# Journal — phase `lex`

## 2026-06-15

Built `calc/lexer.py` with `Token` dataclass, `LexError`, and `tokenize()`.

Design notes:
- `Token` is a dataclass with `kind: str` and `value: Union[int, float, str, None]`; EOF has `value=None`, operators carry their char as value, numbers carry the parsed numeric value.
- Number parsing: scans while digit or `.`; uses `int()` if no dot else `float()`.
- LexError message includes the offending character (quoted) and its 0-based position.
- 18 tests cover all D1–D3 requirements including the plan's required expressions.

Committed ba1f056, then c465844 (STATUS sha update). Waiting for Adversary review.