_value = app.config.get('my-config-key')


.. _Another Do-It-Yourself Framework: http://docs.webob.org/en/latest/do-it-yourself.html