Floor / Ceil methods for Number and Period Inconsistency

[Nosferican]

I just noticed that the floor / ceil methods for Number and Period have the positional arguments inverted.

floor(Int, 3.5)
using Dates
floor(DateTime("2000-01-31"), Month)

From Slack, Dates should be consistent with Base.

Related: #18574

[dkarrasch]

Out of curiosity, what has changed that the arguments against this proposal from #18574 are no longer valid?

[Nosferican]

I think just people keep getting confused by the order... Would it be too bad to have both methods?

ceil(dt::TimeType, p::Period) -> TimeType
ceil(p::Period, dt::TimeType) -> TimeType
ceil(x::Period, precision::T) where T <: Union{TimePeriod, Week, Day} -> T
ceil(precision::T, x::Period) where T <: Union{TimePeriod, Week, Day} -> T

[rapus95]

As I already posted in #34810 (comment) I'm thinking that the original considerations in #18574 were right and the current behaviour is not inconsistent but missing functionality. It can be seen easily if we add a function which can round to a period while converting the time type:

round(Date, now(), Month, RoundUp) == Date(2020, 4) #at least for the month of the post 😄

Tl;Dr: rather add those missing functions and make only the 2nd argument a required one

[Moelf]

I'd like to work on this, if I understand this correctly, we'd like the following to pass as tests in the end?

now_ = now()
for p in (Year, Month, Day)
    for r in (RoundUp, RoundDown)
        @test round(Date, now_, p, r) == round(Date(now_), p, r)
    end
    @test floor(Date, now_, p) == round(Date, now_, p, RoundDown)
    @test ceil(Date, now_, p)  == round(Date, now_, p, RoundUp)
end