Sanity has this concept of “portable text”. It’s basically an AST (Abstract Syntax Tree) generated via their rich text editor (called blocks), which you can render in any format (HTML, JSX, Markdown…).

In this article, I want to walk through automating the creation of a table of contents for the headings contained in a portable text tree. The idea goes likethis:

  1. Walk to the AST to find heading nodes.
  2. Construct a small data structure that represents the heading outline.
  3. Use it to render a table of contents.

Let’s start here, with the body prop containing the portable text queried from Sanity:

const BlogPost = props => {
  return <PortableText value={props.body} />
}

I’ll be using React in this article, but the core logic is framework-agnostic and applicable regardless of how you render your components.

Finding headings

The first ting we need is a way to extract heading nodes from that data tree. To do so, we need a way to walk the tree, test every node, and collect the ones that match a function.

This is how we would create such a function:

  • We “reduce” the tree of nodes into an array of relevant nodes with Array.prototype.reduce.
  • We test every node with our matcher: if it matches, we keep it.
  • If the node has children, we recursively them.
const filter = (ast, match) =>
  ast.reduce((acc, node) => {
    if (match(node)) acc.push(node)
    if (node.children) acc.push(...filter(node.children, match))
    return acc
  }, [])

Now, we can create a findHeadings function that look for nodes with a style prop like h2, h3

const findHeadings = ast => filter(ast, node => /h\d/.test(node.style))

Note that style has nothing to do with the style HTML attribute. It’s a property called style on Portable Text nodes which may contain things like normal, h2, h3, etc.

Edit from October 1st, 2022: Simeon Griggs, from the Sanity team, came up with a clever way to retrieve headings directly in groq by leveraging new groq features. It avoids doing it in JavaScript like before, and could be faster for very large trees since groq is typically quite performant.

*[ _type == "article" ] {
  body,
  "headings": body[length(style) == 2 && string::startsWith(style, "h")]
}

Nesting headings

Now, we want a function that nests these headings properly based on their level. This is surprisingly difficult to do, so I decided to rely on the code of outline-audit I wrote in 2016, which essentially does the same thing. Here is a compact version:

const get = (object, path) => path.reduce((prev, curr) => prev[curr], object)
const getObjectPath = path =>
  path.length === 0
    ? path
    : ['subheadings'].concat(path.join('.subheadings.').split('.'))

const parseOutline = ast => {
  const outline = { subheadings: [] }
  const headings = findHeadings(ast)
  const path = []
  let lastLevel = 0

  headings.forEach(heading => {
    const level = Number(heading.style.slice(1))
    heading.subheadings = []

    if (level < lastLevel) for (let i = lastLevel; i >= level; i--) path.pop()
    else if (level === lastLevel) path.pop()

    const prop = get(outline, getObjectPath(path))
    prop.subheadings.push(heading)
    path.push(prop.subheadings.length - 1)
    lastLevel = level
  })

  return outline.subheadings
}

We now have an array of top-level headings, and each of these headings has its own subheadings in its subheadings prop. Pretty neat! Here is an example:

Rendering

We have everything we need to render our table of contents in the frontend!

const BlogPost = props => {
  const outline = parseOutline(props.body)

  return (
    <>
      <TableOfContents outline={outline} />
      <PortableText value={props.body} />
    </>
  )
}

And finally, our TableOfContents component:

const getChildrenText = props =>
  props.children
    .map(node => (typeof node === 'string' ? node : node.text || ''))
    .join('')

const TableOfContents = props => (
  <ol>
    {props.outline.map(heading => (
      <li>
        <a href={'#' + heading._key}>{getChildrenText(heading)}</a>
        {heading.subheadings.length > 0 && (
          <TableOfContents outline={heading.subheadings} />
        )}
      </li>
    ))}
  </ol>
)

A couple of things to note here:

  • We extract the text from the heading node with this getChildrenText function.
  • We recursively render table of contents components for subheadings (if any).
  • The styling is left at your discretion.

Customizing anchors

Right now, we are using the Sanity node key (the _key property) as the ID for our headings. It’s okay, but it doesn’t make for great URLs (e.g. /your-path#b4282a9f0b2e). It can also generate invalid IDs since keys can start with a number, which is not allowed in HTML.

We can tweak our findHeadings function to provide more information for each node. Sanity uses speakingurl to generate slugs under-the-hood, so there are good chances it’s already in your bundle. We can use it to transform the heading text into a slug (e.g. “Customizing anchors” would become “customizing-anchors”).

const findHeadings = ast =>
  filter(ast, node => /h\d/.test(node.style)).map(node => {
    const text = getChildrenText(node)
    const slug = speakingurl(text)

    return { ...node, text, slug }
  })

And we can update our component:

<a href={'#' + heading.slug}>{heading.text}</a>

That’s it folks! I hope it helps you generating table of contents for your portable text. Feel free to reach out on Twitter if you have any question!