Controllers and Routing

Pecan uses a routing strategy known as object-dispatch to map an HTTP request to a controller, and then the method to call. Object-dispatch begins by splitting the path into a list of components and then walking an object path, starting at the root controller. You can imagine your application’s controllers as a tree of objects (branches of the object tree map directly to URL paths).

Let’s look at a simple bookstore application:

from pecan import expose

class BooksController(object):
    @expose()
    def index(self):
        return "Welcome to book section."

    @expose()
    def bestsellers(self):
        return "We have 5 books in the top 10."

class CatalogController(object):
    @expose()
    def index(self):
        return "Welcome to the catalog."

    books = BooksController()

class RootController(object):
    @expose()
    def index(self):
        return "Welcome to store.example.com!"

    @expose()
    def hours(self):
        return "Open 24/7 on the web."

    catalog = CatalogController()

A request for /catalog/books/bestsellers from the online store would begin with Pecan breaking the request up into catalog, books, and bestsellers. Next, Pecan would lookup catalog on the root controller. Using the catalog object, Pecan would then lookup books, followed by bestsellers. What if the URL ends in a slash? Pecan will check for an index method on the last controller object.

To illustrate further, the following paths:

└── /
    ├── /hours
    └── /catalog
         └── /catalog/books
            └── /catalog/books/bestsellers

route to the following controller methods:

└── RootController.index
    ├── RootController.hours
    └── CatalogController.index
         └── BooksController.index
            └── BooksController.bestsellers

Exposing Controllers

You tell Pecan which methods in a class are publically-visible via expose(). If a method is not decorated with expose(), Pecan will never route a request to it.

expose() can be used in a variety of ways. The simplest case involves passing no arguments. In this scenario, the controller returns a string representing the HTML response body.

from pecan import expose

class RootController(object):
    @expose()
    def hello(self):
        return 'Hello World'

A more common use case is to specify a template and a namespace:

from pecan import expose

class RootController(object):
    @expose('html_template.mako')
    def hello(self):
        return {'msg': 'Hello!'}
<!-- html_template.mako -->
<html>
    <body>${msg}</body>
</html>

Pecan also has built-in support for a special JSON renderer, which translates template namespaces into rendered JSON text:

from pecan import expose

class RootController(object):
    @expose('json')
    def hello(self):
        return {'msg': 'Hello!'}

expose() calls can also be stacked, which allows you to serialize content differently depending on how the content is requested:

from pecan import expose

class RootController(object):
    @expose('json')
    @expose('text_template.mako', content_type='text/plain')
    @expose('html_template.mako')
    def hello(self):
        return {'msg': 'Hello!'}

You’ll notice that we called expose() three times, with different arguments.

@expose('json')

The first tells Pecan to serialize the response namespace using JSON serialization when the client requests /hello.json or if an Accept: application/json header is present.

@expose('text_template.mako', content_type='text/plain')

The second tells Pecan to use the text_template.mako template file when the client requests /hello.txt or asks for text/plain via an Accept header.

@expose('html_template.mako')

The third tells Pecan to use the html_template.mako template file when the client requests /hello.html. If the client requests /hello, Pecan will use the text/html content type by default; in the absense of an explicit content type, Pecan assumes the client wants HTML.

Specifying Explicit Path Segments

Occasionally, you may want to use a path segment in your routing that doesn’t work with Pecan’s declarative approach to routing because of restrictions in Python’s syntax. For example, if you wanted to route for a path that includes dashes, such as /some-path/, the following is not valid Python:

class RootController(object):

    @pecan.expose()
    def some-path(self):
        return dict()

To work around this, pecan allows you to specify an explicit path segment in the expose() decorator:

class RootController(object):

    @pecan.expose(route='some-path')
    def some_path(self):
        return dict()

In this example, the pecan application will reply with an HTTP 200 for requests made to /some-path/, but requests made to /some_path/ will yield an HTTP 404.

route() can also be used explicitly as an alternative to the route argument in expose():

class RootController(object):

    @pecan.expose()
    def some_path(self):
        return dict()

pecan.route('some-path', RootController.some_path)

Routing to child controllers can be handled simliarly by utilizing route():

class ChildController(object):

    @pecan.expose()
    def child(self):
        return dict()

class RootController(object):
    pass

pecan.route(RootController, 'child-path', ChildController())

In this example, the pecan application will reply with an HTTP 200 for requests made to /child-path/child/.

