demo(11-float-money-equality): 11 — Floating-point equality on a money amount

[EricCogen]

11 — Floating-point equality on a money amount

Expected verdict: ❌ Fails — GauntletCI should fire GCI0049 (float/double equality comparison).

What changed

A new LegacyDiscountCalculator works in double (because it was
ported from a JavaScript pricing prototype) and tests for an exact
zero-net result with ==:

public sealed class LegacyDiscountCalculator
{
    public double NetTotal(double subtotal, double discount)
    {
        var net = subtotal - discount;
        if (net == 0.0)
        {
            return 0.0;
        }
        return net;
    }
}

For most inputs subtotal - discount is exactly representable, but for
values like 0.1 + 0.2 - 0.3 the result is 5.55e-17, not 0.0, and
the comparison silently returns false.

Why this is risky

• Money values must never be compared with == in double/float.
  Binary floating-point cannot exactly represent most decimal fractions,
  so a "zero net" check returns the wrong answer for a single rounding
  ulp difference.
• The bug is path-dependent: it fires only for some discount/subtotal
  pairs, so it sails through unit tests that use round numbers and
  surfaces in production for the one customer with $19.99 and a
  $19.99 discount.
• The fix is either an epsilon comparison (Math.Abs(net) < 1e-9) or
  — much better for money — using decimal end-to-end.

What GauntletCI catches

GCI0049 Float/Double Equality Comparison — == (or !=) on the
same line as a floating-point literal in a non-test file.

[github-actions]

GauntletCI found issues in this PR. See inline comments for details.

[github-actions]

GCI0049 — Float/Double Equality Comparison

Direct equality comparison on a floating-point value — use an epsilon threshold instead

Evidence:

Line 8: if (net == 0.0)

⚠️ Why it matters: Floating-point arithmetic is inexact due to binary representation. Two values that are mathematically equal may differ by a tiny rounding error, causing == to return false and != to return true unexpectedly.

💡 Suggested action: Compare with an epsilon: Math.Abs(a - b) < 1e-9. For financial calculations use decimal instead of float/double.

_{Confidence: Medium | Severity: Info}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7265b3e33b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Avoid exact equality on floating-point totals

Replace this == check with a tolerance-based comparison (or switch to decimal) because subtotal - discount in double can produce tiny residuals for mathematically equal amounts (for example, values derived from 0.1 + 0.2 versus 0.3). In those cases this branch is skipped and a near-zero nonzero result leaks out, which is incorrect for zero-net pricing logic.

Useful? React with 👍 / 👎.