While working on an algorithm problem today, I received an issue on an open source repo. I was asked about a build error for one of the templates in Hakyll CSS Garden. This repo is a collection of themes I put together for a static site generator called hakyll. The build error says:
[ERROR] Missing field site_description in context for item archive.html
A bit of background and dynamic values in template files
Hakyll written in haskell and the only package I am aware of (as of May 2016) that is available for generating static sites from markdown or latex with Haskell. There are many other options that allow you to blog using the Haskell language but for the idea of converting text files to a site with consistent theming, this is it. What I want to focus on explaining today is the template system.
The template files are html snippets with variables such as author
or body
inside html tags. Most static site generators have a directory for templates
, though sometimes you might see this folder called layout, or includes as well. The terminology mainly depends on what the framework author decided - in the case of hakyll, template is the default word. This can be changed easily within the generator site.hs
file.
Each of the variables such as author
or body
refers to string values, that the generator will plug in with proper values. Where do these values come from?
1) You can save the string values as a metadata field at the top of posts pages such as:
This is the easiest way in my opinion, but unless you are using a template file that pre-fills the information for you, it’s easy to forget which meta fields are needed in your theme.
2) You can save the string value in your site.hs file like so:
This approach hardcodes the values so that every file generated will have these metadata values. Note that you can save the values in another form of data structure as well, whatever you’re comfortable with in Haskell.
The advantage of this method is that when any of these values change you only have to change 1 file instead of every single page.
3) Lastly, you can save this information in a config file. It’s probably the best practice as far as content management - separating the infrastructure from content and metadata. Do you need to do this for a personal blog? That’s the author’s preference.
Jekyll uses the _config.yml
to save the configuration information, and the example for this blog can be seen here. I don’t know how to read from files with hakyll yet so until then I’ll leave the code as an exercise for the reader
How to fix the issue
As you can see from the hackage docs, the defaultContext
(link) comes with the default set of values:
body
- any metadata field (i.e. blog post date, tags)
url
path
title
As you can see, the error message was about the template variable site_description
, which is not included by default, so when the template contains this variable you’ll need to connect the value to the generator using one of the methods mentioned above.
References