Warning: Under Construction. I am using this site as my playground as I buildĀ gatsby-mdx and gatsby themes. During this time I will also repost new contentĀ to Medium

How MDX transforms into JSX

MDX is at the core of gatsby-mdx and unified is at the core of MDX... but if you're not immersed in the tech it can be hard to understand what MDX is, how it transforms markdown into JSX, and thus also why it's powerful.

Vocabulary

If we import @mdx-js/mdx, the core package for MDX, we can transform arbitrary markdown strings into JSX. MDX does this through a multi-step, multi-AST process. During this process we'll touch mdast, hast, and jsx. mdast is usually referred to as remark while hast is usually referred to as rehype.

Now that we've gotten our vocab out of the way let's take a look at some markdown content

Markdown

This markdown is fairly standard. It has a header, some paragraph content and a code block.

# a title
some paragraph content
```js some-prop
const some = 'code'
```

If we run this markdown through mdx using mdx.sync (mdx is async by default but we're toning it down a bit for this blog post) we end up with the following JSX output.

export default class MDXContent extends React.Component {
constructor(props) {
super(props);
this.layout = null;
}
render() {
const { components, ...props } = this.props;
return (
<MDXTag name="wrapper" components={components}>
<MDXTag name="h1" components={components}>{`a title`}</MDXTag>
<MDXTag
name="p"
components={components}
>{`some paragraph content`}</MDXTag>
<MDXTag name="pre" components={components}>
<MDXTag
name="code"
components={components}
parentName="pre"
props={{
className: "language-js",
metastring: "some-prop",
"some-prop": true
}}
>{`const some = 'code'
`}</MDXTag>
</MDXTag>
</MDXTag>
);
}
}

MDX outputs a React component authored as a class. The layout in the constructor is an MDX layout, which we'll go over another time. For now, know that it's in the constructor because we need a stable reference to the layout component if there is one.

MDXTag is a special component that handles rendering any dom-level components. These are the ones that you typically write in lowercase when using React regularly. MDXTag is what lets us replace the rendering of any of these elements with out own React components. There are also a couple names with special handling that we can replace, wrapper and inlineCode.

Each MDXTag takes the components prop as an argument. This is the same components prop you might have used to define your own React components to handle rendering a pre, a, or p tag.

The props prop are the props that get passed to the underlying element when rendered. In this example, the code element gets these props. Notice how one of the props was parsed from the meta string of the code block. this is a standard markdown feature.

{
className: "language-js",
metastring: "some-prop",
"some-prop": true
}

The other ASTs

That doesn't tell us much about how we got here though. So let's create two plugins: one remark and one hast. These plugins do nothing but print out the AST they receive and return it. This allows us to see the mdast and hast output before it gets transformed into JSX.

const mdPlugin = () => ast => console.log(ast) || ast;
const hastPlugin = () => ast => console.log(ast) || ast;
console.log(
mdx.sync(mdxContent, { mdPlugins: [mdPlugin], hastPlugins: [hastPlugin] })
);

mdast

First the mdast. Since mdast is a markdown ast, we'll see very high level objects here such as heading, text, and paragraph. There are no HTML tags yet. There's also positional data that tells us where each node starts and ends.

Nodes like heading can have additional metadata on them like depth, which indicates what level of heading it is.

the mdast then gets transformed into HAST using a set of custom handlers for code blocks, ES2015 imports, and some other nodes.

{
type: "root",
children: [
{
type: "heading",
depth: 1,
children: [
{
type: "text",
value: "a title",
position: {
start: {
line: 1,
column: 3,
offset: 2
},
end: {
line: 1,
column: 10,
offset: 9
},
indent: []
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 1,
column: 10,
offset: 9
},
indent: []
}
},
{
type: "paragraph",
children: [
{
type: "text",
value: "some paragraph content",
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 23,
offset: 33
},
indent: []
}
}
],
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 23,
offset: 33
},
indent: []
}
},
{
type: "code",
lang: "js",
meta: "some-prop",
value: "const some = 'code'",
position: {
start: {
line: 5,
column: 1,
offset: 35
},
end: {
line: 7,
column: 4,
offset: 74
},
indent: [1, 1]
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 8,
column: 1,
offset: 75
}
}
}

HAST

Next the HAST. Since HAST is an HTML AST, we'll see how the original heading in mdast translated to an h1 tag in HAST. Notice how the depth property was used to translate the heading into an h1 specifically. The HAST also gets transformed in a similar way to the mdast earlier using a toJSX function.

