Liking cljdoc? Tell your friends :D

Stasis Powerpack

A batteries-included Clojure toolkit for building static web sites. It combines the excellent Stasis with some useful tools, and provides the wiring so you can focus on coding your pages - exactly the way you want.

Install

With tools.deps:

no.cjohansen/powerpack {:mvn/version "2024.11.26"}

With Leiningen:

[no.cjohansen/powerpack "2024.11.26"]

Status

Powerpack is based on several sites built on roughly the same architecture over a 5 year period. The specific Powerpack code-base has been developed alongside a 5000 page static page and has been thoroughly put through its paces. APIs are stable, development experience is top notch, and the production export is truly production ready. Use at will!

What does it do?

Stasis Powerpack reads your raw content from Markdown (including mapdown) and EDN files (by default, you can include more file types as needed) into an in-memory Datomic database, and configures the web server with an asset optimization pipeline, image manipulation, live reload during development, and optionally more stuff - like code syntax highlighting.

Powerpack doesn't have any opinions on how you construct pages, there are no themes or design. It's still a toolkit, but the overall dataflow is lined up for you, and the web server is tuned well for both development and production. It comes with an export function that exports all your pages, as well as images and assets, ready for deployment to your serverless static web host of choice.

What's in the box?

Stasis Powerpack combines the following libraries:

Stasis - static web site toolkit
Optimus - asset optimization pipeline
Datomic - relational database
Imagine - image manipulation (crop, resize, greyscale, etc)
m1p - i18n
Mapdown - key/value markdown files
Beholder - file watching (live reload)

Powerpack data flow overview

In Powerpack, you put content in Markdown files and EDN files. These are read and transacted into an in-memory Datomic database. Powerpack defines a few schema attributes it needs to address pages, but other than that you'll need to define your own schema and implement some mapping to prepare raw content for the database. The files can contain whatever you want -- supporting data, a single web page, multiple web pages -- it's up to you.

Powerpack will allow browsing to any entity that has a :page/uri. You must add these attributes to any content you want to be browsable.

When you request a URL from the development server, Powerpack will find the page in the database and call on your code to render a response. This response can be any valid ring response. Powerpack provides some convenience library functions to help you build HTML pages, but you're free to ignore them. Before serving the page to the user, Powerpack will perform some post processing on the pages.

Powerpack currently only post processes HTML pages. When it does, images will be optimized, or passed through Imagine, open graph URLs will be qualified with your production domain name, code blocks will be highlighted, etc. Powerpack has an extension point for this post processing.

Getting started

To build a site with Powerpack, you must provide some configuration, a Datomic schema and a function that renders pages. There is a comprehensive step-by-step getting started in a separate repo.

Configuration reference

`:site/title` (required)

The site title, a string. Used for the default title tag.

`:site/default-locale`

Defaults to :en.

The default locale for pages, a keyword. When the current page has no :page/locale, this is used for i18n lookups, and as the lang attribute on the html element.

`:site/base-url`

Your production URL. This is used to qualify open graph URLs.

`:datomic/schema-file` (required)

Defaults to resources/schema.edn.

The file that contains the Datomic schema.

`:powerpack/content-dir`

Defaults to content.

The directory from which to read content files.

`:powerpack/build-dir`

Defaults to target/powerpack.

The directory to export the site to.

`:powerpack/render-page`

A function that is called to render a page. See rendering pages.

`:datomic/uri`

Defaults to "datomic:mem://powerpack".

The Datomic database URI.

`:imagine/config`

Image transformation configuration, refer to the Imagine Readme.

`:m1p/dictionaries`

A map of {locale dictionary-files} for i18n dictionaries. The map should be keyed by keyword locales, e.g. :en, and the value should be a vector of paths to EDN files with dictionaries, e.g.:

:m1p/dictionaries {:en ["src/powerblog/pages/en.edn"
                        "src/powerblog/sections/en.edn"]}

Refer to the m1p Readme for information about dictionaries.

`:m1p/dictionary-fns`

Additional m1p dictionary functions.

`:m1p/k`

The keyword used for i18n interpolation. Defaults to :i18n, as in

[:h1 [:i18n ::greeting]]

`:optimus/assets`

A vector of Optimus asset configurations. The vector should contain one or more maps with the same options you would pass to Optimus' load-assets, e.g.:

:optimus/assets [{:public-dir "public"
                  :paths ["/images/logo.png"
                          "/images/photo.jpg"]]

`:optimus/bundles`

A map of Optimus bundle configurations. During HTML post processing, CSS bundles will be added to the head and JavaScript bundles will be added to the end of body.

Example:

:optimus/bundles {"app.css"
                  {:public-dir "public"
                   :paths ["/styles.css"]}}

`:optimus/options`

Options passed to Optimus optimizations. Can aditionally contain the following keys to control which optimizations are used:

:minify-css-assets? defaults to true
:minify-js-assets? defaults to true

`:powerpack/asset-targets`

A list of selectors for where Powerpack should look for asset URLs in your rendered HTML pages. Powerpack uses this information to replace asset paths like /preview-small/images/climbing.jpg with a cache-busted version such as /image-assets/preview-small/fb6a746aee13f753872432da49c32a1cd019a334/images/climbing.jpg.