Routing Based on Request Method

The generic argument to expose() provides support for overloading URLs based on the request method. In the following example, the same URL can be serviced by two different methods (one for handling HTTP GET, another for HTTP POST) using generic controllers:

from pecan import expose


class RootController(object):

    # HTTP GET /
    @expose(generic=True, template='json')
    def index(self):
        return dict()

    # HTTP POST /
    @index.when(method='POST', template='json')
    def index_POST(self, **kw):
        uuid = create_something()
        return dict(uuid=uuid)

Pecan’s Routing Algorithm

Sometimes, the standard object-dispatch routing isn’t adequate to properly route a URL to a controller. Pecan provides several ways to short-circuit the object-dispatch system to process URLs with more control, including the special _lookup(), _default(), and _route() methods. Defining these methods on your controller objects provides additional flexibility for processing all or part of a URL.

Routing to Subcontrollers with _lookup

The _lookup() special method provides a way to process a portion of a URL, and then return a new controller object to route to for the remainder.

A _lookup() method may accept one or more arguments, segments of the URL path to be processed (split on /). _lookup() should also take variable positional arguments representing the rest of the path, and it should include any portion of the path it does not process in its return value. The example below uses a *remainder list which will be passed to the returned controller when the object-dispatch algorithm continues.

In addition to being used for creating controllers dynamically, _lookup() is called as a last resort, when no other controller method matches the URL and there is no _default() method.

from pecan import expose, abort
from somelib import get_student_by_name

class StudentController(object):
    def __init__(self, student):
        self.student = student

    @expose()
    def name(self):
        return self.student.name

class RootController(object):
    @expose()
    def _lookup(self, primary_key, *remainder):
        student = get_student_by_primary_key(primary_key)
        if student:
            return StudentController(student), remainder
        else:
            abort(404)

An HTTP GET request to /8/name would return the name of the student where primary_key == 8.

Falling Back with _default

The _default() method is called as a last resort when no other controller methods match the URL via standard object-dispatch.

from pecan import expose

class RootController(object):
    @expose()
    def english(self):
        return 'hello'

    @expose()
    def french(self):
        return 'bonjour'

    @expose()
    def _default(self):
        return 'I cannot say hello in that language'

In the example above, a request to /spanish would route to RootController._default().

Defining Customized Routing with _route

The _route() method allows a controller to completely override the routing mechanism of Pecan. Pecan itself uses the _route() method to implement its RestController. If you want to design an alternative routing system on top of Pecan, defining a base controller class that defines a _route() method will enable you to have total control.

Interacting with the Request and Response Object

For every HTTP request, Pecan maintains a thread-local reference to the request and response object, pecan.request and pecan.response. These are instances of pecan.Request and pecan.Response, respectively, and can be interacted with from within Pecan controller code:

@pecan.expose()
def login(self):
    assert pecan.request.path == '/login'
    username = pecan.request.POST.get('username')
    password = pecan.request.POST.get('password')

    pecan.response.status = 403
    pecan.response.text = 'Bad Login!'

While Pecan abstracts away much of the need to interact with these objects directly, there may be situations where you want to access them, such as:

  • Inspecting components of the URI
  • Determining aspects of the request, such as the user’s IP address, or the referer header
  • Setting specific response headers
  • Manually rendering a response body

Specifying a Custom Response

Set a specific HTTP response code (such as 203 Non-Authoritative Information) by modifying the status attribute of the response object.

from pecan import expose, response

class RootController(object):

    @expose('json')
    def hello(self):
        response.status = 203
        return {'foo': 'bar'}

Use the utility function abort() to raise HTTP errors.

from pecan import expose, abort

class RootController(object):

    @expose('json')
    def hello(self):
        abort(404)

abort() raises an instance of WSGIHTTPException which is used by Pecan to render default response bodies for HTTP errors. This exception is stored in the WSGI request environ at pecan.original_exception, where it can be accessed later in the request cycle (by, for example, other middleware or Custom Error Documents).

If you’d like to return an explicit response, you can do so using Response:

from pecan import expose, Response

class RootController(object):

    @expose()
    def hello(self):
        return Response('Hello, World!', 202)

Extending Pecan’s Request and Response Object

The request and response implementations provided by WebOb are powerful, but at times, it may be useful to extend application-specific behavior onto your request and response (such as specialized parsing of request headers or customized response body serialization). To do so, define custom classes that inherit from pecan.Request and pecan.Response, respectively:

class MyRequest(pecan.Request):
    pass

class MyResponse(pecan.Response):
    pass

and modify your application configuration to use them:

from myproject import MyRequest, MyResponse

app = {
    'root' : 'project.controllers.root.RootController',
    'modules' : ['project'],
    'static_root'   : '%(confdir)s/public',
    'template_path' : '%(confdir)s/project/templates',
    'request_cls': MyRequest,
    'response_cls': MyResponse
}

Mapping Controller Arguments

In Pecan, HTTP GET and POST variables that are not consumed during the routing process can be passed onto the controller method as arguments.

