Multi-line opening tags break sibling tag detection

Author: jlevyCreated Jan 15, 2026Updated Jan 15, 2026

What happened?

Multi-line opening tags break sibling tag detection. When an opening tag spans multiple lines, subsequent sibling tags get incorrectly nested as children instead of remaining at the same level.

Expected: Both documents below should produce identical ASTs since they are semantically equivalent per the Markdoc tag specification, which explicitly allows newlines within tags.

Actual:

Single-line: Both fields at depth 3 (siblings) ✓
Multi-line: Field "a" at depth 1, field "b" at depth 4 (nested inside "a") ✗

To reproduce

Two semantically identical documents—the only difference is line wrapping in the opening tag:

Single-line opening tag (parses correctly):

markdown

{% field id="a" attr="x" %}{% /field %}
{% field id="b" %}{% /field %}

Multi-line opening tag (parses incorrectly):

markdown

{% field id="a"
attr="x" %}{% /field %}
{% field id="b" %}{% /field %}

Test script:

javascript

import Markdoc from '@markdoc/markdoc';

const singleLine = `{% field id="a" attr="x" %}{% /field %}
{% field id="b" %}{% /field %}`;

const multiLine = `{% field id="a"
attr="x" %}{% /field %}
{% field id="b" %}{% /field %}`;

function getFieldDepths(doc) {
  const ast = Markdoc.parse(doc);
  let fields = [];
  function walk(node, depth = 0) {
    if (node.type === 'tag' && node.tag === 'field') {
      fields.push({ id: node.attributes?.id, depth });
    }
    for (const child of node.children || []) walk(child, depth + 1);
  }
  walk(ast);
  return fields;
}

console.log('Single-line:', getFieldDepths(singleLine));
// → [ { id: 'a', depth: 3 }, { id: 'b', depth: 3 } ]  ✓ Same depth

console.log('Multi-line:', getFieldDepths(multiLine));
// → [ { id: 'a', depth: 1 }, { id: 'b', depth: 4 } ]  ✗ Different depths!

Version

0.5.4

Additional context

Root Cause

In src/tokenizer/plugins/annotations.ts, the block tag rule:

Detects that the opening tag spans lines 0-1:
```
{% field id="a"
attr="x" %}{% /field %}
```
Advances state.line past those lines
Content after %} on line 1 (the {% /field %} closing tag) is never tokenized

The closing tag is lost, so the block-level tag_open has no matching tag_close, and all subsequent content becomes its children.

Token-level Evidence

javascript

const Tokenizer = Markdoc.Tokenizer;
const tokenizer = new Tokenizer();

// Single-line: both tags inline, properly paired
tokenizer.tokenize(singleLine);
// → paragraph containing inline: tag_open, tag_close, tag_open, tag_close

// Multi-line: closing tag for "a" is missing!
tokenizer.tokenize(multiLine);
// → tag_open (field a, block-level)
// → paragraph containing inline: tag_open (field b), tag_close (field b)
// → NO tag_close for field a!

Workaround

Keep opening tags on a single line. Do not wrap attributes across multiple lines within {% ... %} delimiters.

Source: markdoc/markdoc

View original on GitHub View discussion on GitHub