node package manager


Blog like it's 1999


Journo = module.exports = {}

Journo is a blogging program, with a few basic goals. To wit:

  • Write in Markdown.

  • Publish to flat files.

  • Publish via Rsync.

  • Maintain a manifest file (what's published and what isn't, pub dates).

  • Retina ready.

  • Syntax highlight code.

  • Publish a feed.

  • Quickly bootstrap a new blog.

  • Preview via a local server.

  • Work without JavaScript, but default to a fluid JavaScript-enabled UI.

... let's go through these one at a time:

Write in Markdown

We'll use the excellent marked module to compile Markdown into HTML, and Underscore for many of its goodies later on. Up top, create a namespace for shared values needed by more than one function.

marked = require 'marked'
= require 'underscore'
shared = {}

To render a post, we take its raw source, treat it as both an Underscore template (for HTML generation) and as Markdown (for formatting), and insert it into the layout as content.

Journo.render = (post, source) ->
  catchErrors ->
    do loadLayout
    source or= fs.readFileSync postPath post
    variables = renderVariables post
    markdown  = _.template(source.toString()) variables
    title     = detectTitle markdown
    content   = marked.parser marked.lexer markdown
    shared.layout _.extend variables{titlecontent}

A Journo site has a layout file, stored in layout.html, which is used to wrap every page.

loadLayout = (force) ->
  return layout if not force and layout = shared.layout
  shared.layout = _.template(fs.readFileSync('layout.html').toString())

Publish to Flat Files

A blog is a folder on your hard drive. Within the blog, you have a posts folder for blog posts, a public folder for static content, a layout.html file for the layout which wraps every page, and a journo.json file for configuration. During a build, a static version of the site is rendered into the site folder, by rsyncing over all static files, rendering and writing every post, and creating an RSS feed.

fs = require 'fs'
path = require 'path'
{spawnexec} = require 'child_process' = ->
  do loadManifest
  fs.mkdirSync('site') unless fs.existsSync('site')
  exec "rsync -vur --delete public/ site"(err, stdout, stderr) ->
    throw err if err
    for post in folderContents('posts')
      html = Journo.render post
      file = htmlPath post
      fs.mkdirSync path.dirname(file) unless fs.existsSync path.dirname(file)
      fs.writeFileSync filehtml
    fs.writeFileSync "site/feed.rss"Journo.feed()

The config.json configuration file is where you keep the configuration details of your blog, and how to connect to the server you'd like to publish it on. The valid settings are: title, description, author (for RSS), url, publish (the user@host:path location to rsync to), and publishPort (if your server doesn't listen to SSH on the usual one).

An example config.json will be bootstrapped for you when you initialize a blog, so you don't need to remember any of that.

loadConfig = ->
  return if shared.config
    shared.config = JSON.parse fs.readFileSync 'config.json'
  catch err
    fatal "Unable to read config.json"
  shared.siteUrl = shared.config.url.replace(/\/$/'')

Publish via rsync

Publishing is nice and rudimentary. We build out an entirely static version of the site and rysnc it up to the server.

Journo.publish = ->
  rsync 'site/images/'path.join(shared.config.publish'images/')->
    rsync 'site/'shared.config.publish

A helper function for rsyncing, with logging, and the ability to wait for the rsync to continue before proceeding. This is useful for ensuring that our any new photos have finished uploading (very slowly) before the update to the feed is syndicated out.

rsync = (from, to, callback) ->
  port = "ssh -p #{shared.config.publishPort or 22}"
  child = spawn "rsync"['-vurz''--delete''-e'portfromto]
  child.stdout.on 'data'(out) -> console.log out.toString()
  child.stderr.on 'data'(err) -> console.error err.toString()
  child.on 'exit'callback if callback

Maintain a Manifest File

The "manifest" is where Journo keeps track of metadata -- the title, description, publications date and last modified time of each post. Everything you need to render out an RSS feed ... and everything you need to know if a post has been updated or removed.

manifestPath = 'journo-manifest.json'
loadManifest = ->
  do loadConfig
  shared.manifest = if fs.existsSync manifestPath
    JSON.parse fs.readFileSync manifestPath
  do updateManifest
  fs.writeFileSync manifestPathJSON.stringify shared.manifest

We update the manifest by looping through every post and every entry in the existing manifest, looking for differences in mtime, and recording those along with the title and description of each post.

updateManifest = ->
  manifest = shared.manifest
  posts = folderContents 'posts'
  delete manifest[postfor post of manifest when post not in posts
  for post in posts
    stat = fs.statSync postPath post
    entry = manifest[post]
    if not entry or entry.mtime isnt stat.mtime
      entry or= {pubtime: stat.ctime}
      entry.mtime = stat.mtime
      content = fs.readFileSync(postPath post).toString()
      entry.title = detectTitle content
      entry.description = detectDescription contentpost
      manifest[post= entry

Retina Ready

In the future, it may make sense for Journo to have some sort of built-in facility for automatically downsizing photos from retina to regular sizes ... But for now, this bit is up to you.

Syntax Highlight Code

We syntax-highlight blocks of code with the nifty highlight package that includes heuristics for auto-language detection, so you don't have to specify what you're coding in.

{Highlight} = require 'highlight'
  highlight: (code, lang) ->
    Highlight code

Publish a Feed

We'll use the rss module to build a simple feed of recent posts. Start with the basic author, blog title, description and url configured in the config.json. Then, each post's title is the first header present in the post, the description is the first paragraph, and the date is the date you first created the post file.