The list contains maps of {:selector :attr} where :selector is a CSS selector to find relevant nodes, and :attr is the string name of the attribute that may contain asset URLs.

Example:

[{:selector ["img[src]"]
  :attr "src"}
 ,,,
 ]

During HTML post processing, this is used to find all img elements that has a src attribute, and optimize the path in that attribute if applicable.

The default asset targets are in powerpack.app/default-asset-targets.

`:powerpack/content-file-suffixes`

Defaults to ["md" "edn"].

What file suffixes to grep for in :powerpack/content-dir. If you add more suffixes here, you must also implement powerpack.ingest/parse-file for them, see parsing content files.

`:powerpack/create-ingest-tx`

A function that receives a string file name and the parsed contents, and returns transaction data to be transacted into Datomic. See the step by step tutorial for an example.

`:powerpack/dev-assets-root-path`

Additional asset path to include only during development. This can be used to serve dev resources such as an unoptimized ClojureScript build during development. This path is ignored during export.

`:powerpack/get-context`

A function that receives no arguments, and returns a map of data to add to the context argument that is passed to :powerpack/render-page. This can be used to add external dependencies such as another database connection, the current time (e.g. a fresh (Instant/now) for every request), etc.

`:powerpack/live-reload-route`

Defaults to "/powerpack/live-reload".

The route that Powerpack uses for the live reload route. Can be changed if it interferes with your own URLs.

`:powerpack/log-level`

Defaults to :info. May be set to :debug.

`:powerpack/on-ingested`

A function that is called after Powerpack updates the database - e.g. once after initial bootup, and whenever you edit content files. Receives the Powerpack app and a list of derefed results from datomic.api/transact. You can use this information to find what data was added:

:powerpack/on-ingested
(fn [powerpack results]
  ;; All datoms added to the database
  (mapcat :tx-data results)

  ;; The database before the ingest
  (:db-before (first results))

  ;; The database after the ingest
  (:db-after (last results)))

`:powerpack/on-started`

A function that is called after Powerpack boots up. Receives the Powerpack app as its only argument.

`:powerpack/on-stopped`

A function that is called after Powerpack shuts down. Receives the Powerpack app as its only argument.

`:powerpack/page-post-process-fns`

Additional post processor functions. A function that receives the context and should return a map of selector to fn. See the syntax highlighter for an example.

`:powerpack/port`

Defaults to 5050.

What port to run the Powerpack app on.

`:powerpack/resource-dirs`

Defaults to ["resources"].

Directories where you have resources. Powerpack will watch these directories for changes to live reload your app. Should correspond to the resource dirs you have in :paths and extra-paths in your deps.edn.

`:powerpack/source-dirs`

Defaults to ["src"].

Directories where you have source code. Powerpack will watch these directories for changes to live reload your app. Should correspond to all the source dirs you have in :paths and extra-paths in your deps.edn.

Rendering pages

:powerpack/render-page is a function that receives two arguments: context and page. It should return one of the following:

A full ring response
A string, which will be treated as an HTML response
Hiccup, which will result in an HTML response
Any Clojure data, which will result in an EDN response

context is a map with at least the following keys:

:uri the request URI
:app/db a Datomic database value
:i18n/dictionaries prepared m1p dictionary maps
:powerpack/app the full Powerpack app map
:optimus-assets the resolved Optimus assets

page is a Datomic entity map that contains :page/uri, and any other keys you added to the page in question.

NB! When you return a full Ring response, beware that your exported static site will not be able to specify content-type at will - those will be inferred from the file extension in most static web servers. So if you intend to return JSON, include .json in the :page/uri.

Ring responses that specify a content-type of either application/json or application/edn can return arbitrary Clojure data as its :body, and Powerpack will automatically stringify it.

Any HTML response that is a full document (e.g. includes the html element) will be post processed.

HTML post processing

Powerpack performs post processing on all HTML responses. The post processing is designed to be helpful, unobtrusive and only make relatively objective improvements. For each of the improvements Powerpack tries to make, it will only do so if you have not done anything similar yourself - e.g. Powerpack will add a meta tag to set the default viewport, but only if you haven't done so yourself.

Set HTML lang

sets the lang attribute on the HTML element from the :page/locale (with a fallback to :site/default-locale).

Add open graph prefix

Sets the prefix attribute on the HTML element with the og: short hand for open graph.

Add a title

Adds a title element if missing, and if your page has a :page/title. Adds :site/title behind a pipe after the page title.

Ensure utf-8

Adds a meta tag with the charset attribute.

Ensure viewport

Adds a meta tag with name="viewport" and content set to width=device-width, initial-scale=1.0.

