interval A overlap interval B

[yinshiyi]

Goal to output that part of none overlap. AKA, the part of A that is not overlapping with B.

setdiff(a,b) a 2023-01-09 09:00:00 PST--2023-01-09 17:00:00 PST b 2023-01-09 11:00:00 PST--2023-01-09 12:30:00 PST

# insert reprex here
Error in setdiff.Interval(a, b) : 
  Cases 1 result in discontinuous intervals.

setdiff(b,a) resulted in output as b

Any suggestions, thank you

[vspinu]

We don't have this funtionality currently. If we are getting into setoperators with intervals then the big question how do we treat vector intervals. You are asking for element wise operation, but R set operations work on vectors.

[DavisVaughan]

I think you should use ivs for this, which has full support for both match-like set ops and pairwise set ops

library(ivs)

a <- iv(
  start = as.POSIXct("2023-01-09 09:00:00", tz = "America/Los_Angeles"),
  end = as.POSIXct("2023-01-09 17:00:00", tz = "America/Los_Angeles")
)
b <- iv(
  start = as.POSIXct("2023-01-09 11:00:00", tz = "America/Los_Angeles"),
  end = as.POSIXct("2023-01-09 12:30:00", tz = "America/Los_Angeles")
)

# Note that `b` is fully contained within `a`
a
#> <iv<datetime<America/Los_Angeles>>[1]>
#> [1] [2023-01-09 09:00:00, 2023-01-09 17:00:00)
b
#> <iv<datetime<America/Los_Angeles>>[1]>
#> [1] [2023-01-09 11:00:00, 2023-01-09 12:30:00)

# So this makes two intervals.
# It treats each vector (`a` and `b`) as one full set of intervals.
# `a` and `b` don't have to be the same length, i.e. they are compared in a match()-like way
# like setdiff()
iv_difference(a, b)
#> <iv<datetime<America/Los_Angeles>>[2]>
#> [1] [2023-01-09 09:00:00, 2023-01-09 11:00:00)
#> [2] [2023-01-09 12:30:00, 2023-01-09 17:00:00)

# This isn't allowed.
# Both `a` and `b` must be the same size, and removing `b` from `a` must
# result in a single interval, which isn't possible here.
iv_pairwise_difference(a, b)
#> Error in `iv_pairwise_difference()`:
#> ! Can't compute a difference when `y` is completely contained within `x`.
#> ℹ This would result in two distinct intervals for a single observation.
#> ℹ Location 1 contains this issue.

I'm not sure that lubridate should add much more functionality with intervals

[vspinu]

Indeed. Given that this is not a common request and that there is a very good tooling available it's not wise to pursue it here.