Function parse 

pub fn parse(input: &str) -> ParseResult

The driver loop (§13.2 “Parsing HTML documents”, the “tokenization and tree construction” step): parses input into a Document tree with per-node source positions, feeding it through the tokenizer and handing each token to the tree builder, applying the two pieces of feedback tree construction sends back to the tokenizer — a state switch (Tokenizer::switch_to, for RCDATA/RAWTEXT/script-data/PLAINTEXT elements) and the foreign-content flag (Tokenizer::set_in_foreign_content, consulted only by CDATA-section handling).

The tokenizer’s iterator yields exactly one Eof token and then ends (None) on the next call, so the loop needs no separate condition for when to stop feeding it tokens. TreeBuilder::stop_parsing (§13.2.7 “The end”) still runs once, explicitly, right after — its one tree-shape-relevant step (“pop all the nodes off the stack of open elements”) isn’t implied by the loop simply ending.

Returns ParseResult, not a bare Document, as of Phase 07 (plan/07-parse-errors.md) — errors covers every tokenizer-level parse error (src/tokenizer.rs’s error() call sites) plus, as of Phase 08 (plan/08-tree-construction-errors.md), the tree-construction-level (§13.2.6) conditions listed there. Both sources are merged and sorted by source position, so errors is always in document order regardless of which stage produced each entry (the two stages interleave: the tokenizer runs ahead of the tree builder token by token).