Adds open graph metas

Adds open graph meta tags, if your page has one or more of the following attributes:

:open-graph/description
:open-graph/title or :page/title
:open-graph/image

Adds favicon link tags

If your Optimus assets includes any of the following paths:

/favicon.ico
/favicon-16x16.png
/favicon-32x32.png
/apple-touch-icon.png

Powerpack will insert the corresponding link elements linking to them.

Adds Optimus bundles

Optimus CSS bundles are added to the head of the document. JavaScript bundles are added at the end of the body element.

HTTP Redirects

Being a static site generator, Powerpack can't really give you server-side HTTP redirects. But it does what it can when you give it a redirecting ring response, and that is to generate an HTML page with a meta tag using http-equiv="refresh", which causes a redirect in browsers.

To create a redirect, return a Ring response like so:

{:status 301
 :headers {"location" "/elsewhere/"}}

And Powerpack will build an HTML page like so:

<!DOCTYPE html>
<html>
  <head>
    <meta http-equiv="refresh" content="0; url=/elsewhere/">
  </head>
  <body>
    <a href="/elsewhere/">Redirect</a>
  </body>
</html>

Parsing content files

Powerpack parses EDN and markdown (which optionally includes mapdown) out of the box. You can add support for other content file types by implementing the multi-method powerpack.ingest/parse-file. It dispatches on the suffix as a keyword, e.g. :edn, and should return something that can be passed directly to datomic.api/transact, e.g. a vector of transactions (typically maps).

Let's say you want to keep some content in HTML files. Here's one possible way to parse them:

(require '[powerpack.ingest :as ingest])

(defmethod ingest/parse-file :html [db file-name file]
  [{:page/uri (ingest/suggest-url file-name)
    :page/body (slurp file)}])

Rationale

I've built a few Stasis pages in my time, and I keep wiring together the same things. Over time I've accumulated a toolkit of really useful things that I always want when I do static web pages. So I thought it would be worth trying to tie them all up in a neat package.

Normally I'm very adverse to "don't call us, we'll call you" style frameworks, and especially ones that are just a grab-bag of dependencies. So why introduce another such beast into the world?

I think that if the web server is configured well enough, you typically won't need to "own" it when you make static web sites. Your focus will be on crafting content and coding pages. Because the deployment target is a bunch of static files, you won't keep adding middlewares and tweaking the setup. In other words, I suspect it might be possible to "set and forget" the server, thus leaving control of it to a framework like Powerpack might be acceptable.

Based on the experience of building a 5000 page static site with Powerpack in a two person team, I would consider this hypothesis validated.

Powerpack combines a bunch of tools in a package. However, it makes no effort to hide them or "unify" them in any way. They are reasonably configured and wired them together for you, but configuration is passed through, and no features are hidden or proxyed in any way. This should reduce the risk of Powerpack becoming a bottle neck or unnecessary layer of confusion. Obviously, it is a middleman, but the individual tools are used as transparently as possible.

Changelog

2024.11.26

Fix bug where using (powerpack.markdown/render-html md) as the body of a page would cause the page to be rendered as EDN.

2024.11.18

Add :optimus/options

2024.11.05

Change hiccup rendering to chassis.

NB! Potentially breaking change. It breaks my heart, but this change may cause breakage for existing sites. In particular, this change makes Powerpack HTML escape content by default. In prior versions, Powerpack would write any text content verbatim to the generated HTML. This is an unsafe default, which is why this version changes it to be escaped by default.

To render strings unescaped, use powerpack.hiccup/unescape.

2024.09.30

Add support for a custom :m1p/k.

2024.06.26

Improve Powerpack's string parser, avoid parsing strings containing quoted text with read-string.

2024.06.24

Improve Powerpack's reboot heuristics. Regular expressions in the Optimus config would cause Powerpack to always reboot when the file containing the Powerpack config changed. Regexes are now compared as strings, avoiding this issue.

2023.05.31

Gracefully skip backing up the previous export if it fails for some reason (can happen when exporting inside Docker images)

2024.05.18

Allow mapdown strings to be explicitly quoted

2024.01.31

Do not auto-generate og:image:width and og:image:height from the file image dimensions. Instead, control these using the :open-graph/image-width and :open-graph/image-height attributes on the page entity.

2024.01.23

Check for bad links in <link rel="canonical">
Allow for a custom link validator

2024.01.16

Dev-mode improvements:

A little more chatty log (see when starting to process large files, etc)
Avoid printing huge transactions when they fail (only show the first 10 entries)
Disconnect from and delete in-memory database on stop/restart

2024.01.14

Bug fix: Don't try to read external open graph image URLs from disk Bug fix: Fix a case where cross-referencing the same entity in multiple files would sometimes cause the entity to "split". Bug fix: Set Optimus' asset cache to 150ms (down from 250ms)

cljdoc is a website building & hosting documentation for Clojure/Script libraries

Keyboard shortcuts Report a problem cljdoc on GitHub

× close