{
type: "root",
children: [
{
type: "element",
tagName: "h1",
properties: {},
children: [
{
type: "text",
value: "a title",
position: {
start: {
line: 1,
column: 3,
offset: 2
},
end: {
line: 1,
column: 10,
offset: 9
}
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 1,
column: 10,
offset: 9
}
}
},
{
type: "text",
value: "
"
},
{
type: "element",
tagName: "p",
properties: {},
children: [
{
type: "text",
value: "some paragraph content",
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 23,
offset: 33
}
}
}
],
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 23,
offset: 33
}
}
},
{
type: "text",
value: "
"
},
{
type: "element",
tagName: "pre",
properties: {},
children: [
{
type: "element",
tagName: "code",
properties: {
className: ["language-js"],
metastring: "some-prop",
"some-prop": true
},
children: [
{
type: "text",
value: "const some = 'code'
"
}
],
position: {
start: {
line: 5,
column: 1,
offset: 35
},
end: {
line: 7,
column: 4,
offset: 74
}
}
}
],
position: {
start: {
line: 5,
column: 1,
offset: 35
},
end: {
line: 7,
column: 4,
offset: 74
}
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 8,
column: 1,
offset: 75
}
}
}

What about React Components?

It's true that we can import and use React components directly in MDX too, so what does that do to the JSX output? Let's take a look at the following MDX as it goes through the process.

# a title
<SketchPicker />

Notice how the mdast keeps the React component as an HTML node, untouched.

{
type: "root",
children: [
{
type: "heading",
depth: 1,
children: [
{
type: "text",
value: "a title",
position: {
start: {
line: 1,
column: 3,
offset: 2
},
end: {
line: 1,
column: 10,
offset: 9
},
indent: []
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 1,
column: 10,
offset: 9
},
indent: []
}
},
{
type: "html",
value: "<SketchPicker />",
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 17,
offset: 27
},
indent: []
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 4,
column: 1,
offset: 28
}
}
};

Check out how the html block was transformed into a jsx block.

{
type: "root",
children: [
{
type: "element",
tagName: "h1",
properties: {},
children: [
{
type: "text",
value: "a title",
position: {
start: {
line: 1,
column: 3,
offset: 2
},
end: {
line: 1,
column: 10,
offset: 9
}
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 1,
column: 10,
offset: 9
}
}
},
{
type: "text",
value: "
"
},
{
type: "jsx",
value: "<SketchPicker />",
position: {
start: {
line: 3,
column: 1,
offset: 11
},
end: {
line: 3,
column: 17,
offset: 27
},
indent: []
}
}
],
position: {
start: {
line: 1,
column: 1,
offset: 0
},
end: {
line: 4,
column: 1,
offset: 28
}
}
};

Notice how the React component was passed through unaltered.

export default class MDXContent extends React.Component {
constructor(props) {
super(props);
this.layout = null;
}
render() {
const { components, ...props } = this.props;
return (
<MDXTag name="wrapper" components={components}>
<MDXTag name="h1" components={components}>{`a title`}</MDXTag>
<SketchPicker />
</MDXTag>
);
}
}

Fin

Hopefully you have a bit more of an understanding of how MDX content gets transformed into JSX and the steps involved. In a future post we'll cover how the MDXProvider and MDXTag components work together to enable replacing dom element rendering with custom React components.

In the meantime, since you made it this far, have a codesandbox link you can play around with and check the logs for how different content affects the mdast, hast, and JSX. Maybe even try writing your own remark plugins!