In this article, we explain the difference between single-page apps (SPAs) and static sites, and how we can bring the advantages of both worlds together in a dynamic JAMstack app using GatsbyJS and FaunaDB. We will build an application that pulls in some data from FaunaDB during build time, prerenders the HTML for speedy delivery to the client, and then loads additional data at run time as the user interacts with the page. This combination of technologies gives us the best attributes of statically-generated sites and SPAs. 

In short…<deep breath>…auto-scaling distributed websites with low latency, snappy user interfaces, no reloads, and dynamic data for everyone!

Heavy backends, single-page apps, static sites

In the old days, when JavaScript was new, it was mainly only used to provide effects and improved interactions. Some animations here, a drop-down there, and that was it. The grunt work was performed on the backend by Perl, Java, or PHP. 

This changed as time went on: client code became heavier, and JavaScript took over more and more of the frontend until we finally shipped mostly empty HTML and rendered the whole UI in the browser, leaving the backend to supply us with JSON data.

This led to a neat separation of concerns and allowed us to build whole applications with JavaScript, called Single Page Applications (SPAs). The most important advantage of SPAs was the absence of reloads. You could click on a link to change what’s displayed, without triggering a complete reload of the page. This in itself provided a superior user experience. However, SPAs increased the size of the client code significantly; a client now had to wait for the sum of several latencies:

• Serving latency: retrieving the HTML and JavaScript from the server where the JavaScript was bigger than it used to be 
• Data loading latency: loading additional data requested by the client
• Frontend framework rendering latency: once the data is received, a frontend framework like React, Vue, or Angular still has to do a lot of work to construct the final HTML 

A royal metaphor

We can analogize the loading a SPA with the building and delivery of a toy castle. The client needs to retrieve the HTML and JavaScript, then retrieve the data, and then still has to assemble the page. The building blocks are delivered, but they still need to be put together after they’re delivered.

If only there were a way to build the castle beforehand…

Enter the JAMstack

JAMstack applications consist of JavaScript, APIs and Markup. With today’s static site generators like Next.js and GatsbyJS, the JavaScript and Markup parts can be bundled up into a static package and deployed via a Content Delivery Network (CDN) that delivers files to a browser. A CDN geographically distributes the bundles, and other assets, to multiple locations. When a user’s browser fetches the bundle and assets, it can receive them from the closest location on the network, which reduces the serving latency. 

Continuing our toy castle analogy, JAMstack apps are different from SPAs in the sense that the page (or castle) is delivered pre-assembled. We have a lower latency since we receive the castle in one piece and no longer have to build it. 

Making static JAMstack apps dynamic with hydration

In the JAMstack approach, we start with a dynamic application and prerender static HTML pages to be delivered via a speedy CDN. But what if a fully static site is not sufficient and we need to support some dynamic content as the user interacts with individual components, without reloading the entire page? That’s where client-side hydration comes in.

Hydration is the client-side process by which the server-side rendered HTML (DOM) is “watered” by our frontend framework with event handlers and/or dynamic components to make it more interactive. This can be tricky because it depends on reconciling the original DOM with a new virtual DOM (VDOM) that’s kept in memory as the user interacts with the page. If the DOM and VDOM trees do not match, bugs can arise that cause elements to be displayed out of order, or necessitate rebuilding the page.

Luckily, libraries like GatsbyJS and NextJS have been designed so as to minimize the possibility of such hydration-related bugs, handling everything for you out-of-the-box with only a few lines of code. The result is a dynamic JAMstack web application that is simultaneously both faster and more dynamic than the equivalent SPA. 

One technical detail remains: where will the dynamic data come from?

Distributed frontend-friendly databases!

JAMstack apps typically rely on APIs (ergo the “A” in JAM), but if we need to load any kind of custom data, we need a database. And traditional databases are still a performance bottleneck for globally distributed sites that are otherwise delivered via CDN, because traditional databases are only located in one region. Instead of using a traditional database, we’d like our database to be on a distributed network, just like the CDN, that serves the data from a location as close as possible to wherever our clients are. This type of database is called a distributed database. 

In this example, we’ll choose FaunaDB since it is also strongly consistent, which means that our data will be the same wherever my clients access it from and data won’t be lost. Other features that work particularly well with JAMstack applications are that (a) the database is accessed as an API (GraphQL or FQL) and does not require you to open a connection, and (b) the database has a security layer that makes it possible to access both public and private data in a secure way from the frontend. The implications of that are we can keep the low latencies of JAMstack without having to scale a backend, all with zero configuration. 

Let’s compare the process of loading a hydrated static site with the building of the toy castle. We still have lower latencies thanks to the CDN, but also less data since most the site is statically generated and therefore requires less rendering. Only a small part of the castle (or, the dynamic part of the page) needs to be assembled after it has been delivered:

Example app with GatsbyJS & FaunaDB

Let’s build an example application that loads data from FaunaDB at build time and renders it to static HTML, then loads additional dynamic data inside the client browser at run time. For this example, we use GatsbyJS, a JAMstack framework based on React that prerenders static HTML. Since we use GatsbyJS, we can code our website completely in React, generate and deliver the static pages, and then load additional data dynamically at run time. We’ll use FaunaDB as our fully managed serverless database solution. We will build an application where we can list products and reviews. 

