Concepts

KindaXML is designed to annotate text without building a deep DOM. The parser emits a flat list of text segments and optional markers, recovering from common model mistakes.

Core ideas

• Annotation-first: tags annotate text spans rather than create nested trees.
• Deterministic recovery: malformed tags still map to predictable spans.
• Shallow structure: open tags auto-close when the next tag starts to avoid runaway nesting.

Output model

The parser returns:

• text: tag-stripped text
• segments: ordered runs of text, each with zero or more annotations
• markers: zero-width annotations for self-closing tags

Example (conceptual):

[
  {"text": "We shipped ", "annotations": []},
  {"text": "last week", "annotations": [{"tag": "cite", "attrs": {"id": "1"}}]},
  {"text": ".", "annotations": []}
]

Runnable example

Run:

cargo run -q --example basic

Expected output (excerpt):

=== Inline span ===
Original text:
We shipped <cite id="1">last week</cite>.

Parsed text:
We shipped last week.
Segments:
- 'We shipped '
- 'last week' (cite [id="1"])
- '.'

CI tip: compare the excerpt above with the matching section in the command output to verify the annotation model.