Support signbit(::Dates.Period)

[jishnub]

signbit(x) is defined to

Return true if the value of the sign of x is negative, otherwise false.

Since sign is defined for a Period, it makes sense for signbit to be defined as well. After this,

julia> sign(Day(1))
1

julia> signbit(Day(1))
false

[LilithHafner]

Seems somewhat reasonable, what's the motivating usecase for adding this feature?

[jishnub]

I don't have a very compelling use case, but I had noticed this when trying to construct an infinite range of dates (i.e. an arbitrarily long sequence of dates starting at a specified point). The current implementation of ranges in InifniteArrays uses signbit, so having this defined improves composability.

[LilithHafner]

Sure, why not

[jishnub]

@halleysfifthinc Could you elaborate why you dislike this proposal?

[halleysfifthinc]

This is completely my own interpretation, since the documentation doesn't explicitly state this, but I understood signbit to only/mainly apply to primitive types (returning the actual sign bit), and sign should be used for all other sign needs (including extending for non-primitive types).

[jishnub]

signbit is already defined for e.g. Rational, which isn't really a primitive type, although it's still a number. There's also a signbit method for Real, so non-primitive number types get the fallback implementation. Now, a Period isn't a number, but it's documented as a simple wrapper for an Int, so perhaps it's reasonable to extend the definition?

That being said, I understand the concern, but I wonder if we want to be strict about this.

[fingolfin]

LGTM