I decided a while ago that I wanted to to start a blog, but despite GitHub Pages natively supporting Jekyll, I was put off because I didn't want to setup Ruby and the other dependencies just for a static site. Sometime later I was reading this article and I noticed it said:
Proudly powered by Pelican, which takes great advantage of Python.
Python is my main language so, Pelican was perfect.
I started playing with it in a toy repository. I liked it a lot, but the one thing I didn't like was that the publish step used another tool, ghp-import, to push the rendered site to the gh-pages branch, and it had to be run manually. I decided to experiment with GitHub Actions, which recently started allowing Pages deployments via Actions. In another toy repository I tested the basics:
- python and pip
- pip can install packages
- can run a make command
That all, worked fine, so I started experimenting with building a pelican site and publishing to Pages. You can find the workflow here. It is heavily based on the Jekyll example in structure, and the use of the first party actions:
The difference is in the pelican build part. I wanted to avoid using the 3rd party pelican build actions from the marketplace, because I saw some that required generating and feeding in a GitHub token, and most were based on the old branch based Pages method. I felt there should be a simpler way, and It turned out I was right! In the end this was all it needed:
- run: pip install "pelican[markdown]"
- name: Generate Pelican
run: pelican content -o output -s publishconf.py -e SITEURL='"${{ steps.pages.outputs.base_url }}"'
In the end I stopped using the default Makefile. It was easier to pass in the SITEURL in the build command, and since I mainly use Windows, I didn't use it much locally.
This site is now published automatically whenever I push a commit to main! ✅