A small, super-fast, event-driven, fault-tolerant, html tag-soup tokenizer that works in node but which is mainly intended as a lightweight client-side parser for small HTML snippets. You pass it a string which is supposed to contain HTML, and it emits events telling you what things it finds. A tokenizer emits events such as "here's an attribute" or "here's an opening tag," however these won't necessarily reflect the structure of well-formed HTML document. For example <foo <foo will produce two opening-tag events in a row. If that bothers you, then what you probably want is a parser, not a tokenizer, so keep scrolling down.