Using WhiteNoise with any WSGI application¶

Note

These instructions apply to any WSGI application. However, for Django applications you would be better off using the WhiteNoiseMiddleware class which makes integration easier.

To enable WhiteNoise you need to wrap your existing WSGI application in a WhiteNoise instance and tell it where to find your static files. For example:

from whitenoise import WhiteNoise

from my_project import MyWSGIApp

application = MyWSGIApp()
application = WhiteNoise(application, root="/path/to/static/files")
application.add_files("/path/to/more/static/files", prefix="more-files/")

On initialization, WhiteNoise walks over all the files in the directories that have been added (descending into sub-directories) and builds a list of available static files. Any requests which match a static file get served by WhiteNoise, all others are passed through to the original WSGI application.

See the sections on compression and caching for further details.

WhiteNoise API¶

class WhiteNoise(application, root=None, prefix=None, \**kwargs)¶

Parameters:

application (callable) – Original WSGI application
root (str) – If set, passed to add_files method
prefix (str) – If set, passed to add_files method
**kwargs – Sets configuration attributes for this instance

WhiteNoise.add_files(root, prefix=None)¶

Parameters:

root (str) – Absolute path to a directory of static files to be served
prefix (str) – If set, the URL prefix under which the files will be served. Trailing slashes are automatically added.

Compression Support¶

When WhiteNoise builds its list of available files it checks for corresponding files with a .gz and a .br suffix (e.g., scripts/app.js, scripts/app.js.gz and scripts/app.js.br). If it finds them, it will assume that they are (respectively) gzip and brotli compressed versions of the original file and it will serve them in preference to the uncompressed version where clients indicate that they that compression format (see note on Amazon S3 for why this behaviour is important).

WhiteNoise comes with a command line utility which will generate compressed versions of your files for you. Note that in order for brotli compression to work the Brotli Python package must be installed.

Usage is simple:

$ python -m whitenoise.compress --help
usage: compress.py [-h] [-q] [--no-gzip] [--no-brotli]
                   root [extensions [extensions ...]]

Search for all files inside <root> *not* matching <extensions> and produce
compressed versions with '.gz' and '.br' suffixes (as long as this results in
a smaller file)

positional arguments:
  root         Path root from which to search for files
  extensions   File extensions to exclude from compression (default: jpg,
               jpeg, png, gif, webp, zip, gz, tgz, bz2, tbz, xz, br, swf, flv,
               woff, woff2)

optional arguments:
  -h, --help   show this help message and exit
  -q, --quiet  Don't produce log output
  --no-gzip    Don't produce gzip '.gz' files
  --no-brotli  Don't produce brotli '.br' files

You can either run this during development and commit your compressed files to your repository, or you can run this as part of your build and deploy processes. (Note that this is handled automatically in Django if you’re using the custom storage backend.)

Caching Headers¶

By default, WhiteNoise sets a max-age header on all responses it sends. You can configure this by passing a max_age keyword argument.

WhiteNoise sets both Last-Modified and ETag headers for all files and will return Not Modified responses where appropriate. The ETag header uses the same format as nginx which is based on the size and last-modified time of the file. If you want to use a different scheme for generating ETags you can set them via you own function by using the add_headers_function option.

Most modern static asset build systems create uniquely named versions of each file. This results in files which are immutable (i.e., they can never change their contents) and can therefore by cached indefinitely. In order to take advantage of this, WhiteNoise needs to know which files are immutable. This can be done using the immutable_file_test option which accepts a reference to a function.

The exact details of how you implement this method will depend on your particular asset build system but see the option documentation for a simple example.

Once you have implemented this, any files which are flagged as immutable will have “cache forever” headers set.

Index Files¶

When the index_file option is enabled:

Visiting /example/ will serve the file at /example/index.html
Visiting /example will redirect (302) to /example/
Visiting /example/index.html will redirect (302) to /example/

If you want to something other than index.html as the index file, then you can also set this option to an alternative filename.

Using a Content Distribution Network¶

See the instructions for using a CDN with Django . The same principles apply here although obviously the exact method for generating the URLs for your static files will depend on the libraries you’re using.

Redirecting to HTTPS¶

WhiteNoise does not handle redirection itself, but works well alongside wsgi-sslify, which performs HTTP to HTTPS redirection as well as optionally setting an HSTS header. Simply wrap the WhiteNoise WSGI application with sslify() - see the wsgi-sslify documentation for more details.

Configuration attributes¶

These can be set by passing keyword arguments to the constructor, or by sub-classing WhiteNoise and setting the attributes directly.

autorefresh¶

Default:: False

Recheck the filesystem to see if any files have changed before responding. This is designed to be used in development where it can be convenient to pick up changes to static files without restarting the server. For both performance and security reasons, this setting should not be used in production.

max_age¶

Default:: 60

Time (in seconds) for which browsers and proxies should cache files.

The default is chosen to be short enough not to cause problems with stale versions but long enough that, if you’re running WhiteNoise behind a CDN, the CDN will still take the majority of the strain during times of heavy load.

Set to None to disable setting any Cache-Control header on non-versioned files.

index_file¶

Default:: False

If True enable index file serving. If set to a non-empty string, enable index files and use that string as the index file name.

mimetypes¶

Default:: None

A dictionary mapping file extensions (lowercase) to the mimetype for that extension. For example:

{'.foo': 'application/x-foo'}

Note that WhiteNoise ships with its own default set of mimetypes and does not use the system-supplied ones (e.g. /etc/mime.types). This ensures that it behaves consistently regardless of the environment in which it’s run. View the defaults in the media_types.py file.

In addition to file extensions, mimetypes can be specified by supplying the entire filename, for example:

{'some-special-file': 'application/x-custom-type'}

charset¶

Default:: utf-8

Charset to add as part of the Content-Type header for all files whose mimetype allows a charset.

allow_all_origins¶

Default:: True

Toggles whether to send an Access-Control-Allow-Origin: * header for all static files.

This allows cross-origin requests for static files which means your static files will continue to work as expected even if they are served via a CDN and therefore on a different domain. Without this your static files will mostly work, but you may have problems with fonts loading in Firefox, or accessing images in canvas elements, or other mysterious things.

The W3C explicitly state that this behaviour is safe for publicly accessible files.

add_headers_function¶

Default:: None

Reference to a function which is passed the headers object for each static file, allowing it to modify them.

For example:

def force_download_pdfs(headers, path, url):
    if path.endswith('.pdf'):
        headers['Content-Disposition'] = 'attachment'

application = WhiteNoise(application,
                         add_headers_function=force_download_pdfs)

The function is passed:

headers: A wsgiref.headers instance (which you can treat just as a dict) containing the headers for the current file
path: The absolute path to the local file
url: The host-relative URL of the file e.g. /static/styles/app.css

The function should not return anything; changes should be made by modifying the headers dictionary directly.

immutable_file_test¶

Default:: return False

Reference to function, or string.

If a reference to a function, this is passed the path and URL for each static file and should return whether that file is immutable, i.e. guaranteed not to change, and so can be safely cached forever.

If a string, this is treated as a regular expression and each file’s URL is matched against it.

Example:

def immutable_file_test(path, url):
    # Match filename with 12 hex digits before the extension
    # e.g. app.db8f2edc0c8a.js
    return re.match(r'^.+\.[0-9a-f]{12}\..+$', url)

The function is passed:

path: The absolute path to the local file
url: The host-relative URL of the file e.g. /static/styles/app.css