Let’s look at an outline of what we have to do to get our example app up and running and then go through every step in detail.

1. Set up a new database
2. Add a GraphQL schema to the database
3. Seed the database with mock-up data
4. Create a new GatsbyJS project
5. Install NPM packages
6. Create the server key for the database
7. Update GatsbyJS config files with server key and new read-only key
8. Load the pre-rendered product data at build time
9. Load the reviews at run time

1. Set up a new database

Before you start, create an account on dashboard.fauna.com. Once you have an account, let’s set up a new database. It should hold products and their reviews, so we can load the products at build-time and the reviews in the browser. 

2. Add a GraphQL schema to the database

Next, we use the server key to upload a GraphQL schema to our database. For this, we create a new file called schema.gql that has the following content:

type Product {
  title: String!
  description: String
  reviews: [Review] @relation
}

type Review {
  username: String!
  text: String!
  product: Product!
}

type Query {
  allProducts: [Product]
}

You can upload your schema.gql file via the FaunaDB Console by clicking “GraphQL” on the left sidebar, and then click the “Import Schema” button.

Upon providing FaunaDB with a GraphQL schema, it automatically creates the required collections for the entities in our schema (products and reviews). Besides that, it also creates the indexes that are needed to interact with those collections in a meaningful and efficient manner. You should now be presented with a GraphQL playground where you can test out 

3. Seed the database with mock-up data

To seed our database with products and reviews, we can use the Shell at dashboard.fauna.com: 

To create some data, we’ll use the Fauna Query Language (FQL), after that we’ll continue with GraphQL to build are example application. Paste the following FQL query into the Shell to create three product documents:

Map(
  [
    { title: "Screwdriver", description: "Drives screws." },
    { title: "Hair dryer", description: "Dries your hair." },
    { title: "Rocket", description: "Flies you to the moon and back." }
  ],
  Lambda("product",
    Create(Collection("Product"), { data: Var("product") })
  )
);

We can then write a query that retrieves the products we just made and creates a review document for every product document:

Map(
  Paginate(Match(Index("allProducts"))),
  Lambda("ref", Create(Collection("Review"), {
    data: {
      username: "Tina",
      text: "Good product!",
      product: Var("ref")
    }
  }))
);

Both types of documents will be loaded via GraphQL. However, there is a significant difference between products and reviews. The former will not change a lot and is relatively static, while the second is user-driven. GatsbyJS allows us to load data in two ways:

• data that is loaded at build time which will be used to generate the static site. 
• data that is loaded live at request time as a client visits and interacts with your website. 

In this example, we chose to let the products be loaded at build time, and the reviews to be loaded on-demand in the browser. Therefore, we get static HTML product pages served by a CDN that the user immediately sees. Then, as our user interacts with the product page, we load the data for the reviews. 

4. Create a new GatsbyJS project

The following command creates a GatsbyJS project based on the starter template:

$ npx gatsby-cli new hello-world-gatsby-faunadb
$ cd hello-world-gatsby-faunadb

5. Install npm packages

In order to build our new project with Gatsby and Apollo, we need a few additional packages. We can install the packages with the following command: 

 $ npm i gatsby-source-graphql apollo-boost react-apollo

We will use gatsby-source-graphql as a way to link GraphQL APIs into the build process. Using this library, you can make a GraphQL call from which the results will be automagically provided as the properties for your react component. That way, you can use dynamic data to statically generate your application. The apollo-boost package is an easily configurable GraphQL library that will be used to fetch data on the client. Finally, the link between Apollo and React will be taken care of by the react-apollo library.

6. Create the server key for the database

We will create a Server key which will be used by Gatsby to prerender the page. Remember to copy the secret somewhere since we will use it later on. Protect server keys carefully, they can be used to create, destroy, or manage the database to which they are assigned. To create the key we can go to the fauna dashboard and create the key in the security tab. 

7. Update GatsbyJS config files with server and new read-only keys

To add the GraphQL support to our build process, we need to add the following code into our graphql-config.js inside the plugins section where we will insert the FaunaDB server key which we generated a few moments ago. 

