• Background & Theory
• Installation
• Example Usage
• APIs
• Frontmatter & Custom Processing
• Caveats
• Security
• License

If you are using MDX within a Next.js app, you are probably using the Webpack loader. This means that you have your MDX files locally and are probably using next-mdx-enhanced in order to be able to render your MDX files into layouts and import their front matter to create index pages.

This workflow is fine, but introduces a few limitations that we aim to remove with next-mdx-remote:

• The file content must be local. You cannot store MDX files in another repo, a database, etc. For a large enough operation, there will end up being a split between those authoring content and those working on presentation of the content. Overlapping these two concerns in the same repo makes a more difficult workflow for everyone.
• You are bound to filesystem-based routing. Your pages are generated with urls according to their locations. Or maybe you remap them using exportPathMap, which creates confusion for authors. Regardless, moving pages around in any way breaks things -- either the page's url or your exportPathMap configuration.
• You will end up running into performance issues. Webpack is a JavaScript bundler, forcing it to load hundreds/thousands of pages of text content will blow out your memory requirements. Webpack stores each page as a distinct object with a large amount of metadata. One of our implementations with a couple hundred pages hit more than 8GB of memory required to compile the site. Builds took more than 25 minutes.
• You will be limited in the ways you are able to structure relational data. Organizing content into dynamic, related categories is difficult when your entire data structure is front matter parsed into javascript objects and held in memory.

So, next-mdx-remote changes the entire pattern so that you load your MDX content not through an import, but rather through getStaticProps or getServerProps -- you know, the same way you would load any other data. The library provides the tools to serialize and hydrate the MDX content in a manner that is performant. This removes all of the limitations listed above, and does so at a significantly lower cost -- next-mdx-enhanced is a very heavy library with a lot of custom logic and some annoying limitations. Early testing has shown build times reduced by 50% or more.

# using npm
npm i next-mdx-remote

# using yarn
yarn add next-mdx-remote

import renderToString from 'next-mdx-remote/render-to-string'
import hydrate from 'next-mdx-remote/hydrate'

import Test from '../components/test'

const components = { Test }

export default function TestPage({ source }) {
  const content = hydrate(source, { components })
  return <div className="wrapper">{content}</div>
}

export async function getStaticProps() {
  // MDX text - can be from a local file, database, anywhere
  const source = 'Some **mdx** text, with a component <Test />'
  const mdxSource = await renderToString(source, { components })
  return { props: { source: mdxSource } }
}

While it may seem strange to see these two in the same file, this is one of the cool things about next.js -- getStaticProps and TestPage, while appearing in the same file, run in two different places. Ultimately your browser bundle will not include getStaticProps at all, or any of the functions it uses only on the server, so renderToString will be removed from the browser bundle entirely.

This library exposes two functions, renderToString and hydrate, much like react-dom. These two are purposefully isolated into their own files -- renderToString is intended to be run server-side, so within getStaticProps, which runs on the server/at build time. hydrate on the other hand is intended to be run on the client side, in the browser.

• renderToString(source: string, { components?: object, mdxOptions?: object, scope?: object })

  renderToString consumes a string of MDX along with any components it utilizes in the format { ComponentName: ActualComponent }. It also can optionally be passed options which are passed directly to MDX, and a scope object that can be included in the mdx scope. The function returns an object that is intended to be passed into hydrate directly.

  renderToString(
    // Raw MDX contents as a string
    '# hello, world',
    // Optional parameters
    {
      // The `name` is how you will invoke the component in your MDX
      components: { name: React.ComponentType },
      // MDX's available options at time of writing pulled directly from
      // https://github.com/mdx-js/mdx/blob/master/packages/mdx/index.js
      mdxOptions: {
        remarkPlugins: [],
        rehypePlugins: [],
        hastPlugins: [],
        compilers: [],
        filepath: '/some/file/path',
      },
      scope: {},
    }
  )

  Visit https://github.com/mdx-js/mdx/blob/master/packages/mdx/index.js for available mdxOptions.

• hydrate(source: object, { components?: object })

  hydrate consumes the output of renderToString as well as the same components argument as renderToString. Its result can be rendered directly into your component. This function will initially render static content, and hydrate it when the browser isn't busy with higher priority tasks.

  hydrate(
    // The direct return value of `renderToString`
    source,
    // Should be the exact same components that were passed to `renderToString`
    {
      components: { name: React.ComponentType },
    }
  )

Markdown in general is often paired with frontmatter, and normally this means adding some extra custom processing to the way markdown is handled. Luckily, this can be done entirely independently of next-mdx-remote, along with any extra custom processing necessary.

Let's walk through an example of how we could process frontmatter out of our MDX source:

import renderToString from 'next-mdx-remote/render-to-string'
import hydrate from 'next-mdx-remote/hydrate'

import matter from 'gray-matter'

import Test from '../components/test'

const components = { Test }

export default function TestPage({ source, frontMatter }) {
  const content = hydrate(source, { components })
  return (
    <div className="wrapper">
      <h1>{frontMatter.title}</h1>
      {content}
    </div>
  )
}

export async function getStaticProps() {
  // MDX text - can be from a local file, database, anywhere
  const source = `
---
title: Test
---

Some **mdx** text, with a component <Test name={title}/>
  `

  const { content, data } = matter(source)
  const mdxSource = await renderToString(content, { components, scope: data })
  return { props: { source: mdxSource, frontMatter: data } }
}

Nice and easy - since we get the content as a string originally and have full control, we can run any extra custom processing needed before passing it into renderToString, and easily append extra data to the return value from getStaticProps without issue.

There's only one caveat here, which is that import cannot be used inside an MDX file. If you need to use components in your MDX files, they should be provided through the second argument to the hydrate and renderToString functions.

Hopefully this makes sense, since in order to work, imports must be relative to a file path, and this library allows content to be loaded from anywhere, rather than only loading local content from a set file path.

This library evaluates a string of JavaScript on the client side, which is how it hydrates the MDX content. Evaluating a string into javascript can be a dangerous practice if not done carefully, as it can enable XSS attacks. It's important to make sure that you are only passing the mdxSource input generated by the render-to-string function to hydrate, as instructed in the documentation. Do not pass user input into hydrate.

If you have a CSP on your website that disallows code evaluation via eval or new Function(), you will need to loosen that restriction in order to utilize the hydrate function, which can be done using unsafe-eval. It's also worth noting that you do not have to use hydrate on the client side, but without it, you will get a server-rendered result, meaning no ability to react to user input, etc.

Mozilla Public License Version 2.0