HTML Formatter & Minifier

HTML is the markup language every web page is built from, and unlike XML it is famously forgiving: browsers will render HTML with unclosed tags, unquoted attributes, and elements in technically invalid places, silently patching things up as they go. That leniency is convenient for browsers but means HTML in the wild is often inconsistently formatted — auto-generated markup from a CMS, a template engine, or a minifier is frequently delivered as one giant unbroken line, and hand-written HTML often mixes indentation styles when multiple people or tools have touched it. Because HTML's parsing rules are so permissive and full of special cases (voidelements like <br> and <img> that never have a closing tag, elements like <li> that browsers will auto-close for you, content models that vary per element), there is no built-in browser API equivalent to DOMParser/XMLSerializer that guarantees a clean re-indent while preserving exact semantics — feeding real-world HTML through an XML-style parser will break on things that are perfectly normal in HTML (an unclosed <img> tag, an unencoded standalone & in text, a <p> that browsers auto-close when a new block starts). So this tool takes a different approach: a small hand-rolled tokenizer that walks the markup character by character, recognising opening tags, closing tags, self-closing/void elements, text, comments and doctype declarations, and reconstructs indentation based on nesting depth. Format mode walks the token stream and emits one tag (or text run) per line, increasing indentation by two spaces on each opening tag and decreasing it on each matching closing tag, while treating known void elements (br, img, input, hr, meta, link, and the rest of the HTML5 void element list) as never opening a new indentation level. Minify mode strips HTML comments (<!-- ... -->) entirely, then collapses runs of whitespace between tags down to nothing, producing compact output while leaving the text content of the page (words a user would actually read) intact. A hand-rolled tokenizer like this cannot fully replicate everything a browser's HTML5 parsing algorithm does — the real specification (the WHATWG HTML parsing algorithm) is thousands of lines of rules covering implied tag closing, table-specific parsing quirks, foreign content (SVG/MathML embedded in HTML), and error recovery for almost any malformed input imaginable. This tool handles the markup patterns that make up the overwhelming majority of real HTML — nested elements, attributes with single/double/no quotes, void elements, comments, and text content — but it does not reformat the *contents* of <pre>, <script>, <style> or <textarea> elements, because whitespace inside those is significant (a <pre> block's formatting is meant to be preserved verbatim, and JavaScript/CSS inside <script>/<style> is not HTML and re-indenting it as if it were tags would corrupt it). Those elements' inner content is passed through untouched, with only the outer tags treated as normal HTML tokens. For everyday tasks — cleaning up markup copied from a browser's "view source", tidying a component template, or shrinking a static HTML snippet before pasting it somewhere size-constrained — this lightweight, dependency-free approach is fast, entirely private (nothing leaves your browser), and good enough. For anything going into a production build pipeline, a proper tool such as Prettier (formatting) or html-minifier-terser (minification) will handle far more edge cases correctly.