Differences smaller than Dates.Milliseconds are ignored

Internals & Design
, ,
1

I have a really hard time understanding this:

julia> using Dates

julia> dt1 = DateTime(1)
0001-01-01T00:00:00

julia> for timeperiod in (Hour, Minute, Second, Millisecond, Microsecond, Nanosecond)
           dt2 = dt1 + oneunit(timeperiod)
           @show dt1 ≠ dt2
       end
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = false
dt1 ≠ dt2 = false

Why are differences smaller than milliseconds ignored…? Has this been extensively covered somewhere…?

2

The underlying integer simply cannot represent that kind of a precision. Cf

using Dates
d = DateTime(1)
d2 = d + oneunit(Microsecond)
d.instant.periods.value == d2.instant.periods.value
3

OK, so I think I understand why it doesn’t work.

help?> DateTime
search: DateTime datetime2unix datetime2rata datetime2julian ISODateTimeFormat unix2datetime

  DateTime

  DateTime wraps a UTInstant{Millisecond} and interprets it according to the proleptic Gregorian
  calendar.

But man, I can’t help think that it’s misleading to be able to add a Nanosecond, in fact 10^6 - 1 of them without noticing any difference. Hmm…

4

Like, if I want to be able to do arithmetic on DateTime and Nanosecond I guess I should convert everything to Milliseconds…

5

Cf

julia> 1.0 + eps()/2
1.0
2 Likes
6

Touché…

7

That is a really good point. OK. I’ll deal. Thanks again @Tamas_Papp!!!

1 Like
8

You could give TimesDates.jl a shot:

julia> using TimesDates,Dates

julia> dt1 = TimeDate(1)
0001-01-01T00:00:00

julia> for timeperiod in (Hour, Minute, Second, Millisecond, Microsecond, Nanosecond)
                  dt2 = dt1 + oneunit(timeperiod)
                  @show dt1 ≠ dt2
              end
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
dt1 ≠ dt2 = true
2 Likes