Ernestas Poškus

Technical blog

"We must view with profound respect the infinite capacity of the human mind to resist the introduction of useful knowledge." - Thomas R. Lounsbury

| github | goodreads | linkedin | twitter |

ansible 2 / elasticsearch 2 / kernel 2 / leadership 1 / linux 2 / mnemonics 1 / nginx 1 / paper 40 / personal 5 / rust 1 / tools 2 /

Serving fresh assets using Nginx location rewrite

WC 425 / RT 2min


Recently I have stumbled upon a problem to serve fresh/new assets for user web application.

As Phil Karlton said:

There are only two hard things in Computer Science: cache invalidation and naming things.

Historically fresh assets problem was approached either by appending appending url query params (?v=20130102) or renaming/hashing asset file completely (/css/default-2j9alkjan2k2.css).

Former is most popular one but not elegant since it brings explicit dependency for backend application what fresh/new asset file to include thus requires exact name file to be present on web server.

This draws 5 main disadvantages of completely hashed asset name:

  1. explicit dependency to include
  2. no fallback mechanism
  3. hashed asset name
  4. exact file name presence on web server
  5. removal of stale assets

I came with solution to use Nginx rewrite block that implicitly drops hash of requested file and serves requested asset.

   location @css_assets {
       rewrite ^/css/(.*)\..*\.(.*)$ /css/$1.$2 last;
   }
   location /css/ {
       try_files $uri $uri/ @css_assets;
   }

First location /css/ { block matches path of /css/ which later executes try_files followed by location @css_assets { location block.

Secondly this rewrite rewrite ^/css/(.*)\..*\.(.*)$ /css/$1.$2 last; matches beginning of /css/ path followed by 2 tracked matches.

/css/app.117c7f2fa4b6ea7a2c077a3dbc9662e6b1c278bd.css

In above example first match tracks (app) and second one (css). Matched information constructs new implicitly requested file like below.

/css/app.css

Newly constructed file will be processed by Nginx without redirects and browser knowing original file name.

Constructing asset hash

To tell your application which asset must be served use ENVIRONMENT variable and checksum of asset to be included. Or you can dynamically invalidate/create asset hash for example hourly or daily depending on release cycle.

/css/app.$CSS_ASSET_HASH.css

In my case I use simple function below.

lazy_static! {
    static ref CSS_ASSETS_HASH: String = {
        match env::var("CSS_ASSETS_HASH") {
            Ok(hash) => format!(".{}.", hash),
            Err(_) => ".".to_string(),
        }
    };
}

html! {
  (Css(format!("/css/app{}css", *CSS_ASSETS_HASH)))
}

Above example is actual code used in production. It tries to statically initialize CSS_ASSETS_HASH variable, if expected environment is not defined it fallbacks to dot . else it appends 2 dots between supplied environment variable .ENVIRONMENT_VARIABLE..

My solution eliminates almost all of main disadvantages of most popular way of asset inclusion.

  1. explicit dependency to include - depends if hash is generated dynamically or real checksum of asset file is used.
  2. no fallback mechanism - if environment is not defined or any hash is supplied it still fallbacks to original requested asset due to regex catchall.
  3. hashed asset name - asset name is explicit and easily understood
  4. exact file name presence on web server
  5. removal of stale assets - only original file is deployed

Regards.