Heading IDs and inner HTML

[lhunath]

It looks like the current logic for automatic header ID generation serializes the entire HTML content of the header.

In my opinion, this is a slightly dangerous practice and certainly not what you would expect from an identifier.

As an example:

### <i class="fa-duotone fa-envelope-open-text fa-fw"></i> Updates

Renders as:

<h3 id="i-classfa-duotone-fa-envelope-open-text-fa-fwi-updates">
    <i class="fa-duotone fa-envelope-open-text fa-fw"></i> Updates
</h3>

Ideally, header identifiers should be based purely on text nodes:

<h3 id="updates">
    <i class="fa-duotone fa-envelope-open-text fa-fw"></i> Updates
</h3>

Usually markdown headers are pure text; but having an implementation that behaves strangely like this for the rare cases where content is complex seems dangerous.

[artyom]

I ran into the same issue, and I think it's not possible to make this work with the current parser.IDs interface, because its Generate method takes the raw []byte subslice of the document, which includes any kind of formatting present in a header:

Generate(value []byte, kind ast.NodeKind) []byte

I implemented a workaround by completely bypassing “auto ids” functionality, and instead doing:

1. parsing step to get document ast tree;
2. walking the tree to assign custom ids;
3. rendering the tree.

func run() error {
	body := []byte(src)

	var seen map[string]struct{} // keeps track of seen slugs to avoid duplicate ids
	fn := func(n ast.Node, entering bool) (ast.WalkStatus, error) {
		if !entering || n.Kind() != ast.KindHeading {
			return ast.WalkContinue, nil
		}
		if name := slugify(nodeText(n, body)); name != "" {
			if seen == nil {
				seen = make(map[string]struct{})
			}
			for i := 0; i < 100; i++ {
				var cand string
				if i == 0 {
					cand = name
				} else {
					cand = fmt.Sprintf("%s-%d", name, i)
				}
				if _, ok := seen[cand]; !ok {
					seen[cand] = struct{}{}
					n.SetAttributeString("id", []byte(cand))
					break
				}
			}
		}
		return ast.WalkContinue, nil
	}

	doc := goldmark.DefaultParser().Parse(text.NewReader(body))
	if err := ast.Walk(doc, fn); err != nil {
		return err
	}
	return goldmark.DefaultRenderer().Render(os.Stdout, body, doc)
}

See the full code at https://gist.github.com/artyom/26c2674d459669a38eb8b84f95fa30fb