Thursday, 7 December, 2017 UTC


Summary

Gatsby has been getting a lot of recognition and adoption lately, and for good reason. It’s so flexible and it works well with nearly everything.
If you’re on a tight budget and don’t want to sacrifice developer experience or cutting-edge deployments, I’ve landed on a favorite set of tools (Gatsby included, of course) for developing static sites that solves multiple problems at once.
The use case I’ll be covering is a documentation site for our company’s main software product. The site contains a lot of content with hundreds of articles.
Here’s a list of my requirements for this site:
  • Speed - both in development and site performance
  • Ease of use - both in developer experience and content upkeep
  • Searchable content - it is a doc site after all
  • Inexpensive Hosting - maximize value (who doesn’t want this?)
  • Automated continuous deployment
The following is a report on my high-level experience of using Gatsby with Contentful, Netlify and Algolia and the problems they solve — without getting down into code.
Static to the Rescue?
I know it’s bad to make assumptions, but I’m going to assume that if you’re here and reading this, you at least know a little about the perks of static sites and the JAMstack. If not, check out https://jamstack.org/ for a quick breakdown on why web development is, for lack of a better phrase, “going back to the basics.”
Having dealt with my share of CMS headaches with Joomla in the almost-forgotten past, and Wordpress in the more recent years, I’ve been on a quest to simplify things. I don’t want to worry about having a plugin or theme get hacked or the constant nagging to install updates. I also would prefer to not deal with themes at all and just have flexible building blocks to shape my site’s appearance via my own codebase. Static sites excel at these things.
But while static sites give you lots of perks, they do create some extra hurdles over going the traditional route.
via GIPHY
While they’re generally fast out of the box, static sites don’t make their content easily editable. After all, a static site’s content is usually just that — static. That means, traditionally, you would have to edit the page code to make content changes or add a markdown file then fire off a site rebuild command, and finally redeploy. Even though static site generators have solved this in many clever ways, I feel Gatsby solves this problem in a particularly elegant fashion via it’s GraphQL data layer (more on that later) and its vast ecosystem of data source plugins.
Before I jump into the topic of content and “data”, I want to briefly say that building a static site template with React-based architecture and hot module reloading is just plain fun. Gatsby’s CLI gets you going so quickly. It really is a joy to use. Judging from all the twitter comments saying the same thing, I think that’s a common consensus.
Okay, now back to some static hurdles.
A Problem of Content
Our site has a lot of content (~300 articles) that needs to be maintained by non-developers, my co-workers. This meant we needed an approachable interface for copy and content editing. I wanted to make it as convenient as logging into Wordpress and publishing from there, without the Wordpress. So the publishing experience couldn’t rely on creating a file and committing changes to a git repo.
Sidebar: There is Gatsby-Source-Wordpress plugin that pulls in content via a Wordpress API. However, to me, this was not appealing because I was trying to avoid hosting a traditional CMS entirely.
Solution: Contentful + Gatsby
Contentful is a hosted headless CMS with a fantastic user experience. It’s similar to having a backend like Wordpress, but you are fully responsible for the front-end layer. The beauty of Contentful is threefold.
  • Intuitive and attractive UI
  • Simple content modeling
  • Free tier
Dealing with the back-end of Contentful is refreshing and the content modeling really leads the pack when compared to other headless content management systems. It doesn’t feel like something that just gets the job done, it’s actually really nice to use. They also just pushed some great new changes that made it even easier to search and filter our articles on the back-end.
Contentful also happens to offer a generous free tier with useful features for a small company or a few projects. At the time of this writing, you get a few spaces (think projects), ten thousand content records, and five users that can be admins, editors or authors of content. All they ask in return is that you display a Contentful logo on your site footer or attribute them in your code repo.
With that salesy pitch out of the way, how does Contentful work within our Gatsby site?
Our documentation exists in 40 different parent topics and numerous articles in each topic. The largest challenge with this was creating topic-based navigation.
Gatsby’s data handling makes these problems easy to manage by simplifying how you get data to your site from external sources. It’s not entirely unique from other static site generators in that regard — other generators utilize plugins for grabbing content as well, but how you deal with actually pulling it into your React components/pages with GraphQL is beautiful.
After you install the gatsby-source-contentful plugin with NPM and add your Contentful API credentials to the gatsby-config file, the fun begins.
Every time you run Gatsby with the develop command or do a new build with the build command, the plugin fetches any new content from the Contentful API. All this data is then available and ready to query locally in your development environment. This means you can start pulling in Contentful assets and content (assets = images/media, content = pages/articles/text/markdown, etc.) using GraphQL queries right inside the template files.
Sidenote: I had created a blog for my wife with Gatsby prior to this doc site, so I had a little experience with the Gatsby APIs. But I still consider myself a complete newbie when it comes to GraphQL. Lucky for me, Gatsby’s tutorials and community are awesome at answering questions or handling general usage issues.
In a single GraphQL query I was able to pull in all the topics and related article titles set up in my Contentful Content Model for navigation. By leveraging React and some GraphQL, I was able to create a dynamically generated sidebar menu based off the content that got pulled in from Contentful. I have to say, it feels so freeing being about to create static content with dynamic data like this.
The articles themselves are written in markdown in the Contentful editor. They get converted to HTML via a transformer plugin within Gatsby. The markdown editing in Contentful is quite practical with standard WYSIWYG-like editor features. I haven’t heard any complaints from my co-workers.
Searching for a Search Solution
Another “problem” with static sites is the lack of out-of-the-box site search. Most search implementations occur between the server and the database. As a doc site, users typically expect solid search functionality. There are a few frontend only javascript search libraries (like lunr.js) that take a search query and parse a pre-built JSON index of your content.
