Add a note about chroma as a syntax highlighting option.
Add a `help` subcommand.
Add default render methods adapted from personal site.
A static site generator for those who love Markdown and parens.
A long time ago, I wrote a blog engine called Coleslaw. It was written on a lark as I and two friends all resolved to get off Wordpress. We scheduled a lunch and learn for our unwritten blog engines a week later. The initial implementation was thrown together in a weekend. For many years, it was enough.
Collards is a simple static site generator. It offers:
Markdown blog posts with 3bmd
Custom pages built with Spinneret
Local preview with hot reloading via Hunchentoot
Deployment via Rsync with support for subdomains
CLI commands via Clingon to build
, serve
, and deploy
Optional RSS feeds, tag indexes, and year indexes
A straightforward API for custom content types
The official repository and CI builds are available.
While Collards should be portable to any OS, it is only tested on Linux. Notably, the serve command may not work on non-linux OSes.
The only way to install Collards today is building it yourself.
Cloning the repository and running make app
should be enough.
After that, place bin/collards
anywhere on your $PATH
.
While the Makefile builds with SBCL by default, passing LISP=ccl
as an argument to make app
should be sufficient to build with Clozure CL.
Other lisps have not been tested.
For users of Guix, a package is provided in the repo so
running guix shell
will get you a working environment.
At some point, collards may become available via binaries or Quicklisp.
Once collards is installed, building a site involves creating a folder
(or git repo if preferred), adding a .collards
file, and then adding
markdown files to generate posts, lisp files to generate pages, and some
special files to generate RSS feeds or index pages.
Once the config file is written, the collards
command can be run anywhere
within that site folder. Your config might be as simple as:
(:author "Melissa Rogers"
:domain "melissa.website"
:ignored ("drafts")
:title "Adventures in Code")
The config file itself strives to be small, requiring only values for an
Author, Domain, and Title. You can also provide a list of Ignored subfolders.
If you wish to use the deploy
command to rsync your build output to a
server, then you must also supply Server settings. Subdomains can optionally
be specified to send specific directories output to different paths on the
server.
An example config file can be viewed here. The other files in that
example
directory in the collards repo are used as data for the test suite.
For a look at a "production" site that may be more realistic, the repo for
my homepage may be of interest.
When collards builds a site, it will process all files that it recognizes and
copy any other files to an output directory. Then, that output directory can
be freely copied to a public webserver either with the deploy
command built
into collards or manually. The files in the output directory will be organized
the same way they were in the site folder. If all your markdown post files are
in a blog
folder, then the HTML files will go in a blog
output subfolder.
Collards recognizes files by extension (see: Content Types) and currently
knows about .lisp
files which are treated as pages, .md
files which are
treated as posts, and some custom files like rss.feed
which generates RSS
feeds and tag.collection
or recent.collection
which generate indexes of
POSTs with specific tags or in reverse chronological order.
As mentioned above, the examples folder in the project repo or my personal site may be useful to see examples of a live site.
Collards uses Spinneret for HTML generation. No templates are provided by default but this may change in a future version and you are welcome to use my templates as a starting point.
The intent of this is merely to encourage people to design sites according to
their own style and avoid baking in many assumptions. Defining a template is
a matter of defining a COLLARDS.CONTENT:RENDER
method that accepts a content
object and a site object as arguments and generates HTML with spinneret.
Collards will load all lisp files from an init
directory in the site
folder before building so template definition and any desired custom behavior
can be done there.
A simple post template with next/prev navigation might be:
;; Note: I recommend using package nicknames for convenience.
;; I.e. To allow for post:body instead of collards.post:body, etc.
