Elm developer and educator. Founder of Incremental Elm Consulting.
December 10th, 2019
A is for API - Introducing Static HTTP Requests
I'm excited to announce a new feature that brings
If you’ve tried
, you may be thinking, "elm-pages hydrates into a full Elm app... so couldn’t you already make HTTP requests to fetch API data, like you would in any Elm app?" Very astute observation! You absolutely could.
So what's new? It all comes down to these key points:
Let's dive into these points in more detail.
Let's break down how you perform HTTP requests in vanilla Elm, and compare that to how you perform a Static HTTP request with
Anatomy of HTTP Requests in Vanilla Elm
Cmd for an HTTP request on init (or update)
You receive a
with the payload
Store the data in
Tell Elm how to handle
s (including JSON decoding failures)
Anatomy of Static HTTP Requests in elm-pages
function specifies some
data, and a function to turn that data into your
tags for that page
That's actually all of the boilerplate for
There is a lifecycle, because things can still fail. But the entire Static HTTP lifecycle happens
before your users have even requested a page
. The requests are performed at build-time, and that means less boilerplate for you to maintain in your Elm code!
Let's see some code!
Here's a code snippet for making a StaticHttp request. This code makes an HTTP request to the Github API to grab the current number of stars for the
The data is baked into our built code, which means that the star count will only update when we trigger a new build. This is a common JAMstack technique. Many sites will trigger builds periodically to refresh data. Or better yet, use a webhook to trigger new builds whenever new data is available (for example, if you add a new blog post or a new page using a service like Contentful).
Notice that this app's
function are not involved in the process at all! It's also worth noting that we are passing that data into our
function, which allows us to use it in our
tags for the page.
functions are very similar to Elm libraries you've likely used already, such as
. If you don't depend on any StaticHttp data, you use
, similar to how you might use
This is actually the same as our previous example that had a
, except that it doesn't make a request or have the stargazer count data.
A common pattern is to use environment variables in your local environment or your CI environment in order to securely manage auth tokens and other secure data.
provides an API for accessing this data directly from your environment variables. You don't need to wire through any flags or ports, simply use the
module (see the docs for more info)
. It will take care of masking the secret data for you so that it won't be accessible in the bundled assets (it's just used to perform the requests during the build step, and then it's masked in the production assets).
The Static HTTP Lifecycle
If you have a bad auth token in your URL, or your JSON decoder fails, then that code will never run for your
site. Instead, you'll get a friendly
build-time error telling you exactly what the problem was and where it occurred (as you're familiar with in Elm).
These error messages are inspired by Elm's famously helpful errors. They're designed to point you in the right direction, and provide as much context as possible.
Which brings us to our next key point...
Static HTTP requests are performed at build-time. Which means that if you have a problem with one of your Static HTTP requests,
your users will never see it
. Even if a JSON decoder fails,
will report back the decode failure and wait until its fixed before it allows you to create your production build.
Your API might go down, but your Static HTTP requests will always be up (assuming your site is up). The responses from your Static requests are baked into the static files for your
build. If there is an API outage, you of course won't be able to rebuild your site with fresh data from that API. But you can be confident that, though your build may break, your site will always have a working set of Static HTTP data.
Compare this to an HTTP request in a vanilla Elm app. Elm can guarantee that you've handled all error cases. But you still need to handle the case where you have a bad HTTP response, or a JSON decoder fails. That's the best that Elm can do because it can't guarantee anything about the data you'll receive at runtime. But
make guarantees about the data you'll receive! Because it introduces a new concept of data that you get a snapshot of during your build step.
guarantees that this frozen moment of time has no problems before the build succeeds, so we can make even stronger guarantees than we can with plain Elm.
The StaticHttp API also comes with some significant performance boosts. StaticHttp data is just a static JSON file for each page in your
site. That means that:
No waiting on database queries to fetch API data
Your site, including API responses, is just static files so it can be served through a blazing-fast CDN (which serves files from the nearest server in the user's region)
Scaling is cheap and doesn't require an Ops team
intelligently prefetches the Static HTTP data for a page when you're likely to navigate to that page, so page loads are instant and there's no spinner waiting to load that initial data
JSON data, stripping out everything but what you use in your JSON decoder
The JSON optimization is made possible by a JSON parsing library created by Ilias Van Peer. Here's the pull request where he introduced the JSON optimization functionality:
Let's take our Github API request as an example. Our Github API request from our previous code snippet (
) has a payload of 5.6KB (2.4KB gzipped). That size of the optimized JSON drops down to about 3% of that.
You can inspect the network tab on this page and you'll see something like this:
If you click on Github API link above and compare it, you'll see that it's quite a bit smaller! It just has the one field that we grabbed in our JSON decoder.
This is quite nice for privacy and security purposes as well because any personally identifying information that might be included in an API response you consume won't show up in your production bundle (unless you were to explicitly include it in a JSON decoder).
Comparing StaticHttp to other JAMstack data source strategies
You may be familiar with frameworks like Gatsby or Gridsome which also allow you to build data from external sources into your static site. Those frameworks, however, use a completely different approach,
using a GraphQL layer to store data from those data sources
, and then looking that data up in GraphQL queries from within your static pages.
This approach makes sense for those frameworks. But since
is built on top of a language that already has an excellent type system, I wanted to remove that additional layer of abstraction and provide a simpler way to consume static data. The fact that Elm functions are all deterministic (i.e. given the same inputs they will always have the same outputs) opens up exciting new approaches to these problems as well. One of Gatsby's stated reasons for encouraging the use of their GraphQL layer is that it allows you to have your data all in one place. But the
StaticHttp API gives you similar benefits, using familiar Elm techniques like
, etc to massage your data into the desired format.
I'm looking forward to exploring more possibilities for using static data in
. Some things I plan to explore are:
Programatically creating pages using the Static HTTP API
Configurable image optimization (including producing multiple dimensions for
s) using a similar API
Optimizing the page metadata that is included for each page (i.e. code splitting) by explicitly specifying what metadata the page depends on using an API similar to StaticHttp
Getting started with StaticHttp
take a look at this an end-to-end example app that uses the new
to get started.
Or just use the
and start building something cool! Let me know your thoughts on Slack, I'd love to hear from you! Or continue the conversation on Twitter!