nx_js_lex.nx

nx_js_lex.nx -- R-JS-LEX (WB-JS-001 rung 0): the ECMAScript TOKENIZER, the hardware-up floor of the sovereign JS engine (browser-engine-exceed roadmap §2). Founds R-JS-PARSE above it (no-floating law: the lexer is built+gated before any parser rides it). Shape = STATE_MACHINE/STRUCT_WALK over source bytes -- an EXISTING emitter shape, so team-authorable now (NOT blocked on X-AUT-006c). Scans ECMAScript source into a flat token stream (3 i64 slots per token: kind, start-offset, byte-length). Covers the lexical core real JS uses: - whitespace + line-terminators (skipped; ASI is a PARSER concern, rung 1) - // line comments and /* block */ comments (skipped) - identifiers + keywords ($ and _ legal; keyword set classified) - numeric literals: decimal int/float, exponent, 0x/0o/0b prefixes - string literals ' and " with backslash-escape handling (\" \' \\ \n ...) - punctuators incl. multi-char operators (=== !== >>>= => ?. ?? ** etc.) - unterminated string -> ERROR token (honest, never silently swallowed) HONEST RUNG-0 SCOPE (named OPEN, NOT faked): template literals (`...${}`) and regex-literal disambiguation need PARSER context (is `/` a divide or a regex?), so they are R-JS-LEX-0b, built once the parser can feed back context. BigInt `n` suffix + full Unicode IdentifierStart are also 0b. Rung 0 tokenizes ASCII JS. GATE (main): 7 KATs assert the exact token stream (kind + lexeme) on real JS snippets + TAMPER (unterminated string must be ERROR, never a fabricated STRING). Self-validating; exit 0 iff all pass; appends knowledge/status/js_engine.log. license_tier: ORIGINAL (tutor-bootstrap scaffold; team re-authors from the R-JS-LEX data spec via author=organ -- (B)-debt, mirror the h2 capstone note.)