(defmethod content:render ((content post:post)
(site site:site))
(spinneret:with-html
(:doctype)
(:html
(:head
(:title "foo")
(:link :href css-path :rel "stylesheet" :type "text/css"))
(:body
(:div.navigation
(:p "Nav goes here"))
(:section
(:div.article-content
(:raw (post:render-markdown (post:body content))))
(:div.neighbors
(util:when-let (prev (registry:prev content))
(:a :href (content:url-for prev) (content:title prev)))
(util:when-let (next (registry:next content))
(:a :href (content:url-for next) (content:title next)))))
(:div.fineprint
(:hr)
"Unless otherwise credited all material CC-BY-SA"
(site:author site)))))
As mentioned in Adding Content, Collards associates the extension of a file with a particular parser. Markdown is particularly noteworthy both because our parser provides some useful extensions and because it can be used to create Pages as well as Posts based on the frontmatter. A good example of how to write pages using Spinneret will be forthcoming in a future release.
Since Collards uses 3bmd for markdown parsing, their docs will be the
most authoritative source of info on our markdown support. That said,
the behavior of frontmatter in .md
files is unique to collards and
it is valuable to have a brief overview of the useful extensions from
3bmd we have enabled.
It is critical to supply metadata for content generated from markdown. All markdown files intended for processing by Collards are expected to have an initial frontmatter section written in YAML, which you may be familiar with from other blogware like Jekyll.
Frontmatter is separated from the main content by ---
. For example:
---
title: Writing a Static Site Generator
tags: lisp, software-dev
date: 2025-02-03 08:00:00
draft: true
summary: Some experiences writing a static site generator for personal use.
---
## A Sensible Build System
In many ways, static site generators can be viewed as build tools to
turn arbitrary content formats into HTML...
This example demonstrates all the available metadata for our post format. Title is the only required field but comma-separated tags, date, draft, and summary are all supported. Collards does not handle actual Date parsing yet so a date format beginning with YYYY-MM-DD is recommended for now.
The only important attribute not demonstrated in the previous example is the
type
attribute. type
is an escape hatch to allow using markdown files for
generating Pages though nothing prevents you from using it to generate new
kinds of content defined in your config, like a remark.js powered
slideshow. My personal config actually does just that.
Much more common though is the desire to generate a Page from markdown which is useful since they don't participate in tag, year, or recent collections.
---
title: About Me
type: page
---
## Hobbies and Projects
...
Two 3bmd extensions are used: Code blocks and Wikilinks. Code blocks are
highlighted by Colorize but Pygments or Chroma can be used instead
by throwing one of the following snippets in your init files:
(setf 3bmd-code-blocks:*renderer* :pygments)
or
(setf 3bmd-code-blocks:*renderer* :chroma)
For further details, consult the 3bmd documentation.
Wikilinks will come in handy when you want to link between pieces of content
within a collards site. Using a Content ID between double brackets will look
up the specified content in the site registry and link to it. For example:
[[page:about]]
or [[post:2024-reflections]]
.
Answering the siren song of my own laziness, rather than laboriously writing docs about the CLI commands, I have repurposed the usage docs generated by Clingon. If anything is unclear, feel free to ask me questions via email or mastodon. I may be slow to respond.
[function] COLLARDS::COLLARDS/COMMAND
NAME:
collards - A static site generator
USAGE:
collards [global-options] [<command>] [command-options] [arguments ...]
OPTIONS:
--help display usage information and exit
--version display version and exit
COMMANDS:
build Build a collards site.
serve Live reload changed site files and serve them locally.
deploy Sync build output to the server specified in .collards.
help Display help for collards.
AUTHORS:
Brit Butler
LICENSE:
MIT
[function] BUILD/COMMAND
NAME:
build - Build a collards site.
USAGE:
build [options] [arguments ...]
OPTIONS:
--help display usage information and exit
--version display version and exit
-s, --site <VALUE> Path to the site to build [default:
/home/cons/projects/lisp/collards/]
-t, --timing Print the duration of each build step in milliseconds
[function] SERVE/COMMAND
NAME:
serve - Live reload changed site files and serve them locally.
USAGE:
serve [options] [arguments ...]
