DEV Community

Cover image for Create an MDX Plugin To Have My Own Markdown Language
Minh-Phuc Tran
Minh-Phuc Tran

Posted on • Updated on • Originally published at phuctm97.com

Create an MDX Plugin To Have My Own Markdown Language

Yesterday, I migrated my website from plain HTML to Next.js + MDX, to solve the problem of duplications and boilerplates when writing in HTML. However, using Next.js + MDX isn't just about that, it opened a door for me to customize my writing framework with technically no limit (which is why I migrated from Medium/DEV.to/Hashnode to my own website in the first place).

How?

TL;DR: Intercept Next.js and MDX pipeline, parse our custom syntax and write into supported syntax.

Next.js and MDX are designed and created with customization and flexibility in mind.

  • Next.js creates a pipeline to build React server-rendered pages. As long as we are able to convert something into JSX (and JavaScript functions), we can technically use anything (MDX is an example). Next.js is also built on top of Webpack and Babel, which enable you access to the even bigger plugin ecosystems.

  • MDX creates a pipeline to convert Markdown-based syntax into JSX. It is designed and built to work with existing unifiedjs, remark, and rehype ecosystems, which are about compiling content (natural language, Markdown, etc) into structured data. The structured data then can be processed, modified, and written into any existing languages (JSX, MDX, etc).

The combined pipeline looks like this:

  • You define and write custom Markdown documents.

  • Next.js reads the documents as pages, send to MDX.

  • (You intercept and customize here)

  • MDX sends the documents into Remark and Rehype.

  • Remark converts the documents into a data structure called MDXAST.

  • (You intercept and customize here)

  • Rehype converts MDXAST into its data structure called MDXHAST.

  • (You intercept and customize here)

  • Rehype writes final structured data into JSX pages.

  • Next.js statically generates HTML pages.

Some examples of what you can do:

  • Get when a file was first commited into Git and use that as the published date.

  • Based on a file's location and name, determine its layout components.

  • Write your own Github-flavored Markdown syntax that have posts rendered beautifully on both Github and your website.

  • Write a generator that converts your Markdown into formats that are suitable for distributions to different platforms like DEV.to, Hashnode, Medium.

What I did?

Previously, every MDX pages in blog/ directory has to import and export BlogPost component with manually-written JSX props, which have following shortcomings:

  • Being in blog directory should be enough to indicate which layout the MDX pages should use. The import and export are boilerplates.

  • I had to write a path prop to every page so that the canonical and Open Graph URL can be rendered correctly. However, the file location should be sufficient instead of having to write a manually-written prop.

  • The import and export statements are rendered very ugly on Github because Github don't support MDX.

To solve the above problems, I designed the following concept:

  • path, slug, and layout will be infered from the file location. There's no import and duplicated props.

  • Intercept the pipeline after Remark processed Markdown syntax and dynamically add a line importing a coressponding layout component and an export default statement with proper props pre-populated.

  • All other information like SEO description and published time are written in YAML frontmatter so that Github can render properly.

How an article looks in MDX

See full source code:

---
title: "Switch to Next.js and MDX"
description: ">-"
  I switched from plain HTML to using Next.js and MDX to have better ease of
  writing and extensibility.
published time: 2020-12-18
---

## The Problem

To prevent myself from procrastinating, I [started my blog dead simple in plain
HTML][start blog].
Enter fullscreen mode Exit fullscreen mode

How the custom plugin was written (conceptually)

See full source code:

const path = require("path");
const yaml = require("yaml");
const find = require("unist-util-find");

const Components = {
  blog: "BlogPost",
};

const getSubpage = (file) => path.basename(file.dirname);

const getRoute = (file) => {
  const sub = getSubpage(file);

  const Component = Components[sub];
  if (!Component)
    return file.fail(
      `Subpage '${sub}' is invalid. Valid subpages: ${Object.keys(Components)
        .map((it) => `'${it}'`)
        .join(", ")}.`
    );

  const slug = file.stem;
  return {
    Component,
    slug,
    path: `${sub}/${slug}`,
  };
};

module.exports = () => (tree, file) => {
  const frontmatter = find(tree, { type: "yaml" });
  const { title, description, "published time": publishedTime } = yaml.parse(
    frontmatter.value
  );

  const { path, Component } = getRoute(file);
  const props = `{
    path: "${path}",
    title: "${title}",
    description: "${description}",
    publishedTime: new Date("${publishedTime}"),
  }`;

  tree.children.unshift(
    {
      type: "import",
      value: `import ${Component} from "~components/mdx/${Component}";`,
    },
    {
      type: "export",
      default: true,
      value: `export default ${Component}(${props});`,
    }
  );
};
Enter fullscreen mode Exit fullscreen mode

Top comments (0)