Hugo, Org and starting over at a new blog

I’d decided to give up on my free WordPress blog a while back, after having watched a SystemCrafters video on how to blog from Emacs (d’oh, I just have to do everything from Emacs). Of course, daviwil only covered the basics of writing and exporting using Emacs’ built-in packages and later for deploying it on GitHub / Sourcehut pages. And it took me a good year to get down to making the switch but apparently moving to a new place for work is the motivation it required. It felt like cheating to use Jekyll on GitHub and I wanted something that could be deployed to a Sourcehut site easily too (Sourcehut blocks any CDN-based CSS loaded into the site’s HTML) and I wanted something extremely minimal, like Drew Devault’s blog but with even fewer frills — no images anywhere except for if a blog post required them. Looking at his blog source, however, made me realise that there was a lot more to that minimalism than one could see.

This has come through fairly well, so far. I’ve got an org-capture setup for this that links up every new entry captured into a master posts file, adding all the relevant info.

(with-eval-after-load 'org-capture
          (defun org-hugo-new-subtree-post-capture-template ()
            "Returns `org-capture' template string for new Hugo post.
        See `org-capture-templates' for more information."
            (let* ((title (read-from-minibuffer "Post Title: ")) ;Prompt to enter the post title
                   (fname (org-hugo-slug title)))
              (mapconcat #'identity
                         `(
                           ,(concat "* TODO " title)
                           ":PROPERTIES:"
                           ,(concat ":EXPORT_HUGO_BUNDLE: " fname)
                           ":EXPORT_FILE_NAME: index"
                           ":EXPORT_HUGO_AUTO_SET_LASTMOD: t"
                           ":END:"
                           "%?\n")          ;Place the cursor here finally
                         "\n")))

          (add-to-list 'org-capture-templates
              '("h"                ;`org-capture' binding + h
                "Hugo blog post"
                entry
                (file+olp "~/my_gits/brihadeesh.github.io/content-org/blog/posts.org" "Posts")
                (function org-hugo-new-subtree-post-capture-template))))

Exporting to markdown (Org just doesn’t have a good enough support yet), tags and organisation of pages into bundles is handled by ox-hugo ^{1, 2}. The header arguments in the capture template cover everything. With Emacs’s .dir-locals.el feature, a file of that name in the home directory of the blog ensures every new entry or modification into the master posts file gets auto-exported to markdown on save. The contents are quite simple.

;; ~/.dir-locals.el
(("content-org/"
  . ((org-mode . ((eval . (org-hugo-auto-export-mode)))))))

With Emacs’s Org mode, this posts file has subheadings under a Posts header, each of which is a blog post and is exported to a sub-directory under ~/content/posts/ as a lone index.md keeping with the page-bundle kind of organisation.

A tree run for the content directory shows:

$ tree content
content
├── about
│   └── index.md
├── emacs
│   └── index.md
├── emacs-literate-configuration
│   └── index.md
├── _index.md
├── posts
│   ├── a-dark-side-to-pets
│   │   └── index.md
│   ├── introduction
│   │   └── index.md
│   ├── misunderstanding-evolution
│   │   └── index.md
│   └── pets-put-in-context
│       └── index.md
└── publications
    └── index.md

where every sub-directory in the top-level directory has a page of its own while the home-page is the sole _index.md in the same. What I’ve got going feels a little hacky but I’ll figure this out.

Hugo, being a static site generator, creates HTML exports into ~/public and this is what the site uses. All major git hosting services have configurable CI/CD for deploying these to the domain and they’re run automatically if you have a specific file in either

1. the root directory of the repo for Sourcehut called .build.yml
2. ~/.github/workflows/ for GitHub called anything you want with a .yml extension.

Mine uses github-pages and it looks like this:

name: github pages

on:
  push:
    branches:
      - main  # Set a branch that will trigger a deployment
  pull_request:

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
        with:
          submodules: true  # Fetch Hugo themes (true OR recursive)
          fetch-depth: 0    # Fetch all history for .GitInfo and .Lastmod

      - name: Setup Hugo
        uses: peaceiris/actions-hugo@v2
        with:
          hugo-version: 'latest'
          # extended: true

      - name: Build
        run: hugo --minify

      - name: Deploy
        uses: peaceiris/actions-gh-pages@v3
        if: github.ref == 'refs/heads/main'
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          publish_dir: ./public

There’s still a lot to fix

1. heading anchors on top-level pages are superfluous
2. maybe consider switching to a theme-agnostic setup like Drew’s
3. get rid of unnecessary indentation like in the table of contents and headings
4. add anchors even to lower level headers
5. switch to a Sourcehut site (eventually and when I can afford it)

This is but a blog post written, and edited, within half an hour so I likely haven’t covered a lot of important things. I’ll add some links to others’ blog posts that discuss using this or documentation as I come across them.