OPTIONS:
--help display usage information and exit
--version display version and exit
-p, --port <INT> The localhost port to use for site hosting. [default: 4242]
-s, --site <VALUE> The path to the site to serve. [default:
/home/cons/projects/lisp/collards/]
-w, --swank-server Start a swank server to connect a REPL/editor to.
[function] DEPLOY/COMMAND
NAME:
deploy - Sync build output to the server specified in .collards.
USAGE:
deploy [options] [arguments ...]
OPTIONS:
--help display usage information and exit
--version display version and exit
-n, --no-build Skip building the site.
-s, --site <VALUE> The path of the site to build and deploy. [default:
/home/cons/projects/lisp/collards/]
-y, --assume-yes Rather than prompting, run all deploy commands immediately.
Help takes a single topic as an argument and displays the documentation for that topic. If no topic is supplied, the list of available topics is shown.
[function] HELP/COMMAND
NAME:
help - Display help for collards.
USAGE:
help [options] [arguments ...]
OPTIONS:
--help display usage information and exit
--version display version and exit
It is time that we described the API which powers Collards. The docs from this point forward are more concerned with understanding the inner workings of the abstractions and build process than using the CLI tools and building a site.
Before doing anything else, collards needs a site object. The site object holds basic configuration info as well as a registry of Content Objects that will be used to build the site.
A site can be built from any folder with a .collards
file which
serves as a configuration file that the site is created from.
An example collards config can be found here.
[class] SITE
An object to store configuration for the site under
construction such as AUTHOR
, DOMAIN
, TITLE
, and any other arguments. It also
stores a registry, mapping @CONTENT-IDs to @CONTENT-OBJECTS which is queried
during the build process.
[reader] AUTHOR SITE (:AUTHOR)
The name or username of the site author.
[reader] SERVER SITE (:SERVER)
A plist of :host
and :webroot
indicating the
host and root path to deploy the build output to.
[reader] DOMAIN SITE (:DOMAIN)
The domain where the site will be hosted.
[reader] SUBDOMAINS SITE (:SUBDOMAINS)
An list of plists of :path
and :domain
.
This controls the subdomain to which specific paths in the build
output will be deployed.
[reader] TITLE SITE (:TITLE)
The title of the website.
[reader] IGNORED SITE (:IGNORED)
A list of ignored top-level directories in the site. Assumed to be site-relative paths if they do not begin with '/'.
[reader] ASSET-DIR SITE (:ASSET-DIR)
A directory to store site assets in. Assumed to be a site-relative path if it does not begin with '/'.
[reader] BUILD-DIR SITE (:BUILD-DIR)
A directory to store the generated site in. Assumed to be a site-relative path if it does not begin with '/'.
[reader] INIT-DIR SITE (:INIT-DIR)
A directory in the source repo for user code. Lisp files in this directory are loaded first, non-lisp files are ignored. Assumed to be a site-relative path if it does not begin with '/'.
[reader] REGISTRY SITE (= (SERAPEUM:DICT))
Used to store and lookup content during the site build.
The folder with the .collards
file is considered to be the 'site root'.
This is significant because Collards places rendered files in the build
folder relative to where they were under the site root.
As a consequence, it is essential that the .collards
file can be found
and useful to have helpers for determining the relative path of site files.
Errors will be raised if the .collards
file cannot be found or if there
are any missing arguments needed to construct a functioning site.
[function] RELATIVE-PATH PATHNAME
Given a PATHNAME
under the site, return the relative path to the file.
If a .collards
config file cannot be found from traversing the
directories in PATHNAME
, a MISSING-CONFIG-ERROR
will be signaled.
[function] ABSOLUTE-PATH &REST ARGS
Build a path starting at the ROOT-DIR
and appending all supplied ARGS
.
[condition] MISSING-CONFIG-ERROR
An error raised when no .collards
file was found.
