fix: inverting year fraction should give opposite sign

[borisdavid]

Context

While comparing daycount and elgorian, I noticed an inconsistency on daycount side concerning ActualActualAFB which produced a wrong value when to was before from and a 29th February was comprise within the interval [to, from].

Content

Since the year fraction computation is done on a backward induction, the function now inverts from and to if to is before from.

A property test is added to make sure that inverting from and to returns the opposite value (* -1.0).

[RemiMattheyDoret]

Tiny detail. I think it makes the swap a bit more obvious if it is written as

Suggested change

	from, to = to, remaining
	from, to = to, from

[RemiMattheyDoret]

Would sign, directionality, dateOrdering, dateOrderingMultiplier, isAfter, fromAfterTo, isFromAfterTo be any better than alpha? (alpha might be just fine, I'm just sending in ideas).

Sadly, in Go, we cannot multiply by a bool (and we cannot simply cast a bool into an int), as alpha, really, should be a bool. Of course, we could do

sign := true

asInt := func(b bool) int {
    if b {
         return 1
    }
    
    return -1
}

return asInt(sign) * (float64(nbFullYears) + float64(remaining.Sub(from))/computeYearDurationAFB(from, remaining))

but that might be considered as over-engineering for such a small thing.

[olivierjacob]

I asked myself the same question :)
It's generally called signum (at least in Java SDK)

[RemiMattheyDoret]

I like signum!

It is pronounced sig num (with the gueh sound), not sign um (with the nieh sound), right? 😅