# JOURNAL-eval — Builder

## Build log

### Approach

AST walker in `calc/evaluator.py`:
- `Num` → return `_coerce(node.value)`
- `Unary('-', ...)` → `_coerce(-evaluate(operand))`
- `BinOp` → evaluate both sides; for `/`, check `right == 0` before dividing; apply `_coerce` to result

`_coerce(value)`: if `isinstance(value, float) and value == int(value)` → `int(value)`, else pass-through.
This keeps the API return clean (no `2.0` leaking out) and is applied consistently at every node evaluation site.

### Test run (local)

```
python -m unittest -v 2>&1
...
Ran 68 tests in 0.270s
OK
```

All 68 tests pass:
- 18 lexer tests (unchanged)
- 26 parser tests (unchanged)
- 24 evaluator + CLI tests (new)

### CLI spot-check

```
python calc.py "2+3*4"    → 14
python calc.py "(2+3)*4"  → 20
python calc.py "7/2"      → 3.5
python calc.py "4/2"      → 2
python calc.py "1/0"      → error: division by zero (stderr, exit 1)
python calc.py "1 +"      → error: unexpected end of input (stderr, exit 1)
```

### D3 rule rationale

Python `/` always returns `float`. Applying `_coerce` at every evaluate site means:
- `4/2` → `2.0` → `int(2)` = `2`
- `7/2` → `3.5` (not whole → stays float)
- `2+3` → `5` (int arithmetic → already int, _coerce is a no-op)

This is documented in `calc/evaluator.py` module docstring.