Back in 2021 I had a look around and decided to base this incarnation of my blog on an open-source Jekyll theme called “Type”, though I’ve changed and added so much it’s quite some distance from the original as it stands today. As I added blog posts the performance became much worse. Initially I blamed Jekyll for this, until I took a closer look. What I learned was that the blog theme did some things in sub-optimal ways, so over the course of 2024 I have corrected as many of them as I can. Build time dropped from ~12 seconds to ~1 second.
My fixes and optimisations included:
- Optimised SASS to compile once in plugin rather than header of every page (23 Feb)
- Optimised all includes and templates to reduce build time (19 Aug)
- Improved “noun” replacement, automatically-emphasised words (3 Sep)
- Better YouTube embeds using lite-yt-embed (11 Sep)
And my additions along the way:
- Multi-carousel support (14 Jul)
- Automatic transformation of local images urls to CDN urls (19 Aug)
- Automatic smart quotes in post titles (19 Aug)
- Added OpenSearch to allow searching blog from address bar (15 Sep)
You might ask… what’s left to do? Not much I reckon. So I took a closer look at what is involved in me creating a blog post and made a list of the most repetitive, awkward or error prone tasks:
- Entering links, partcularly links to other blog posts
- Entering tags, going from memory or using search across project
- Getting images onto my CDN server, currently copy by SFTP
Let’s solve all these annoyances!
Entering links and tags
You can read about these two in previous blog posts:
- Created Markdown File Linker to make linking easier (8 Oct)
- Created YAML Tag Picker to make tagging easier (8 Oct)
Getting images onto my CDN server
I was still having to copy my images to my CDN server manually, which was a pain. But the software stack on the server was intimidating: I serve my websites using a collection of docker containers. I kept putting it off for a rainy day, but I knew I would evetntually get around to automating it.
The idea was to not upload the images at all, but rather download them to the server directly. I’d use a GitHub Webhook to trigger a git pull
of the latest files onto my server. It took a few hours and a few attempts, but I finally arrived at a fairly elegant system I’m happy with:
- GitHub Webhook that runs on push event
- PHP script in web server docker container receives, validates, and creates a trigger file
- Service on server outside of docker looks for trigger file and does
git pull
GitHub Webhook
- Payload URL: https://www.example.com/webhook.php
- Content type: application/json
- SSL verification: enabled
- Which events would you like to trigger this webhook: Just the push event
- Active: on
Webhook receiver (PHP script)
Webhook git pull watcher (Shell script)
Trigger (System service)
Setup
You can set up the service using:
sudo systemctl daemon-reload
sudo systemctl start git-pull-watcher
sudo systemctl enable git-pull-watcher
Other commands
sudo systemctl restart git-pull-watcher
sudo systemctl status git-pull-watcher
Enjoyed this blog post? Please show some support.
--
Comments: @gingerbeardman