Years ago I figured out that I’d eventually have to migrate my blog from Blogger to something more independent, and based on my previous experience with Wordpress I wasn’t exactly enthusiastic to go down that path.
In 2015 I’ve seen Scott Lowe going from Wordpress to Jekyll and then to Hugo, and decided it might make sense to recreate ipSpace.net blog with a tool that generates static web pages… but never found the time to do it.
Fast forward to early spring 2020. On March 12th we were effectively told to stay at home with more rigorous measures eventually put in place to contain morons who didn’t get the memo, and I decided to resurface that old project. A week later the ipSpace.net blog is generated with Hugo and hosted as a set of static web pages.
Some functionality is still missing, notably:
- Popular posts
- Adding new comments
Unfortunately I don’t see the quarantine lifted any time soon, so they might be done sooner than I would prefer…
Even though I tried to test the setup as much as possible, there might be other broken bits and pieces. If you find them, please contact me (for obvious reasons you can’t write a comment ;).
Of course I missed a few things (or thought they could wait for a bit longer). Here are the fixes already put in place since the blog post was published:
- Atom and RSS feeds work (had to change the URLs, but that shouldn’t impact you if you used feed.ipspace.net URL in your RSS reader)
- Replaced links to ioshints.blogspot.com and blog.ioshints.into. It’s so nice when you could write a script that traverses the whole directory tree and fixes stuff ;)
- Replaced links to category pages in blog posts
- After being annoyed for years, I finally managed to unify tag capitalization (yes, I’m aware this is textbook OCD behavior, but it felt so good ;)
- It turns out working with Google Analytics API is much simpler than I expected. Popular posts are back ;) (added 2020-03-23)
In case you want to know more…
It wasn’t exactly an easy journey. Being of a slightly masochistic persuasion I used Python (instead of Perl) to develop all the glue I needed, including:
- Converting Blogger XML dump into Hugo posts;
- Fetching all Blogger images referred to in various blog posts into the same Git repository and updating the HTML markup on the fly (hint: BeautifulSoup is a great tool)
- Salvage all comments you made in the last 14 years from Blogger XML dump and save them into JSON format
Next step: figuring out Hugo. I wanted to retain the existing web page format, so I had to start by creating my own Hugo theme, and learned more about Go templating than I ever wanted… and then hit the performance roadblocks. Creating blog archive and popular tags sidebars in Hugo took forever (as Hugo had to generate them for every single blog post). Time for more glue - a script that traverses all blog posts and generates:
- Stub pages for monthly and yearly archives;
- Sidebar HTML markup pointing to monthly and yearly archives
- List of recently-used tags and all other tags.
After fixing the sidebar performance, it was time to include existing comments into the blog posts. Dumping 15.000 comments into a single JSON file didn’t seem like a good idea. Creating 4.000 JSON files and have Hugo read them all would probably be even worse. In the end, I solved the dilemma with yet another script that reads per-post comments in JSON format and creates corresponding HTML markup which is then easily included into blog posts with Hugo’s readfile functionality.
- CI/CD configuration that recreates and republishes the whole blog on every change in the publish branch;
- A script that merges master branch into publish branch, and recreates the comments and sidebar markup. That script is run every few minutes, and as a nice side effect triggers publishing of blog posts that have date set into the future.