[condition] MISSING-SETTING-ERROR
An error raised when a required setting is undefined.
The *SITE*
instance is dynamically bound during the build process but
should not be interacted with by user code. LOAD-CONFIG
is used behind
the scenes to create the site instance.
[variable] *SITE* NIL
The SITE
instance under construction.
This should never be used directly by client code. All interactions
with *SITE*
are done through the documented interface and generally
run from CLI tasks which dynamically bind the site.
[function] LOAD-CONFIG PATHNAME
Given a PATHNAME
under a collards project, find and load the site config.
For situations where there is a need to iterate over the files in the site
a helper macro named DO-FILES
is provided. It automatically ignores the build,
asset, and init folders, any .git folder within the site, and hidden files.
A lower-level CALL-WITH-FILES
function is also provided for special cases.
[function] CALL-WITH-FILES DIRECTORY IGNORED FN
Walk the given DIRECTORY
calling FN
on the files within each subdir. Hidden
files and folders are excluded, along with any subpaths of the IGNORED
list.
[macro] DO-FILES (VAR) &BODY BODY
Iterate over the files within *SITE*
, binding each to VAR
and then executing
BODY
. Skips hidden files and folders as well as the subpaths of IGNORED
.
Content Objects are what collards uses to represent files it will render, usually to HTML, to build the site. All Content objects must have at least a source file they are loaded from and a title used to generate a slug.
[class] CONTENT
A simple base class for any object that will be routable within collards. Relies on the title and slug to generate an identifier.
[reader] FILE CONTENT (:FILE)
The source file the content was loaded from.
[reader] SLUG CONTENT (:SLUG)
The slug generated for the content.
[reader] TITLE CONTENT (:TITLE)
The title to use for the content.
As part of HTML generation, collards finds all files to process in the site
folder and builds content objects with them. To make a content object, we
first determine the file extension and call PARSE
on it, then take the result
of parsing and give that to GET-TYPE
which determines what class BUILD
should
make an instance of. NOTE: The collection and feed types are treated
specially since they generate multiple HTML files from a single input file.
Consequently, user-defined content types can be created by defining:
A new subclass of CONTENT
A GET-TYPE
method with an eql-specialized file
extension returning the class to instantiate
A PARSE
method on that extension returning a plist of
initargs
A RENDER
method to accept an instance of the content
and generate HTML (i.e. a template)
Nothing is required to "inform" collards about the new content type because we introspect the content methods at runtime to determine what files have the necessary machinery for parsing and building.
All this could be done in an init file so no upstream modification to collards
is needed. As described in the Templates section, defining
RENDER
methods is left to the user as part of templating.
[generic-function] GET-TYPE METADATA EXTENSION
Given METADATA
and an EXTENSION
, return the symbol of a class
to instantiate. METADATA
can supply a :type
key to override the result.
[generic-function] PARSE PATHNAME EXTENSION
Read PATHNAME
to parse metadata for the given EXTENSION
.
[generic-function] RENDER CONTENT SITE
Generate HTML for the given CONTENT
using SITE
if needed.
[function] FIND-SUBCLASS NAME
Search for an appropriate subclass of CONTENT
matching NAME
.
[function] KNOWN-TYPE? PATHNAME
Search for an applicable PARSE
method for PATHNAME
. If a match is found,
return the extension as a keyword.
[function] BUILD PATHNAME
Make a content object from the supplied PATHNAME
.
Since pages need to have unique, safe URLs we provide ID
and PATH
functions that help determine where content should reside on the site
and provide a shorthand for uniquely identifying it. SLUGIFY
is run on
the TITLE
of all content during BUILD
to help determine its output path.
[function] ID CONTENT
Generate a unique identifier for CONTENT
based on the class name
and slug of the content. See: @CONTENT-ID and @SLUG.
[function] PATH CONTENT
Return the relative path of CONTENT
under the site.
