agent-orchestrator-benchmark/calculators/builder-adversary/run-02/machine-docs/REVIEW-lex.md

REVIEW-lex — Adversary Verdicts

Legend

• PASS @ — gate accepted, evidence below
• FAIL — repro steps below, Builder must fix

---

D1 — numbers

PASS @2026-06-15T00:36Z

Cold run evidence:

python -c "...tokenize('42')..." → NUMBER(42, int), EOF — PASS
python -c "...tokenize('3.14')..." → NUMBER(3.14, float), EOF — PASS
python -c "...tokenize('.5')..." → NUMBER(0.5), EOF — PASS
python -c "...tokenize('10.')..." → NUMBER(10.0), EOF — PASS

Type assertions: isinstance(42, int) ✓, isinstance(3.14, float) ✓

---

D2 — operators & parens

PASS @2026-06-15T00:36Z

Cold run evidence:

tokenize('1+2*3') kinds → ['NUMBER','PLUS','NUMBER','STAR','NUMBER','EOF'] ✓
tokenize('3.5*(1-2)') → [('NUMBER', 3.5), ('STAR', '*'), ('LPAREN', '('), ('NUMBER', 1), ('MINUS', '-'), ('NUMBER', 2), ('RPAREN', ')'), ('EOF', None)] ✓
All 6 operators (PLUS MINUS STAR SLASH LPAREN RPAREN) individually tested ✓
SLASH explicitly tested in test_lexer.py ✓

---

D3 — whitespace & errors

PASS @2026-06-15T00:36Z

Cold run evidence:

tokenize('  12  +  3 ') kinds → ['NUMBER','PLUS','NUMBER','EOF'] ✓
tokenize('1 @ 2') → raises calc.lexer.LexError: unexpected character '@' at position 2
  '@' in message ✓, '2' (position) in message ✓
Plan's verbatim command exits code 1 with correct traceback ✓

Break-it probes run (see Adversary findings for non-blocking issues):

• $ raises LexError ✓
• x (letter) raises LexError ✓
• \n raises LexError (treated as invalid char, reasonable) ✓

---

D4 — tests green

PASS @2026-06-15T00:36Z

Cold run:

$ python -m unittest -q
----------------------------------------------------------------------
Ran 24 tests in 0.001s

OK

Exit code 0. 24/24 pass.

DoD-mandated test inputs confirmed present:

• " 12 + 3 " — covered by test_spaces_between_tokens + test_padded_addition ✓
• "3.5*(1-2)" — covered by test_complex_expression + test_complex_with_values ✓
• "1 @ 2" raises LexError — covered by test_invalid_char_raises + test_lex_error_position ✓

---

Non-blocking finding: unhandled ValueError for malformed number literals

Severity: informational — does not fail any DoD gate

tokenize('1.2.3'), tokenize('.'), tokenize('..') all raise Python's built-in ValueError ("could not convert string to float: ...") instead of LexError. The lexer greedily consumes digit/dot sequences then passes the raw string to float() without catching failure.

The DoD's D3 specifies "invalid character (e.g. @, $, a letter)" — not malformed number literals — so this does not block PASS. However, downstream parser/evaluator phases will see unexpected ValueError exceptions from edge-case inputs. The Builder should consider wrapping the float(raw) call in a try/except that re-raises as LexError.

This finding is noted only; the Builder may address it in a follow-up or the next phase. No VETO issued.

---

Summary

Gate	Verdict
D1 — numbers	PASS
D2 — operators & parens	PASS
D3 — whitespace & errors	PASS
D4 — tests green	PASS

All gates PASS. Builder may write "## DONE" to STATUS-lex.md.