Simple Static Sites

My setup for building and deploying a personal website

Personal websites! Everyone wants to have one, but nobody wants to go through the bother of maintaining, updating them, and fixing them. I thought I’d go through my personal setup, and show how I build and deploy my own site.

Outline

Let’s go through the basics first. I build my site using the static site generator Hugo, and serve my content (for free!) with GitHub Pages. I chose Hugo for its blazing fast performance (~25ms to build my current site), and its focus on giving full control over design, structure and templating. While Hugo, like its competitors, offers a plethora of pre-built templates to choose from, I enjoy having the freedom to change anything to do with my site’s presentation.

Here’s a few of the methods I went through to build and deploy my site. Hugo, like most generators, keeps a subdirectory of your project to store build artifacts (files created at build time). When you call the hugo command, your site will be generated into the public directory. On GitHub, I used a dev branch for these templating files, and the master branch to deploy off of (this is required by GitHub Pages config).

My goal was to eliminate as much effort from my side as possible when it comes to building and deploying, but at the same time ensuring that I maintain a fairly clean solution that avoids ‘dirty’ fixes to make functional ends meet.

Using .gitignore

My first approach was quite naive but effective.

This approach is simple and understandable, but I found a few drawbacks with it. This setup struggles to be portable, as you have to keep setting up the public folder when moving from computer to computer, and at the time it was unwieldy to have to keep on remembering to push after each build.

CI/CD Providers

The next approach was seeing if I could find a solution where all the heavy lifting (building and deploying) would be done through an external service.

I tried two options for this. Travis was my first condsideration because of how easily it ties with GitHub itself, and because I had used it in the past with reasonable success. I just wasn’t quite able to find a solution that I liked however. Although solutions on the internet could get Hugo working in Travis through the use of Python libraries or managing dependencies in Go, there was no simple solution that I both liked and understood well. A dirty fix might have got the job done in this case, but I didn’t feel like letting something I didn’t understand to all the work for me.

I did a little investigating and found that GitLab also offers a GitHub Pages-esque service, and with inbuilt support for Hugo in GitLab CI, I thought I’d found my solution. Through repository mirroring (another fantastic GitLab service!), I would also be able to keep my GitHub Pages site up to date with my separate GitLab Pages site. The stumbling point came when I tried to access the static site contents in the public directory generated by Hugo. As it turns out, GitLab will simply deploy the contents of this directory to nchlswhttkr.gitlab.io, with no push to master being made. Generated files were confined to the build environment, so you couldn’t push them to your master branch at the end of your build cycle.

It should be noted that at the time of development/writing, GitLab is working on a feature to allow CI runners to push back to other repositories during a build. However, this issue had been open for more than two years and a public release did not seem to be on the horizon).

Method 3: (Back) To The Local Repo!

After development stagnated for a while as I studied at university and worked with MonPlan, I learned a bit more about the features of Git, and how to use scripts to automate simple actions. Everything started to click into place when I discovered a little feature of Git called hooks, which allow you to run various configured commands or scripts before/after key actions in the Git lifecycle (commit, push, etc…).

After a series of attempts and subsequent fixes, I was able to set up a script that sets up all the Git hooks I needed, and all that was needed was an init script to run making a fresh clone of the repo. Hugo doesn’t overwrite the .git directory from the public directory, so there’s no need to worry about needing to repeat myself after rebuilding the site. With that, I was able to make a setup that I was both understood well and felt happy using.

Now, pushing an update to my site is as quick and easy as a commit and a push!