Journo.feed = ->
  RSS = require 'rss'
  do loadConfig
  config = shared.config
  feed = new RSS
    title: config.title
    description: config.description
    feed_url: "#{shared.siteUrl}/rss.xml"
    site_url: shared.siteUrl
  for post in sortedPosts()[0...20]
    entry = shared.manifest[post]
      title: entry.title
      description: entry.description
      url: postUrl post
      date: entry.pubtime

Quickly Bootstrap a New Blog

We init a new blog into the current directory by copying over the contents of a basic bootstrap folder.

Journo.init = ->
  here = fs.realpathSync '.'
  if fs.existsSync 'posts'
    fatal "A blog already exists in #{here}"
  bootstrap = path.join(__dirname'bootstrap')
  exec "rsync -vur --delete #{bootstrap} ."(err, stdout, stderr) ->
    throw err if err
    console.log "Initialized new blog in #{here}"

Preview via a Local Server

Instead of constantly rebuilding a purely static version of the site, Journo provides a preview server (which you can start by just typing journo from within your blog).

Journo.preview = ->
  http = require 'http'
  mime = require 'mime'
  url = require 'url'
  util = require 'util'
  do loadManifest
  server = http.createServer (req, res) ->
    rawPath = url.parse(req.url).pathname.replace(/(^\/|\/$)/g'') or 'index'

If the request is for a preview of the RSS feed...

if rawPath is 'feed.rss'
      res.writeHead 200'Content-Type': mime.lookup('.rss')
      res.end Journo.feed()

If the request is for a static file that exists in our public directory...

      publicPath = "public/" + rawPath
      fs.exists publicPath(exists) ->
        if exists
          res.writeHead 200'Content-Type': mime.lookup(publicPath)
          fs.createReadStream(publicPath).pipe res

If the request is for the slug of a valid post, we reload the layout, and render it...

          post = "posts/#{rawPath}.md"
          fs.exists post(exists) ->
            if exists
              loadLayout true
              fs.readFile post(err, content) ->
                res.writeHead 200'Content-Type': 'text/html'
                res.end Journo.render postcontent

Anything else is a 404. (Does anyone know a cross-platform equivalent of the OSX open command?)

              res.writeHead 404
              res.end '404 Not Found'
  server.listen 1234
  console.log "Journo is previewing at http://localhost:1234"
  exec "open http://localhost:1234"

Work Without JavaScript, But Default to a Fluid JavaScript-Enabled UI

The best way to handle this bit seems to be entirely on the client-side. For example, when rendering a JavaScript slideshow of photographs, instead of having the server spit out the slideshow code, simply have the blog detect the list of images during page load and move them into a slideshow right then and there -- using alt attributes for captions, for example.

Since the blog is public, it's nice if search engines can see all of the pieces as well as readers.

Finally, Putting it all Together. Run Journo From the Terminal

We'll do the simplest possible command-line interface. If a public function exists on the Journo object, you can run it. Note that this lets you do silly things, like journo toString but no big deal. = ->
  command = process.argv[2or 'preview'
  return do Journo[commandif Journo[command]
  console.error "Journo doesn't know how to '#{command}'"

Let's also provide a help page that lists the available commands. = Journo['--help'= ->
  console.log """
    Usage: journo [command]
    If called without a command, `journo` will preview your blog.
    init      start a new blog in the current folder
    build     build a static version of the blog into 'site'
    preview   live preview the blog via a local server
    publish   publish the blog to your remote server

And we might as well do the version number, for completeness' sake.

Journo.version = Journo['--version'= ->
  console.log "Journo 0.0.1"

Miscellaneous Bits and Utilities

Little utility functions that are useful up above.

The file path to the source of a given post.

postPath = (post) -> "posts/#{post}"

The server-side path to the HTML for a given post.

htmlPath = (post) ->
  name = postName post
  if name is 'index'

The name (or slug) of a post, taken from the filename.

postName = (post) -> path.basename post'.md'

The full, absolute URL for a published post.

postUrl = (post) -> "#{shared.siteUrl}/#{postName(post)}/"

Starting with the string contents of a post, detect the title -- the first heading.

detectTitle = (content) ->
  _.find(marked.lexer(content)(token) -> token.type is 'heading')?.text

Starting with the string contents of a post, detect the description -- the first paragraph.

detectDescription = (content, post) ->
  desc = _.find(marked.lexer(content)(token) -> token.type is 'paragraph')?.text
  marked.parser marked.lexer _.template("#{desc}...")(renderVariables(post))

Helper function to read in the contents of a folder, ignoring hidden files and directories.

folderContents = (folder) ->
  fs.readdirSync(folder).filter (f) -> f.charAt(0) isnt '.'

Return the list of posts currently in the manifest, sorted by their date of publication.

sortedPosts = ->
  _.sortBy _.without(_.keys(shared.manifest)'')(post) ->

The shared variables we want to allow our templates (both posts, and layout) to use in their evaluations. In the future, it would be nice to determine exactly what best belongs here, and provide an easier way for the blog author to add functions to it.

renderVariables = (post) ->
    posts: sortedPosts()
    post: path.basename(post)
    manifest: shared.manifest

Quick function which creates a link to a Google Map search for the name of the place.

mapLink = (place, additional = '', zoom = 15) ->
  query = encodeURIComponent("#{place}#{additional}")
  "<a href=\"{query}&t=h&z=#{zoom}\">#{place}</a>"

Convenience function for catching errors (keeping the preview server from crashing while testing code), and printing them out.

catchErrors = (func) ->
  try do func
  catch err
    console.error err.stack

Finally, for errors that you want the app to die on -- things that should break the site build.

fatal = (message) ->
  console.error message
  process.exit 1