[function] SLUGIFY STRING
Transform STRING
by first substituting dashes for whitespace as defined
by SERAPEUM:WHITESPACEP
, then removing any non-alphanumeric chars besides
'-' or '/', and downcasing the result.
Finally, there are some internal helpers for determining urls and whether URLs should be relative or absolute, mostly for the build and serve commands.
[variable] *RELATIVE-URLS* NIL
This controls whether URL-FOR
generates relative or absolute URLs. It is
used to ensure that the serve command produces a working local site.
[generic-function] OUTPUT-PATH-FOR CONTENT &KEY EXTENSION
Generate an absolute path for CONTENT
under the build dir.
[generic-function] URL-FOR CONTENT &KEY EXTENSION
Generate a URL for CONTENT
.
The Site registry allows for storing and retrieving loaded content objects as well as providing some helpers to ease HTML generation. There is also an iteration macro provided so that the build command may quickly traverse all known content.
[macro] DO-CONTENT (VAR &KEY (TEST 'IDENTITY) (SORT NIL) (KEY NIL) (SITE 'SITE:*SITE*)) &BODY BODY
Filter the content in the SITE
registry by TEST
, optionally sorting it if a
SORT
is provided. KEY
controls the SORT
key. Once filtered, iterate over
the items, binding each to VAR
and executing BODY
.
Because it is often interesting to view "related" content, there are PREV
and NEXT
generic functions for use when organizing posts chronologically or
generating collections by tag or year.
[generic-function] PREV CONTENT &OPTIONAL SITE
Given a CONTENT
object, find the previous object in the
registry when sorted in reverse chronological order.
[generic-function] NEXT CONTENT &OPTIONAL SITE
Given a CONTENT
object, find the next object in the registry
when sorted in reverse chronological order.
Finally, there is the primary functionality to lookup and register content. While no content objects define custom methods at present, these are left as Generic Functions in case method combination or specialization is useful.
[generic-function] LOOKUP CONTENT-ID &OPTIONAL SITE
Given a CONTENT-ID
, search for a matching piece of content in
the site registry. If no match is found, MISSING-CONTENT-ERROR
is signaled.
[generic-function] REGISTER CONTENT &OPTIONAL SITE
Register the supplied CONTENT
with site. If an existing piece
of content is found that shares the same CONTENT-ID
but has a different source
file, a DUPLICATE-ERROR
condition will be signaled.
[condition] DUPLICATE-ERROR ERROR
When content is supplied for registration and uses a
@CONTENT-ID that has already been registered, a DUPLICATE-ERROR
is signaled.
It is not signaled if both the original and new content have the same path.
[condition] MISSING-CONTENT-ERROR
Signaled when a piece of content could not be located.
[class] ASSET CONTENT:CONTENT
ASSET
s are intended for larger files, usually media or binary
files that need convenient linking but are best kept in cloud storage.
[function] ASSET? OBJECT
Returns t if OBJECT
is an ASSET
.
[class] PAGE CONTENT:CONTENT
PAGE
s are user-defined static web pages with custom layout.
They are intended to provide maximum flexibility by allowing user-supplied
code to generate arbitrary content as part of the build process.
[reader] BODY PAGE (:BODY)
Optional markdown content for producing a page. The POST
template is used instead of RENDER-FN
when a BODY
value is present.
[reader] RENDER-FN PAGE (:RENDER-FN)
A function accepting content and site that generates HTML.
[function] PAGE? OBJECT
Returns t if OBJECT
is a PAGE
.
[macro] DEFINE-PAGE &KEY TITLE RENDER-FN
Given a TITLE
and a symbol RENDER-FN
accepting a PAGE
object and a SITE
,
supply the page metadata for access by LOAD-CONTENT.
[class] POST CONTENT:CONTENT
POST
s are collards way of representing articles with a
common template, usually grouped by topic and date, and part of a journal.
[reader] DATE POST (:DATE)
The date the post was created.
