Matt Wetmore

Accessing list metadata in Hakyll

Like many static site generators, Hakyll allows you to annotate your files with metadata, for use in templates. For example, this post is written in Markdown, and the first lines of the file are

---
title: Accessing list metadata in Hakyll
date: 2015-06-26
comments: true
---

This looks like YAML[1], but Hakyll doesn’t actually use YAML. Instead, by default it parses a simpler syntax, and associates each key to the left of the : with the string to the right. This means that we can’t use the nice list syntax included in YAML, for example:

tags:
 - haskell
 - programming
 - hakyll

However, we can build our own function to parse a simpler syntax: comma-separated, single-line lists. While I don’t believe it’s possible to use the syntax above, let’s settle for less and add the capability to parse comma-separated lists such as

tags: haskell, programming, hakyll

Contexts

Hakyll templates allow us to use dollar-signs ($) to delimit variables which will be replaced when the templates are compiled. For example, part of this post’s HTML is specified in the template post.html, which looks like this:

<div class="main-text">
    $body$
</div>

The $body$ bit is filled in with the contents of the post, after they are compiled from Markdown. But how does the compiler know what to fill in for $body$?

This is where contexts come in. Contexts hold mappings from strings like body to values which will replace them in templates. All of the special forms between $s in Hakyll’s template language derive their meanings from contexts. One such special form is the $for()$ construct, which looks like this:

$for(posts)$
    <li>
        <a href="$url$">$title$</a> - $date$
    </li>
$endfor$

The type of context entry which allows this to work is the listField:

listField :: String -> Context a -> Compiler [Item a] -> Context b

The String is the name of the list used as the argument to for. The Context a is the context which is used for the body of the loop - in the example above, it would be the context providing value for $url$, $title$, and $date$. Next, the Compiler [Item a] is the list of items to iterate over in the loop.

We will use listField to create a list we can iterate over, whose elements are parsed from metadata.

Parsing the metadata

The following function will take a context for each item, and a string to use as the metadata key, and return a context associating that string to a list, parsed from the metadata.

listContextWith :: Context String -> String -> Context a
listContextWith ctx s = listField s ctx $ do
    identifier <- getUnderlying
    metadata <- getMetadata identifier
    let metas = maybe [] (map trim . splitAll ",") $ M.lookup s metadata
    return $ map (\x -> Item (fromFilePath x) x) metas

For the most part, this is a copy of the function which parses tags from metadata fields, found in the Hakyll.Web.Tags module. I’d go on to define

listContext :: String -> Context a
listContext = listContextWith defaultContext

for convenience.

Application: adding “scripts” and “styles” fields

The default Hakyll blog uses a template called default.html[2] to wrap all site content in a consistent look. This template includes the <head> and <body> tags, so everything else on the page can’t use <head> or <body>. If you want to include custom scripts or styling to a particular page, but still keep the consistent look, you can’t place them in the head.

We can use list metadata to handle this. Add the following wherever you want the scripts to show up, say in the <head> tag of default.html:

$if(scripts)$
$for(scripts)$
    <script src="$body$"></script>
$endfor$
$endif$

The $body$ in this case is the contents of the list elements of scripts. It is provided by defaultContext.

Now for any page you want to use the scripts metadata with, compile it with the following context included[3]:

headContext = listContext "scripts" <> defaultContext

Now we can add the following metadata to any pages we want extra scripts on:

---
scripts: /js/custom.js, http://greatlibraryjs.com/source.js
---

  1. YAML Ain’t Markup Language ↩︎

  2. Here is mine. ↩︎

  3. The <> is the infix operator for mappend, the combining operation for monoids. Context is an instance of the Monoid typeclass - combining two contexts returns a context with fields from both, with the left context’s fields getting precedence. ↩︎