{
  resolve: "gatsby-source-graphql",
  options: {
    typeName: "Fauna",
    fieldName: "fauna",
    url: "https://graphql.fauna.com/graphql",
    headers: {
      Authorization: "Bearer <SERVER KEY>",
    },
  },
}

For the GraphQL access to work in the browser, we have to create a key that only has permissions to read data from the collections. FaunaDB has an extensive security layer in which you can define that. The easiest way is to go to the FaunaDB Console at dashboard.fauna.com and create a new role for your database by clicking “Security” in the left sidebar, then “Manage Roles,” then “New Custom Role”:

Call the new custom role ‘ClientRead’ and make sure to add all collections and indexes (these are the collections that were created by importing the GraphQL schema). Then, select Read for each for them. Your screen should look like this:

You have probably noticed the Membership tab on this page. Although we are not using it in this tutorial, it is interesting enough to explain it since it’s an alternative way to get security tokens. In the Membership tab can specify that entities of a collection (let’s say we have a ‘Users’ collection) in FaunaDb are members of a particular role. That means that if you impersonate one of these entities in that collection, the role privileges apply. You impersonate a database entity (e.g. a User) by associating credentials with the entity and using the Login function, which will return a token. That way you can also implement password-based authentication in FaunaDb. We won’t use it in this tutorial, but if that interests you, check the FaunaDB authentication tutorial.

Let’s ignore Membership for now, once you have created the role, we can create a new key with the new role. As before, click “Security”, then “New Key,” but this time select “ClientRead” from the Role dropdown:

Now, let’s insert this read-only key in the gatsby-browser.js configuration file to be able to call the GraphQL API from the browser:

import React from "react"
import ApolloClient from "apollo-boost"
import { ApolloProvider } from "react-apollo"

const client = new ApolloClient({
  uri: "https://graphql.fauna.com/graphql",
  request: operation => {
    operation.setContext({
      headers: {
        Authorization: "Bearer <CLIENT_KEY>",
      },
    })
  },
})

export const wrapRootElement = ({ element }) => (
  <ApolloProvider client={client}>{element}</ApolloProvider>
)

GatsbyJS will render its Router component as a root element. If we want to use the ApolloClient everywhere in the application on the client, we need to wrap this root element with the ApolloProvider component.

8. Load the pre-rendered product data at build time

Now that everything is set up, we can finally write the actual code to load our data. Let’s start with the products we will load at build time.

For this we need to modify src/pages/index.js file to look like this:

import React from "react"
import { graphql } from "gatsby"
Import Layout from "../components/Layout"

const IndexPage = ({ data }) => (
  <Layout>
    <ul>
      {data.fauna.allProducts.data.map(product => (
        <li>{product.title} - {product.description}</li>
      ))}
    </ul>
  </Layout>
)

export const query = graphql`
{
  fauna {
    allProducts {
      data { _id title description }
    }
  }
}
`

export default IndexPage

The exported query will automatically get picked up by GatsbyJS and executed before rendering the IndexPage component. The result of that query will be passed as data prop into the IndexPage component.If we now run the develop script, we can see the pre-rendered documents on the development server on http://localhost:8000/.

 $ npm run develop

9. Load the reviews at run time

To load the reviews of a product on the client, we have to make some changes to the src/pages/index.js: 

import { gql } from "apollo-boost"
import { useQuery } from "@apollo/react-hooks"
import { graphql } from "gatsby"
import React, { useState } from "react"
import Layout from "../components/layout"

// Query for fetching at build-time
export const query = graphql

`
{ 
  fauna { 
    allProducts { 
      data { 
        _id title description
        }
      } 
    }
  }
  `

  // Query for fetching on the client
  const GET_REVIEWS = gql
  `
  query GetReviews($productId: ID!) {
    findProductByID(id: $productId) {
      reviews { 
        data { 
          _id username text
        }
      }
    }
  }
`
const IndexPage = props => {
  const [productId, setProductId] = useState(null)
  const { loading, data } = useQuery(GET_REVIEWS, {
    variables: {
      productId
    },
    skip: !productId,
  })
}

export default IndexPage

Let’s go through this step by step.

First, we need to import parts of the apollo-boost and apollo-react packages so we can use the GraphQL client we previously set up in the gatsby-browser.js file.

Then, we need to implement our GET_REVIEWS query. It tries to find a product by its ID and then loads the associated reviews of that product. The query takes one variable, which is the productId.

In the component function, we use two hooks: useState and useQuery

The useState hook keeps track of the productId for which we want to load reviews. If a user clicks a button, the state will be set to the productId corresponding to that button.

The useQuery hook then applies this productId to load reviews for that product from FaunaDB. The skip parameter of the hook prevents the execution of the query when the page is rendered for the first time because productId will be null.

If we now run the development server again and click on the buttons, our application should execute the query with different productIds as expected.

$ npm run develop

Conclusion

A combination of server-side data fetching and client-side hydration make JAMstack applications pretty powerful. These methods enable flexible interaction with our data so we can adhere to different business needs. 

It’s usually a good idea to load as much data at build time as possible to improve page performance. But if the data isn’t needed by all clients, or too big to be sent to the client all at once, we can split things up and switch to on-demand loading on the client. This is the case for user-specific data, pagination, or any data that changes rather frequently and might be outdated by the time it reaches the user.

In this article, we implemented an approach that loads part of the data at build time, and then loads the rest of the data in the frontend as the user interacts with the page. 

Of course, we have not implemented a login or forms yet to create new reviews. How would we tackle that? That is material for another tutorial where we can use FaunaDB’s attribute-based access control to specify what a client key can read and write from the frontend. 

The code of this tutorial can be found in this repo.