Depending on the signature of the method, these arguments can be mapped explicitly to arguments:

from pecan import expose

class RootController(object):
    @expose()
    def index(self, arg):
        return arg

    @expose()
    def kwargs(self, **kwargs):
        return str(kwargs)
$ curl http://localhost:8080/?arg=foo
foo
$ curl http://localhost:8080/kwargs?a=1&b=2&c=3
{u'a': u'1', u'c': u'3', u'b': u'2'}

or can be consumed positionally:

from pecan import expose

class RootController(object):
    @expose()
    def args(self, *args):
        return ','.join(args)
$ curl http://localhost:8080/args/one/two/three
one,two,three

The same effect can be achieved with HTTP POST body variables:

from pecan import expose

class RootController(object):
    @expose()
    def index(self, arg):
        return arg
$ curl -X POST "http://localhost:8080/" -H "Content-Type: application/x-www-form-urlencoded" -d "arg=foo"
foo

Static File Serving

Because Pecan gives you direct access to the underlying Request, serving a static file download is as simple as setting the WSGI app_iter and specifying the content type:

import os
from random import choice

from webob.static import FileIter

from pecan import expose, response


class RootController(object):

    @expose(content_type='image/gif')
    def gifs(self):
        filepath = choice((
            "/path/to/funny/gifs/catdance.gif",
            "/path/to/funny/gifs/babydance.gif",
            "/path/to/funny/gifs/putindance.gif"
        ))
        f = open(filepath, 'rb')
        response.app_iter = FileIter(f)
        response.headers[
            'Content-Disposition'
        ] = 'attachment; filename="%s"' % os.path.basename(f.name)

If you don’t know the content type ahead of time (for example, if you’re retrieving files and their content types from a data store), you can specify it via response.headers rather than in the expose() decorator:

import os
from mimetypes import guess_type

from webob.static import FileIter

from pecan import expose, response


class RootController(object):

    @expose()
    def download(self):
        f = open('/path/to/some/file', 'rb')
        response.app_iter = FileIter(f)
        response.headers['Content-Type'] = guess_type(f.name)
        response.headers[
            'Content-Disposition'
        ] = 'attachment; filename="%s"' % os.path.basename(f.name)

Handling File Uploads

Pecan makes it easy to handle file uploads via standard multipart forms. Simply define your form with a file input:

<form action="/upload" method="POST" enctype="multipart/form-data">
  <input type="file" name="file" />
  <button type="submit">Upload</button>
</form>

You can then read the uploaded file off of the request object in your application’s controller:

from pecan import expose, request

class RootController(object):
    @expose()
    def upload(self):
        assert isinstance(request.POST['file'], cgi.FieldStorage)
        data = request.POST['file'].file.read()

Thread-Safe Per-Request Storage

For convenience, Pecan provides a Python dictionary on every request which can be accessed and modified in a thread-safe manner throughout the life-cycle of an individual request:

pecan.request.context['current_user'] = some_user
print pecan.request.context.items()

This is particularly useful in situations where you want to store metadata/context about a request (e.g., in middleware, or per-routing hooks) and access it later (e.g., in controller code).

For more fine-grained control of the request, the underlying WSGI environ for a given Pecan request can be accessed and modified via pecan.request.environ.

Helper Functions

Pecan also provides several useful helper functions for moving between different routes. The redirect() function allows you to issue internal or HTTP 302 redirects.

See also

The redirect() utility, along with several other useful helpers, are documented in pecan.core – Pecan Core.

Determining the URL for a Controller

Given the ability for routing to be drastically changed at runtime, it is not always possible to correctly determine a mapping between a controller method and a URL.

For example, in the following code that makes use of _lookup() to alter the routing depending on a condition:

from pecan import expose, abort
from somelib import get_user_region


class DefaultRegionController(object):

    @expose()
    def name(self):
        return "Default Region"

class USRegionController(object):

    @expose()
    def name(self):
        return "US Region"

class RootController(object):
    @expose()
    def _lookup(self, user_id, *remainder):
        if get_user_region(user_id) == 'us':
            return USRegionController(), remainder
        else:
            return DefaultRegionController(), remainder

This logic depends on the geolocation of a given user and returning a completely different class given the condition. A helper to determine what URL USRegionController.name belongs to would fail to do it correctly.