manxu

Did you look at Pelican? I share the frustration with much of Hugo’s infrastructure: the template language is buggy and inscrutable, and the plugin architecture wanting.

I ended up with Hugo, but I considered Pelican. It uses standard Jinja templates, which I find much more rational (but it might just be me) and I recall there were plugins for a lot of things, including different source formats. The code is written in Python, so that even if there isn’t a plugin for a format you need, there probably is a Python library for it and it should be relatively easy to make it a plugin.

Crap, now I want to switch to Pelican…

Hugo watch mode (both server and build) does not produce accurate sites on change and is really meant for development. I find after a developing for a while, I have to kill the process and restart it and then things are “fresh”

From reading the documentation, I strongly have the impression that hugo focuses on being fast on re-render and that the idea is to build and deploy to public site each time there is a change. The big difference is probably whether to render locally and push the generated content, or to push the source markdown and render remotely (which I chose).

I ended up with Hugo, a git repository, and a cron job for the build. I write an article, check it in, the server picks up the git change and rebuilds the site. What I like about the setup is that the server only has the binaries hugo and git, and a shell script for the rebuild. Also, I write in Markdown, add media to the git repository, and articles are published soon after I check in without any remoting on my part.

I did look at WriteFreely after the setup, though. I find the minimalist design very beautiful. Didn’t switch to it, but may look at it again for another project. https://github.com/writefreely/writefreely

There must have been a gallery component, since I only looked at generators that had one available.

Honestly, in hindsight the templates were really not a big deal. Just for fun, I tried converting the Hugo template I used to Pelican, and it was easy for me.

Pelican is solid and mature and I would use it, in hindsight. The only major flaws are that it’s much slower (but makes up for it with incremental builds) and that the community is much, much smaller. On the plus side, Jinja2 is much, much, much better than Go templating (Hugo borrows from Go).

I went the same direction, from WordPress to static site generation. I did the same evaluation as you are trying to do and ended up with Hugo, mostly because there is a lot of support available for it. My runner up was Pelican, because I was fluent in Jinja2, but I didn’t want to mess around with the templates and Hugo’s were prettier. Sue me, I am shallow.

The one regret I have about Hugo is that the templating language is challenging. I am trying to be as neutral as possible, but it seemed like even simple things were complicated to achieve. If someone would come up with a Hugo that speaks Jinja2, I’d be really delighted.

Other than that, conversion from WordPress to Hugo was relatively straightforward, despite needing to find a gallery component and converting menus. Hugo is indeed very fast in processing, which become important when your blog has thousands of articles.

I set up the blog as a private git repository. The server pulls from it, then runs Hugo and a full text search engine, and the content is visible and searchable within five minutes on update.

I think it’s the “basically” part in basically right back where we started that makes the difference. Even if 100% of Fediverse users were on a single instance, once that instance starts pulling Musk moves, the users can move somewhere else.

Only if that instance defederates from everyone else do we get the Twitter situation again. Or, since that’s actually what happened, Truth Social, which I think is just a slightly modified Mastodon instance with federation disabled/deactivated.