Static websites load blazing fast and are fun to make. Having your website calculated for each request is getting out of style and is becoming less necessary.
I went through the process of turning the WordPress version of this blog into a statically served website. This article explains the process behind it and what I did to achieve the current static website.
WordPress was my platform of choice back in 2008 when I started writing and publishing articles. Back then I liked the comfort of not having to write my own content management system for posting blog posts. The amount of plugins available helped me implement stuff I didn’t really know how to make.
Years and a lot of web development related experience later I learned that WordPress is not for me. I explained why in a previous post about rebranding Gaya Design.
With a static approach I could make my blog behave the way I think it should behave. Calculate the content before it will is served, not at the time it is requested, like what happens on a typical blog system. You can kind of prevent calculations using caching tools, but it still is no holy grail.
It was a nice challenge for me, since I never really made a statically generated site before. My weblog was the perfect case for me to go and try it out. I was going to change my blog’s name anyway, so I might as well go ahead and create a new site along with it.
##Planning it all out
There were a few requirements my new weblog had to meet. Since I am quite picky on the quality of my own workflow I decided I wouldn’t make it easy for myself.
The things I really wanted were:
- Generate everything from the source. Do not have any compiled or generated content in my repository.
- Ability to write posts using MarkDown.
- Being able to test and write for my site locally and deploy it with the same content.
- Deployment from my weblog’s repository. No uploading through FTP.
This meant that I had to go and search for a static site generator that meets these needs. If it had blogging support it would be even better.
##Looking for a Static Site Generator
To be more specific, I was really looking for a static blog generator. I played around with a few, one was Jekyll. Which is pretty nice, but I came to the conclusion that I wanted to make a Node.js powered blog. Ruby gems were also driving me nuts, and I am no Ruby specialist, so I stopped there.
I quickly came across Hexo, Assemble and Wintersmith, which are all fine frameworks in their own way.
Before I used Gulp I was using Grunt, and in that time I’ve had some experience using Assemble. Which also has a bit of Gulp support, but it only works half of the time and is not recommended for production environments. This options went out of the window pretty quickly.
When I started using Wintersmith I completely fell in love. It’s so quick and has a great programmable interface to work with. So that works perfectly if I want to integrate Wintersmith in my Gulp streamline.
The templating looked easy enough, and there is a large variety of Wintersmith template plugins available. Along with the dynamically generated content I was going to need, Wintersmith looked like a solid choice. Sorry Hexo! Looks like I didn’t even give you a chance.
##Creating the templates
To process my Sass I used gulp-sass. It runs on libsass, which makes it super fast too. That’s a great thing when you’re using BrowserSync because the newly generated assets will be available in the browser really quickly. Especially when you have two monitors where one holds a browser window and the other one your editor.
Frameworks I used for styling worth mentioning are: Neat for responsive layout grids and Bourbon for standard mixins.
For my SVG images on the site I used a SVG sprite generator. I choose not to include the SVGs as images in my DOM so I could keep styling of icons out of my templates. This way I can keep my HTML clean. The only problem was that I needed a way to calculate the SVG sprite width and height for each element to scale dynamically.
To generate the SVG sprite from a source folder I used svg-sprite, a Node.js module that gathers the SVGS and optimises the output. It also allows you to output a
.sass file to be used. So with that I generated a Sass file that calculated all the necessary widths and heights.
##Working with Wintersmith
The first thing I noticed about Wintersmith is that it, like a lot of other great Node.js projects, uses Jade. I don’t really like it that much, so I looked for an alternative. Nunjucks is a templating language heavily inspired by jinja2, which I was already familiar with. It looked fairly straight-forward, so I went with it. Looking at the examples of Wintersmith it was relatively easy to convert my HTML files to Wintersmith templates.
The only thing I had to do was to create a 404, about and contact page. That was it. Pretty harmless task since it already took care of most of my requirements.
As I said before, Wintersmith works really well as a Node.js module, so I created a watcher in Gulp that looks for changes and regenerates the whole blog on change. The great thing is that it only takes about 1 second for Wintersmith to do that.
##Converting content from WordPress to Wintersmith
Wintersmith is pretty clean in its article structure. It follows a convention that allow me to create a new article in a folder which contains the MarkDown file with the post. All the images that are used in the MarkDown file can be put in the same folder.
The challenge here was to get all the posts from WordPress, add the correct meta data in YAML format and convert the posts to MarkDown syntax while copying all the images. All the code snippets and internal URLs in the posts had to be changed as well.
I wrote a quick convert script for this and it saved me a lot of time. It was a dirty job, but I had to do it.
##Optimising all the assets
From the source folder all the assets had to be optimised before serving them to the browser. This happens after all the files are put in the folder that gets served. After which the server checks all the files and optimises them.
For image optimisation I used imagemin, a pretty quick module that minifies the images that get passed in. This way I don’t have to optimise the images myself before committing them.
To optimise the loading speed of my pages I extracted and inlined the critical-path CSS from my generated pages. This will allow the pages of my site to be loaded with visual content without blocking the loading with the external CSS file. To do this I used critical which scans my pages one by one and injects the correct inline CSS.
##Creating a server to serve my content
What I needed was a simple web server that could serve static files (which were build by my tasks), compress those files and create a fallback for pages that are not found. An easy way to get this done is to use express as the framework.
For serving content the serve-static middleware is really easy to use. For compression of the output I used compression. For the pages that are not found I wrote my own middleware that serves my own
404.html file I created.
All that was left was to build a middleware that handles my contact form submissions. I wrote up a little tool that handles the given data, puts it in an email and sends it to me. It responds with a simple JSON object telling the sender everything worked out.
##Deploying it all
I went a bit into deploying Node.js applications and came across dokku. It behaves like Heroku in a way that I can deploy my code on a repository that is hosted on my server and let dokku do the rest.
Once I push code to my dokku server through Git it will start to install all the dependencies defined in
package.json. It will than execute the npm scripts I defined in my
package.json file. It will build the site, optimise everything and test if it worked afterwards. Then it will switch my current running version with the new one. Perfect!
Now I can finally keep all my source the same on every machine and deploy my article whenever I want them to.
Have suggestions or found better ways to go static? Feel free to reach out on Twitter.