How to Write a PluginΒΆ

This document is for people who want to write a plugin for Aspen. If you only want to use Aspen with existing plugins, then ... what?

Negotiated and rendered resources have content pages the bytes for which are transformed based on context. The user may explicitly choose a renderer per content page, with the default renderer per page computed from its media type. Template resources derive their media type from the file extension. Negotiated resources have no file extension by definition, so they specify the media type of their content pages in the resource itself, on the so-called “specline” of each content page, like so:

[---] text/plain
Greetings, program!
[---] text/html
<h1>Greetings, program!</h1>

A Renderer is instantiated by a Factory, which is a class that is itself instantied with one argument:

configuration   an Aspen configuration object

Instances of each Renderer subclass are callables that take five arguments and return a function (confused yet?). The five arguments are:

factory         the Factory creating this object
filepath        the filesystem path of the resource in question
raw             the bytestring of the page of the resource in question
media_type      the media type of the page
offset          the line number at which the page starts

Each Renderer instance is a callable that takes a context dictionary and returns a bytestring of rendered content. The heavy lifting is done in the render_content method.

Here’s how to implement and register your own renderer:

from aspen.simplates.renderers import Renderer, Factory

class Cheese(Renderer):
    def render_content(self, context):
        return self.raw.replace("cheese", "CHEESE!!!!!!")

class CheeseFactory(Factory):
    Renderer = Cheese

request_processor.renderer_factories['excited-about-cheese'] = CheeseFactory(request_processor)

Put that in your startup script. Now you can use it in a negotiated or rendered resource:

[---] via excited-about-cheese
I like cheese!

Out will come:

I like CHEESE!!!!!!!

If you write a new renderer for inclusion in the base Aspen distribution, please work with Aspen’s existing reloading machinery, etc. as much as possible. Use the existing template shims as guidelines, and if Aspen’s machinery is inadequate for some reason let’s evolve the machinery so all renderers behave consistently for users. Thanks.