Shaleen Jain

· 3 min read

Prefetching Data with Next.js and Apollo

Update: My pull requests implementing the techniques discussed here were merged into the zeit/next.js with-data-prefetch and with-apollo examples. You can easily combine the various examples (including the above two) to achieve your desired functionality.

For most modern web apps, network speed and latency is still the biggest constraint for high performance and great UX. Universally rendered Isomorphic applications do a great job to solve that to some extent without sacrificing interactivity by processing the complete DOM of the first request on the server and subsequent requests on the client. But there is a lot more we can do.

Modern web apps already have their core application logic on the client by the time the first request completes. From there on anticipating the user action and selectively prefetching the data can lead to great performance and UX improvements.

Apollo and Next is the stack of our choice, having the necessary capabilities, to demonstrate our concept.

Adam Soffer’s introduction summarizes Apollo and Next with their concerns perfectly:

Apollo is a GraphQL client that’s concerned exclusively with the data layer of our application; it cares about efficiently fetching our data. Next, on the other hand, is a minimalistic framework for server-rendered React applications that’s concerned exclusively with the UI layer of our application; it cares about efficiently rendering our UI.

🔗 Tying together Next.js and Apollo

Next has a unique way of loading data, every page has a lifecycle hook function getInitialProps which is called while loading on the client as well as the server. This is a very nice separation of our data logic from the UI logic and is what enables us to use the full power of Apollo and Next as both of these frameworks make no assumptions about each others layer and provides the perfect abstraction to connects these two layers.

Apollo client’s react library react-apollo provides a helper function getDataFromTree which recursively traverses the supplied React component and executes every graphql() query it finds, storing the result in its local cache (configured by us).

The excellent example from the Next.js repo shows how we can combine these two functionalities by creating a HOC which wraps getDataFromTree within getInitialProps and returning the cached data as a plain Javascript Object

Here we make sure to use getDataFromTree on both the client and for SSR, altering the example like so:

import React from 'react'
import PropTypes from 'prop-types'
import { ApolloProvider, getDataFromTree } from 'react-apollo'
import Head from 'next/head'

import initApollo from './initApollo'

export default ComposedComponent => {
  return class WithData extends React.Component {
    static displayName = `WithData(${getComponentDisplayName(
    static propTypes = {
      serverState: PropTypes.object.isRequired

    static async getInitialProps(ctx) {
      // Initial serverState with apollo (empty)
      let serverState

      // Evaluate the composed component's getInitialProps()
      let composedInitialProps = {}
      if (ComposedComponent.getInitialProps) {
        composedInitialProps = await ComposedComponent.getInitialProps(ctx)

      // Run all GraphQL queries in the component tree
      // and extract the resulting data
      const apollo = initApollo();

      try {
        // Run all GraphQL queries
        await getDataFromTree(
          <ComposedComponent ctx={ctx} {...composedInitialProps} />,
            router: {
              asPath: ctx.asPath,
              pathname: ctx.pathname,
              query: ctx.query
            client: apollo,
      } catch (error) {
        // Prevent Apollo Client GraphQL errors from crashing SSR.
        // Handle them in components via the data.error prop:

      if (!process.browser) {
        // getDataFromTree does not call componentWillUnmount
        // head side effect therefore need to be cleared manually

      // Extract query data from the Apollo store
      serverState = {
        apollo: {
          data: apollo.cache.extract()

      return {

    constructor(props) {
      this.apollo = initApollo(

    render() {
      return (
        <ApolloProvider client={this.apollo}>
          <ComposedComponent {...this.props} />

🔗 Prefetching Data

Next has in-built support for prefetching pages by adding the prefetch prop to the <Link> component or imperatively calling Router.prefetch('/dynamic')}. However this fetches the page’s JS code but not its data/initial props.

This results in instant page loads but, some would argue even worse, a loading indicator shown for however long it takes to get the relevant page data from your API.

Since we know our stack we can optimise performance using that knowledge. We know that with every call to getInitialProps of a component wrapped in our WithData HOC, our queries are cached in the local ApolloClient store. With that all we have to do is get a reference to our top level page component and manually call its getInitialProps method!

import Router from 'next/router'
import { format, resolve, parse } from 'url'

export default prefetch = async (href) => {
  // if  we're running server side do nothing
  if (typeof window === 'undefined') return

  const url =
    typeof href !== 'string'
      ? format(href)
      : href

  const { pathname } = window.location

  const parsedHref = resolve(pathname, url)

  const { query } =
    typeof href !== 'string'
      ? href
      : parse(url, true);

  // get component reference
  const Component = await Router.prefetch(parsedHref)

  // fetch the component props
  // and cache locally, handled within getInitialProps
  if (Component && Component.getInitialProps) {
    const ctx = { pathname: href, query, isVirtualCall: true }
    await Component.getInitialProps(ctx)

We can then call this throughout our application, like for example: <a onMouseOver={() => prefetch("/product?sku=0001")}>

This will load not only our product.js page but also all the specific graphql queries into the ApolloClient store.

Liked this article? Consider donating to support my work.

Got any questions or comments? Drop me a message on Twitter @shalzzj

Sign up for my newsletter to be the first to know about a new post