[reader] TAGS POST (:TAGS)
A comma separated list of tags for the post.
[reader] BODY POST (:BODY)
The main content of the post.
[reader] DRAFT? POST (:DRAFT)
Whether the post should be published or not.
[reader] SUMMARY POST (:SUMMARY)
An optional short summary of the post topic.
[function] POST? OBJECT
Returns t if OBJECT
is a POST
.
[function] ALL-POSTS &OPTIONAL (SITE SITE:*SITE*)
Given SITE
, return a list of POST
objects in reverse chronological order.
[function] RENDER-MARKDOWN MARKDOWN
Given a MARKDOWN
formatted string, use 3bmd to render it and treat
any wikilinks as objects to be looked up via URL-FOR in the registry.
[class] FEED CONTENT:CONTENT
A generic feed class for generating Atom or RSS.
[function] FEED? OBJECT
Returns T
if OBJECT
is a FEED
.
Most sites have interest in grouping subsets of content together by different means. Tags are one way to accomplish this but collards also models groupings by date as a kind of collection.
Collections are generated in a second phase after all initial content is loaded. Currently tag collections and annual collections are generated. Only post objects participate in collections.
It is worth noting that collections, once generated, are stored in the
registry along with other content but they are not loaded by CONTENT:BUILD
like other content objects, instead relying on GENERATE
. tag:foo-bar
and
year:2024
are examples of collection content-ids.
[class] COLLECTION CONTENT:CONTENT
A simple wrapper class for managing groups of content.
[function] COLLECTION? OBJECT
Returns t if OBJECT
is a COLLECTION
.
[reader] KIND COLLECTION (:KIND)
The kind of collection to build after content is loaded.
[accessor] ITEMS COLLECTION (:ITEMS)
The content items in the collection.
[generic-function] GATHER COLLECTION-TYPE SITE
Collect all the usages of COLLECTION-TYPE
in SITE
posts.
[generic-function] POPULATE COLLECTION SITE
Add the relevant posts on SITE
to the given COLLECTION
.
[function] GENERATE &KEY (SITE SITE:*SITE*)
Generate collections from the registry after all other content is loaded.
[glossary-term] Content Objects
A Content Object represents a file collards
will process that
produces a unique URL on the site. A content object must store at least
the path to its source file and a title to use during URL generation.
[glossary-term] Content Type
A Content Type is a data type representing a kind of file that collards knows how to recognize (parse) and generate a page for (render). Out of the box, Collards has the following content types: Pages, Posts, Feeds, Collections, and Assets.
It is possible for users to define their own Content Types in their
init files by supplying a subclass of COLLARDS.CONTENT:CONTENT
along
with a COLLARDS.CONTENT:GET-TYPE
method, a COLLARDS.CONTENT:PARSE
method,
and a COLLARDS.CONTENT:RENDER
method returning HTML.
The GET-TYPE
method should be EQL
-specialized on the file extension,
for example :md
for markdown files, and return the class to
instantiate (e.g. with FIND-CLASS
), not a symbol.
The PARSE
method should read the supplied file and return a plist
of metadata including at minimum :file
and :title
attributes.
[glossary-term] Content ID
A Content ID is a string uniquely identifying a piece of
content managed by collards
. It is constructed from the
Content Type and Slug of a piece of content, joined by a ':'.
For example, page:about
or post:favorite-albums-of-2023
.
A notable limitation of the current scheme for content IDs is that titles must form a unique slug for all content of the same type. i.e. There cannot be two pages with the same slug, even if they are in different folders.
[glossary-term] Slug
A Slug is the unique identifying part of a web address, such as '/favorite-albums.html'. Collards does not consider file extensions as part of slugs and generates a slug for each piece of content when it is registered by transforming the title with a SLUGIFY function.
Users can feel free to redefine COLLARDS.CONTENT:SLUGIFY
within their
init files since init files are loaded before site processing occurs.