I’ve published tempo version 0.5.0 that makes the changes outlined earlier. A few more related changes are noted in the changelog.
Your example revisited
Now your example returns the expected “clipped” period of intersection.
It answers the question "The time when I was both in Japan and enrolled is the period from September 30, 2018 — when enrollment started — to March 31, 2019, when my Japan trip ended."
iex(1)> import Tempo.Sigils
Tempo.Sigils
iex(2)> japan = ~o"2018-05-01/2019-03-31[Asia/Tokyo]"
~o"2018Y5M1D/2019Y3M31D[Asia/Tokyo]"
iex(3)> enrolled = ~o"2018-09-30/2020-06"
~o"2018Y9M30D/2020Y6M"
iex(4)> {:ok, both} = Tempo.intersection(japan, enrolled)
{:ok, #Tempo.IntervalSet<[~o"2018Y9M30D/2019Y3M31D[Asia/Tokyo]"]>}
New members_overlapping/2 function
The old intersection/2 behaviour is now in members_overlapping/2.
It answers the question “is my Japan trip one of the items that overlaps my enrollment window?”
{:ok, hits} = Tempo.members_overlapping(japan, enrolled)
#=> {:ok, #Tempo.IntervalSet<[~o"2018Y5M1D/2019Y3M31D[Asia/Tokyo]"]>}
Thanks for reporting this - I think the result is an important improvement in applying expected behaviour in set operations.
