django-medusa

Allows rendering a Django-powered website into a static website a la Jekyll, Movable Type, or other static page generation CMSes or frameworks. django-medusa is designed to be as simple as possible and allow the easy(ish) conversion of existing dynamic Django-powered websites -- nearly any existing Django site installation (not relying on highly-dynamic content) can be converted into a static generator which mirror's that site's output.

Given a "renderer" that defines a set of URLs (see below), this uses Django's built-in TestClient to render out those views to either disk, Amazon S3, or to Google App Engine.

At the moment, this likely does not scale to extremely large websites.

Optionally utilizes the multiprocessing library to speed up the rendering process by rendering many views at once.

For those uninterested in the nitty-gritty, there are tutorials/examples in the docs dir:

• Tutorial 1: Hello World

Renderer classes

Renderers live in renderers.py in each INSTALLED_APP.

Simply subclassing the StaticSiteRenderer class and defining get_paths works:

from django_medusa.renderers import StaticSiteRenderer

class HomeRenderer(StaticSiteRenderer):
    def get_paths(self):
        return frozenset([
            "/",
            "/about/",
            "/sitemap.xml",
        ])

renderers = [HomeRenderer, ]

A more complex example:

from django_medusa.renderers import StaticSiteRenderer
from myproject.blog.models import BlogPost


class BlogPostsRenderer(StaticSiteRenderer):
    def get_paths(self):
        paths = ["/blog/", ]

        items = BlogPost.objects.filter(is_live=True).order_by('-pubdate')
        for item in items:
            paths.append(item.get_absolute_url())

        return paths

renderers = [BlogPostsRenderer, ]

Or even:

from django_medusa.renderers import StaticSiteRenderer
from myproject.blog.models import BlogPost
from django.core.urlresolvers import reverse


class BlogPostsRenderer(StaticSiteRenderer):
    def get_paths(self):
        # A "set" so we can throw items in blindly and be guaranteed that
        # we don't end up with dupes.
        paths = set(["/blog/", ])

        items = BlogPost.objects.filter(is_live=True).order_by('-pubdate')
        for item in items:
            # BlogPost detail view
            paths.add(item.get_absolute_url())

            # The generic date-based list views.
            paths.add(reverse('blog:archive_day', args=(
                item.pubdate.year, item.pubdate.month, item.pubdate.day
            )))
            paths.add(reverse('blog:archive_month', args=(
                item.pubdate.year, item.pubdate.month
            )))
            paths.add(reverse('blog:archive_year', args=(item.pubdate.year,)))

        # Cast back to a list since that's what we're expecting.
        return list(paths)

renderers = [BlogPostsRenderer, ]

Renderer backends

Disk-based static site renderer

Example settings:

INSTALLED_APPS = (
    # ...
    # ...
    'django_medusa',
)
# ...
MEDUSA_RENDERER_CLASS = "django_medusa.renderers.DiskStaticSiteRenderer"
MEDUSA_MULTITHREAD = True
MEDUSA_DEPLOY_DIR = os.path.abspath(os.path.join(
    REPO_DIR,
    'var',
    "html"
))

S3-based site renderer

Example settings:

INSTALLED_APPS = (
    # ...
    # ...
    'django_medusa',
)
# ...
MEDUSA_RENDERER_CLASS = "django_medusa.renderers.S3StaticSiteRenderer"
MEDUSA_MULTITHREAD = True
AWS_ACCESS_KEY = ""
AWS_SECRET_ACCESS_KEY = ""
MEDUSA_AWS_STORAGE_BUCKET_NAME = "" # (also accepts AWS_STORAGE_BUCKET_NAME)

Be aware that the S3 renderer will overwrite any existing files that match URL paths in your site.

The S3 backend will force "index.html" to be the Default Root Object for each directory, so that "/about/" would actually be uploaded as "/about/index.html", but properly loaded by the browser at the "/about/" URL.

BONUS: Additionally, the S3 renderer keeps the "Content-Type" HTTP header that the view returns: if "/foo/json/" returns a JSON file (application/json), the file will be uploaded to "/foo/json/index.html" but will be served as application/json in the browser -- and will be accessible from "/foo/json/".

App Engine-based site renderer

Example settings:

INSTALLED_APPS = (
    # ...
    # ...
    'django_medusa',
)
# ...
MEDUSA_RENDERER_CLASS = "django_medusa.renderers.GAEStaticSiteRenderer"
MEDUSA_MULTITHREAD = True
MEDUSA_DEPLOY_DIR = os.path.abspath(os.path.join(
    REPO_DIR,
    'var',
    "html"
))
GAE_APP_ID = ""

This generates a app.yaml file and a deploy directory in your MEDUSA_DEPLOY_DIR. The app.yaml file contains the URL mappings to upload the entire site as a static files.

App Engine generally follows filename extensions as the mimetype. If you have paths that don't have an extension and are not HTML files (i.e. "/foo/json/", "/feeds/blog/", etc.), the mimetype from the "Content-Type" HTTP header will be manually defined for this URL in the app.yaml path.

Usage

1. Install django-medusa into your python path (TODO: setup.py) and add django_medusa to INSTALLED_APPS.
2. Select a renderer backend (currently: disk or s3) in your settings.
3. Create renderer classes in renderers.py under the apps you want to render.
4. django-admin.py staticsitegen
5. ???
6. Profit!

Example

From the first example in the "Renderer classes" section, using the disk-based backend.

$ django-admin.py staticsitegen
Found renderers for 'myproject'...
Skipping app 'django.contrib.syndication'... (No 'renderers.py')
Skipping app 'django.contrib.sitemaps'... (No 'renderers.py')
Skipping app 'typogrify'... (No 'renderers.py')

Generating with up to 8 processes...
/project_dir/var/html/index.html
/project_dir/var/html/about/index.html
/project_dir/var/html/sitemap.xml