Credence.Pattern.PreferErlangFloat (credence v0.7.0)

Replaces float-coercion arithmetic tricks with explicit :erlang.float/1.

LLMs (and developers) use x * 1.0, x / 1.0, x + 0.0, or x - 0.0 to coerce a number to a float. These are arithmetic idioms borrowed from Python; :erlang.float/1 expresses the same intent explicitly and works whether the input is an integer (converts) or already a float (returns it unchanged).

The rewrite preserves the float result — it does not delete the coercion. Deleting * 1.0 (an earlier mistake of a now-merged sibling rule) would turn 6.0 back into 6, a value-kind change; wrapping in :erlang.float/1 keeps the value identical for every number.

Applies to any operand shape — bare variables (n * 1.0) and compound expressions alike (Enum.sum(list) * 1.0, (a + b) * 1.0).

Detected patterns

x * 1.0      1.0 * x
x / 1.0
x + 0.0      0.0 + x
x - 0.0

Note: 0.0 - x is NOT flagged — it negates, not coerces.

Behaviour note

For every numeric operand the rewrite is exact (same float value). A non-number operand raises in both forms — x * 1.0 raises ArithmeticError while :erlang.float(x) raises ArgumentError — so the code crashes either way and only the exception module differs. (A non-number operand at a float-coercion site is already-broken code.)

Bad

defp to_float(n) when is_integer(n), do: n * 1.0
avg = total / count * 1.0

Good

defp to_float(n) when is_integer(n), do: :erlang.float(n)
avg = :erlang.float(total / count)