datamark vs Raw Frontmatter

The simplest approach to structured Markdown is extracting frontmatter with regex and parsing the YAML yourself.

The manual approach

const match = content.match(/^---\n([\s\S]*?)\n---\n/);
const frontmatter = match ? yaml.parse(match[1]) : {};

This works for simple cases, but it quickly breaks down:

• No body parsing — you still have unstructured text
• No typing — frontmatter is any
• No validation — typos in YAML are silent
• No positions — you can't point to where an error occurred

The datamark approach

import { parse } from "datamark";

const doc = parse(content);
// doc.frontmatter is parsed YAML
// doc.root is a structured section tree

For typed, validated data, use the Format SDK:

import { datamark, inlineText } from "datamark";
import * as z from "zod";

const MyFormat = datamark({
  schema: z.object({ title: z.string() }),
  parse(doc) {
    const h1 = doc.root.children.find(n => n.type === "section") as any;
    const title = h1 ? inlineText(h1.heading.children) : "";
    return { title };
  },
});

When to use raw frontmatter

If you literally only need one key from the frontmatter and nothing else, raw extraction is fine. For anything more complex, datamark gives you structure, types, and error handling for free.