/ #Hugo #static site generators 

A tale of two site generators

I made my website back in 2008 as a way to build an online presence. The site began as a WordPress blog hosted on wordpress.com. After years of living at wordpress.com, I moved to my own server for a short stint. Eventually, I discovered GitHub Pages and started my journey into static site generators to avoid the burden of managing my own hardware. Today, I’ve moved from a custom static site generator to Hugo. This post explores why I would make that kind of change.

In 2013, the landscape of static site generators was different. There were not a ton of choices. I really wanted to run a Python tool and my best option was Pelican, which I ran on my own server and even contributed back code. Unfortunately, I wanted something more than a blog, and I couldn’t figure out how to make Pelican work in the way I needed.

Without a tool to do want I wanted, the most fun thing I could think to do was build my own. Making my own tool would be a great way to learn about the specifics of static site generators and get something that did exactly what I desired.

The result of my effort was handroll. Today, handroll has all sorts of features that I’ve needed for my site. Features like:

  • Markdown, reStructuredText, and Textile support
  • A development server that does incremental build updates
  • Blog plugin to generate feeds and listing pages
  • OpenGraph and Twitter metadata plugins
  • Jinja2 templating
  • Sass support
  • A plugin and extension system with well documented extension points

handroll served me well, but I gave up on it a few weeks ago. Why would I do that?

The short answer is community.

Never underestimate the power of a robust community.

When I created handroll, I tried to build a community by documenting the tool well on Read the Docs. I even attempted to add translations to the command line interface for a large number of languages. Google Translate was the source of the translations so I apologize for anyone looking at those strings today. handroll never got traction. It’s hard to get noticed even when something is pretty good.

Because handroll never established a community, it meant that I was doing everything. A competent developer working in his spare time can’t compete with a massive community of people focused on a single project. Many hands make light work or something.

In contrast, Hugo took off in popularity. The project has 540 contributors and 28k stars on GitHub. That kind of attention helps remove all the rough edges.

Having lots of users causes these kinds of changes to happen (to an average project):

  • The user experience is better.
  • The documentation is better.
  • The amount of help on Stackoverflow is better.
  • The diversity of contributions is better.

That last bullet is the one that won the day for Hugo for me against my own project.

My site was looking fairly stale from a design perspective. I have a reasonable skillset with CSS, but I’m not a designer. Without a community, there was no hope of handroll ever having a set of themes built by designers. Because Hugo is so popular, it attracted the attention of developers and designers alike. Hugo has an entire Themes page that shows off many high quality themes that are ready to use. This was super appealing to me.

I set out and created a branch that would try to use Hugo. My posts were already in Markdown with a YAML frontmatter so my conversion process was pretty easy.

As a bonus, I got to add things like tags and categories that I never had time to implement myself for handroll.

Now, you’re reading the result. I’m really loving two things about this:

  1. I don’t have to maintain a tool just to maintain a website.
  2. The speed and features of Hugo are really nice.

Thanks to the Hugo community for making something easy and fun to use!

Author

Matt Layman

Matt Layman is a software engineer from Frederick, MD. He is an open source software maintainer and advocate for the Python programming language.