I could have created this index by tying into the onPostBuild Gatsby API. This event fires after all the pages are generated, so all page nodes are available to parse to create a search index.
I quickly decided that this approach wouldn’t work well for our case because of the vast number of articles. The index file alone would have been rather large and present a hefty chunk of the site download which I felt was antithetical to the performance benefits of using Gatsby (or even any static site). I needed something that would operate client-side, but that had the search brains still residing in the cloud somewhere. Though an option, I didn’t have time to build my own solution.
Solution: Algolia + Gatsby
This solution was a bit of a trial and error process for me. I had seen many doc sites use Algolia online in my dev travels. I knew it offered a usable free tier (again with logo attribution) with what seemed like enough free API calls to suit our user base. What I didn’t know was how I would get all my content indexed properly. Algolia does have helpful documentation on indexing.
It was more of a challenge of how to break down the content in the articles into the chunks that the index requires. The Algolia Docs state that your index records shouldn’t be more than 10kb each which roughly equates to a paragraph or two each. This suddenly became a challenge in how to parse my article content by section. There wasn’t a good example out there of how to accomplish this.
Eventually I settled on an HTML to JSON library that turned the page hierarchy into a parseable JSON object. I set up a script in the onPostBuild Gatsby event API that took the built HTML from each article. The library did the magic of converting the HTML to JSON, then I iterated through the JSON. While keeping track of the last linked heading (h tag), I set the index record’s page link accordingly for each article section. The index was then uploaded to Algolia via their node.js client.
It wasn’t pretty, but it worked.
I ended up coupling this indexing method with React InstantSearch. This is Algolia’s official React component library for utilizing with their service. The final product was a search input with highlighted suggested results that allowed the user to click one of those results and be taken directly to a parent heading on a particular article.
Not bad.
Though after I wired all this up, it turned out that I was having some issues with my implementation that left me seeking help from support. I was receiving emails about quota usage, and I was sure I wasn’t even coming close to actual usage limits.
This was ironically about the same time I discovered DocSearch by Algolia. So as any good developer would do, I scrapped all my hard work and just signed up for DocSearch. Long story short is that they crawl your site every 24 hours, updating the index for you. You include a script tag that ties their API to your search input. Style it with some CSS. Bam. Done.
And it worked better out of the box than my implementation did. I felt kinda stupid after going through all the effort because I realized that the Reactjs.org repo had the answer right in their source. They use DocSearch instead of rolling their own indexing and search interface. Oh well.
Hosting & Deployment
A great thing about static sites is that you can host them just about anywhere. You get a folder of pre-built files that you can throw up on any web server and you’re done. You can even dump it in an Amazon S3 bucket and save a ton of money with very little work.
So while hosting is easy, static sites adds a manual step for deploying changes made to a site’s code or content — unlike WordPress or other traditional CMSes where a content change goes live immediately.
If you don’t configure some form of deployment automation, you have to manually initiate a build and upload it yourself. I wanted a continuous development workflow — push a commit to my code repo and have Gatsby run its build in the cloud and automatically deploy the new version of the site to a hosting provider.
Can this be done with AWS? Sure, with a bit of configuration and more grunt work in setup. Can it be done elsewhere with little to no configuration? How about for free?
Luckily I already knew the answers to these questions because I had already discovered Netlify for some other projects.
Solution: Netlify + Gatsby
Plugging any static site into the Netlify workflow is a no brainer, but after I found Gatsby, I knew there was no other option for me. These two pair so well together!
Netlify recently changed their pricing to improve what was already an awesome hosting per dollar value. I can’t get through this section without a bullet list of why Netlify is so fantastic.
  • Free to use for personal/commercial project (seriously great free tier)
  • Push button HTTPS via Let’s Encrypt built in
  • Fast Global CDN
  • Support for custom domains
  • Atomic deploys
  • A crazy-cool, integrated build engine
  • And a lot more…
  • Did I mention you get all this for FREE?
via GIPHY
And back to using it with Gatsby.
After you link your Netlify site to a specific code repository, Netlify’s build bots take care of the rest. From that point on, when you push a change to your repo, the build bot says, “Hey look! a change! I need to run the gatsby build command…”, then it follows the package.json (or yarn file) in the repo and downloads the necessary dependencies if they aren’t already cached, and then builds out the static site.
And during this build process, Gatsby’s clever APIs are taking care of grabbing content from Contentful and generating the static pages for the articles. Awesome. When this process is done, you can even be notified via Slack, or Email.
Netlify is the easy-button for your deploy/hosting problem.
Coupled with your Gatsby site, the site performance is outstanding. Both perceived performance and measurable performance are there. You get time-to-first-byte measurements in the single digits. The code-splitting and pre-fetching perks of Gatsby also help to make your site score well in web performance tests. All out of the box.
Now, to come full circle I needed something that would trigger a new site build when someone edited or added an article from Contentful. Again, Contentful and Netlify to the rescue.
Contentful has webhook functionality built-in and lets you fire off a request when an action is taken on a piece of content or when content is created. Perfect. With this hook, Contentful talks to Netlify when there’s a change or addition, and Netlify rebuilds the site and deploys it.
It’s a match made in JAMstack heaven.
Is it Magic?
Gatsby makes building the site fun and painless. It gets out of the way and lets you be creative with your craft — yet also adds some insane perks, like responsive image handling and lazy loading, with very little effort on your part. Contentful lets you focus on your content the same way Gatsby lets you focus on developing, and Netlify really “just works.” It lets you hit a few buttons and leaves you with a “Is it really this easy?” kind of feeling.
Now, hopefully, our customers will feel the same with our site.
via GIPHY