Adam Pearce github twitter email rss

Incremental Rebuilds and Hot Reloading: 60 Lines of Literate Code for Static Blogging

For five years, I was frustrated by every blogging engine I tried.

WordPress made it difficult to embed inline interactive charts. Octopress’s predefined css was hard to disable and pasting Stack Overflow instructions on installing gems without understanding what renv or rvm were eventually broke my ruby installation. Metalsmith was easier, but I never managed to successfully configure the rss plugin.

And none of alternatives I looked at supported hot reloading.

Writing my own blogging software seemed like epitome of yak shaving. I thought it would be difficult, too, until I came across Jeremy Ashkenas’s Jorno and Rich Harris’s Svelte blog last summer. Using their code as a starting point, I spent a lazy Sunday simplifying my setup.

Now this site is built with just 60 lines of code. And they’re run directly off of this post.

How It Works

Each post is a markdown file in the source/_posts folder. The posts get read in, parsed and written out to public/ as an HTML file using one of templates from source/_templates.

Static files that don’t need preprocessing, like images or javascript, are copied directly from source/ to public/ with rysnc in preperation for publishing.

var fs = require('fs')
var {exec, execSync} = require('child_process')

var public = `${__dirname}/../../public`
var source = `${__dirname}/../../source`

function rsyncSource(){
  exec(`rsync -a --exclude _posts --exclude _templates ${source}/ ${public}/`)

Markdown is converted to HTML with marked and syntax highlighted by highlight.js.

var hljs = require('highlight.js')
var marked = require('marked')
  highlight: (code, lang) => hljs.highlight(lang, code).value,
  smartypants: true

Files in the _templates directory, currently rss.xml, sitemap.xml and post.html, are ES6 template strings. eval turns them into functions that can be passed data.

var templates = {}
readdirAbs(`${source}/_templates`).forEach(path => {
  var str = fs.readFileSync(path, 'utf8')
  var templateName = path.split('_templates/')[1]
  templates[templateName] = d => eval('`' + str + '`')

function readdirAbs(dir){ return fs.readdirSync(dir).map(d => dir + '/' + d) }

Each post file in the source/_posts folder is read in with parsePost and saved to the posts array.

Instead of having to install and configure a plugin, I created an rss feed by passing the array of posts to the rss.xml template and writing out a file.

var posts = readdirAbs(`${source}/_posts`).map(parsePost)
fs.writeFileSync(public + '/rss.xml',  templates['rss.xml'](posts))
fs.writeFileSync(public + '/sitemap.xml', templates['sitemap.xml'](posts))

Passed the path of a post, parsePost reads the title, url, date and publish status from front matter at the top of the post. The markdown body is converted to an HTML fragment and an object representing the post is returned.

function parsePost(path){
  var [top, body] = fs.readFileSync(path, 'utf8')
    .replace('---\n', '')

  var post = {html: marked(body)}
  top.split('\n').forEach(line => {
    var [key, val] = line.split(/: (.+)/)
    post[key] = val

  return post

writePost takes a post object, creates a folder for it in public/, runs it through a template and writes out the post to index.html.

function writePost(post){
  var dir = public + post.permalink
  if (!fs.existsSync(dir)) execSync(`mkdir -p ${dir}`)
  fs.writeFileSync(`${dir}/index.html`, templates[post.template](post))

And that’s all the code that’s needed to build the blog!

To get it all on the internet npm run publish runs lit-node on this post to regenerate everything locally, then uses rsync again to copy the public directory to a remote folder that’s being statically served.

"scripts": {
  "publish": "lit-node source/_posts/ && 
    rsync -a public/",
  "start": "lit-node source/_posts/ --watch & 
    cd public/ && hot-server"

npm run start runs hot-server in the public folder and runs this post with the --watch flag. Changes in the source directory rerun rsyncSource, which copies the the update file to public, triggering hot-server’s file watch and passing the file to the browser along a websocket. A little Rube Goldberg, but still plenty fast and simpler than rewriting hot-server here.

Edits to a post rebuild just that post, making hot-server trigger a page reload.

if (process.argv.includes('--watch')){
  require('chokidar').watch(source).on('change', path => {
    if (path.includes('_posts/')) writePost(parsePost(path))

I don’t spend much time looking at sitemap.xml or tweaking the templates, so they’re not hooked up to automatically update. I’ve tried to only implement exactly what I need without any unnecessary abstractions to keep the code easy to work with. Writing both the code and content lets you aggressively cut corners.

Make Your Own

I’m not totally sold on literate programming yet. I quite liked the having all the code fit on one screen and ⌘-B doesn’t work out of the box. But I’ve been asked a couple of times for advice on putting words and charts on the internet without Medium App Nag getting plastered all over it. Hopefully this post shows how far a little glue code can go when paired with a folder of markdown files, rsync and a static server.

If you’d like to try it without futzing with the literate bits, there’s a javascript only version.