article Markdown to HTML Converter star

How It Works

This converter follows the CommonMark specification — a precisely defined version of Markdown that resolves the ambiguities left open by the original 2004 Markdown description (how nested lists interact with loose paragraphs, what happens when a marker sits inside a blockquote, and so on). Parsing happens in two distinct passes. The first pass is block-level: the raw text is split into structural units — paragraphs, ATX headings (# through ######), fenced code blocks, blockquotes, ordered and unordered lists, horizontal rules — purely by reading line structure and indentation, without looking at what is inside each block yet. The second pass is inline-level: within each block, the parser scans for emphasis markers (*italic*, **bold**), inline code spans (`code`), and link or image syntax ([text](url)), rewriting them as the matching HTML tags. This two-phase design explains why an asterisk sitting inside a fenced code block is left untouched — code blocks are fully resolved in the first pass, before inline emphasis rules ever reach their contents.

Example: given ## Release Notes followed by a blank line and - **v2.1**: fixes, block-level parsing first identifies a heading and a list with one item; inline-level parsing then turns **v2.1** into <strong>v2.1</strong> inside the <li>. Tables follow the GFM (GitHub Flavored Markdown) extension layered on top of CommonMark: a header row, a --- separator row whose colon placement (:--, :-:, --:) sets column alignment, and any number of data rows below.

What to Know

• Raw HTML is escaped, not executed — pasting a script tag or an inline event handler renders as visible text instead of running, a deliberate safety choice for pasted or untrusted content.
• Loose vs. tight lists change the output — a blank line between list items forces CommonMark to wrap each item in its own <p> (a "loose" list); no blank lines keep items as plain inline content (a "tight" list), which changes vertical spacing in the rendered HTML.
• Very deep nesting has a practical ceiling — four-plus levels of nested lists or reference-style link definitions are valid CommonMark, but are edge cases better served by a dedicated build-time parser than a browser-based live preview.