Building a GUI for managing middleman blogs

the site will still be fast, really fast

Jul 30, 2015

Can we make static site generators work like a normal site? Lets take a look at how we'd start to implement Slow Data and Fast Sites using the example of building an admin UI for a Middleman blog.


Given a middleman app with middleman-blog and middleman-blog-drafts configured:

  1. Add middleman-blog-ui in your Gemfile. This document describes how middleman-blog-ui works.
  2. Add activate :blog_ui in config.rb
  3. Start middleman server
  4. Visit http://localhost:4567/admin.

And now you are living in the fabulous world where you can, from your browser:

  • Edit existing drafts and posts
  • Create new drafts
  • Publish drafts into posts
  • Run some basic git commands.
  • Run middleman build
  • Run middleman deploy

Why do this

This is explained in more detail in the Slow Data and Fast Sites post, but basically what we're trying to do it seperate out runtime tooling with development time tooling. Future work will be around differentiating when components are rendered and from which data source.

Static sites have a huge runtime benefit, but actually setting them up and running them can really suck unless you already know how to do it. This is a proof of concept of how to add a CMS to a static site so that it's easier to manage. This doesn't address the "pain in the ass to setup" part, but it helps with the usability.

Lets walk through what we're doing if we were to implement this in your own app directly. The gem packages this up with a slightly different directory scheme, but the code below is the same but slightly easier to understand.

The basic idea

We're going to add a static JavaScript application in the source/admin directory which will interact with the local filesystem and middleman install. This application will consume data from your middleman app on the local system, in the "development" environment, such as the lists of published and draft articles, and will interact with a API server that will be mounted as part of the middleman preview server. This will allow the Javascript app to change the filesystem, which will in turn be monitored and rendered by the middleman application.

Examples of the commands that can be triggered are:

  • create new draft
  • publish draft into article
  • git status
  • bundle exec middleman build

G<!-- app -->appStatic Middleman Site<!-- admin -->adminAdmin JS FrontEnd App<!-- admin->app -->admin->app<!-- api -->apiSinatra API Server<!-- admin->api -->admin->api<!-- api->app -->api->app<!-- cmds -->cmdsMiddleman Commands<!-- api->cmds -->api->cmds<!-- cmds->app -->cmds->app

Let go through each of the components as if we were building them in our app directly. The gem does a lot of this for you, but it's always nice to see how it works.

Static Middleman Site

We're going to add our files in source/admin, and source/javascript/admin and source/stylesheets/admin.scss. All of these files will be rendered in development mode, so when the server is running locally and you load up http://localhost:4567/admin we can use the middleman build process to load up our UI. This, in turn, will use AJAX to pull in data from the running middleman instance.

Inside of config.rb, we tell middleman about our pages, and we tell the build process to ignore these files as part of the build process since we don't want to release them into production:

page "/admin/*html", layout: "admin/layout"

# Build-specific configuration
configure :build do
  ignore '/admin/*'
  ignore '/stylesheets/admin/*'
  ignore '/javascripts/admin/*'

If they were not ignored, then you would be deploying a read-only version of the admin tools, since the production build would include the javascript application but not the API which is needed to make changes.

This will allow us to create files like /source/admin/drafts.json.erb:

<% d = drafts.collect do |d|
    { path: d.path, title: d.title }
  {drafts: d }.to_json

That we can load from our JS app like:

$.get( '/admin/drafts.json' ).then( function( data ) { console.log( data ) } );

Note that when we do middleman build these files will not be generated, since we've specified them as ignore. These files only are in the :development environment.

The Sinatra App

That's fine for static data, or data that can be exposed using normal middleman ways of doing things, but we'll also want to be able to interact with the filesystem and run commands. It's also going to be useful to interact with the middleman Sitemap as it understands it, so we can build a simple sinatra app that acts as a bridge. The current version of the code I'm using for this blog can be found on github.

