[Bug]: toDecimal fails to print a leading negative sign when the first unit is negative zero

[jparise]

Is there an existing issue for this?

• I have searched the existing issues

Current behavior

#690 fixed the the general case of formatting negative unit values, but there is one outstanding specific case where formatting is still incorrect: when the first unit is -0, the resulting string doesn't include the leading negative sign.

This occurs because the (default) toString formatter is String, and String(-0) produces '0'.

toDecimal(dinero({ amount: -1, currency: USD }) === '0.01'

Expected behavior

toDecimal(dinero({ amount: -1, currency: USD }) === '-0.01'

Steps to reproduce

Here's a failing test case:

it('returns the negative amount for the last unit in decimal format', () => {
  const d = dinero({ amount: -1, currency: USD });

  expect(toDecimal(d)).toEqual('-0.01');
});

Expected: "-0.01"
Received: "0.01"

Version

2.0.0-alpha12

Environment

Node.js 16.14.1

Code of Conduct

• I agree to follow this project's Code of Conduct

[jparise]

This wasn't an issue prior to the toUnit / toFormat to toUnits / toDecimal change because the code previous deal with values like -0.01 rather than the decomposed units array [-0, .01].

---

I spent some time looking at this. Assuming we want to localize the solution to toDecimal and continue to use String as the default toString formatter, it seems like we'd need a special (gross) case like:

if (isFirst && Object.is(unit, -0) /* && !unitAsPaddedString.startsWith('-') */) {
    return `-${unitAsString}`;
}

That works for native JavaScript numbers, but it won't work for bigint or Big.js amounts. We'd need a more generic (Calculator-based?) solution.

I tried constructing a generic "negative zero" reference value:

const zero = calculator.zero();
const negativeOne = calculator.decrement(zero);
const negativeZero = calculator.multiply(zero, negativeOne);

... but that doesn't work because (a) the Object.is identity test doesn't work for Big.js instances and equality checks can't distinguish between 0 and -0 values.

It might be possible for sign to help us out, perhaps by getting it to return -0 for negative zero values and using the Object.is test combined with a calculator.zero() comparison, but that just moves the existing (unsolved) problem into sign.

I'm happy to help explore other solutions.

[sarahdayan]

Hey @jparise, good catch!

I think it's fine that toUnits returns a positive zero, since it's equivalent in the language and trying to work against it will generate clumsy, unreliable code.

We can fix this in toDecimal by checking two things:

1. units contains at least one negative value (they should all be negative if the original amount is negative, aside from zeros of course)
2. The first entry of the units array is zero

If both conditions are satisfied, we can pad the returned value with '-'.

I have a working prototype, but if you want to contribute, I'll let you take a stab at it!

[jparise]

Thank you! This was the important insight that I was missing:

units contains at least one negative value (they should all be negative if the original amount is negative, aside from zeros of course)

I have a working version based on your suggestions that I'll push shortly. I appreciate the opportunity to finish the puzzle. 😉