First we need to require and mount our app on config.rb. This is done like so:

require 'lib/apiserver.rb'

map "/api" do
  run ApiServer

The commands we want to support are:

  • get /api/post - load a specific post and metadata
  • post /api/post - save a post and update metadata
  • post /api/images - upload an image (for drag and drop) (not implemented)
  • post /api/draft - create a new draft post
  • post /api/publish - change a draft into an article
  • post /api/update - run git pull origin master 2>&1
  • post /api/diff - run git diff 2>$1
  • post /api/status - run git status 2>&1
  • post /api/build - run bundle exec middleman build 2>&1
  • post /api/deploy - run bundle exec middleman deploy 2>&1

To reiterate right now the code is in more of a proof of concept, butso please feel free to make suggestions. Lets take a look at a few of these methods in lib/api_server.rb:

require 'sinatra/base'
require 'sinatra/json'

class ApiServer < Sinatra::Base
  get '/' do
    "This is the api server"

  get '/post' do
    app = load_app "Looking up #{params[:path]}"

    file = app.sitemap.find_resource_by_path params[:path] if params[:path]

    if !file
      status 404
      json error: "Unknown path #{params[:path]}"
      raw = file.source_file
      body = raw.gsub( /^---\n.*?---\n*/m, "" ) # Remove the preyaml

      json meta:, content: body

  def load_app
    opts = {}

    app = ::Middleman::Application.server.inst do
      set :environment, opts[:environment].to_sym if opts[:environment]



  def logger
    ::Middleman::Logger.singleton( 1 )

The Javascript App

I also wanted to see how to build this in React and Reflux to start exploring how we can start to add build-type rendering into the equation. While this doesn't do any server side rendering, its still good to explore how a React app actually works.

The main html file simple calls React.render( React.createElement(App, null), document.getElementById('content') ); to setup the page. The App component works off of the pathStore component, which keeps track of what page the user is currently editing. Lets take a quick look at how the DashboardDraftList component works:

@DashboardDraftList = React.createClass
  mixins: [Reflux.connect(draftStore)],

  getInitialState: ->
    drafts: []

  componentDidMount: -> updateDraftList()

  render: ->
    drafts = (item) ->
      <li key={item.path}><a onClick={viewPath.bind(this, item.path)}>{item.title}</a></li>

    <div className="maincontent">

      <ul className="nav nav-pills nav-stacked">

The first thing that this does is wire itself up to the draftStore store which manages the communication to the server and the state received from the server. When this changes, draftStore will trigger a state change and all components that have been connected will receive a setState message, which will then call the render method of the component keeping the UI up to date. As part of the React component life cycle componentDidMount, we call the updateDraftList() action when the component is first loaded on the page. This is defined in drafts store:

@updateDraftList = Reflux.createAction
  asyncResult: true

@draftStore = Reflux.createStore
  init: ->
    @drafts = []
    @listenTo( updateDraftList, @onUpdateDrafts )

  onUpdateDrafts: ->
    request.get "/admin/drafts.json", (err, response) =>
      if response.ok
        updateDraftList.completed( response.body )
        @drafts = response.body
        @trigger @drafts

The first thing we do is to define and export a Reflux action called updateDraftList. This action is what the componentDidMount function triggeres. Inside of the draftStore we listen for that action, and when it get it, call onUpdateDrafts. This uses request.get to load up the json from /admin/drafts.json and then pushes it out to all of the components that have subscribed to it.

The other components are defined here and the different stores can be found here. This is my first real React app so I'm sure there's cleaner/better ways of doing things...

Putting it all together

When I was working on this I did it out of the source directory of this blog, but I wanted to package it up into a gem so that other folks could use it. I pulled the relavent code out from into middleman-blog-ui/source and then created a middleman extension to put the various templates in the sitemap, tell sprockets where to find the javascript files, and setup coffee script.

The full code is found on github, but to explain how it works:

  1. Define SOURCE_DIR
  2. Map the sinatra app ApiServer in the extentions initialize method, using
  3. Register the CoffeeScript processing using ::Sprockets.register_preprocessor 'application/javascript', ::Sprockets::CoffeeReact
  4. Use after_configuration to register the application templates, by creating Middleman::Sitemap::Resource objects mapping the sitemap urls with the filesytem paths.
  5. Adding each of the javascript and stylesheet files to app.sprockets.

Next steps

This is mainly a proof of concept, but I'm using it right now to write this post. The app itself needs

  1. A decent UI
  2. Support for other static generators other than middleman
  3. A concept of users
  4. Shared drafts
  5. Better error handling

If you are at all interested in this, please let me know!