clj-time.core/overlaps? provides inconsistent results.

[HennieLouw]

The overlaps function in core has a 'convenience' operation to provide four ReadablePartial instances instead of two Interval instances.

(defn overlaps? [^ReadablePartial start-a ^ReadablePartial end-a  start-b ^ReadablePartial end-b])

However, the implementation is not following the rules of Joda Time.

Intervals are inclusive of the start instant and exclusive of the end.
* An interval overlaps another if it shares some common part of the
* datetime continuum.

Can be reproduced with:

(let [a-start (clj-time.coerce/to-date-time "2019-05-25T00:30:00.000-00:00")
      a-end   (clj-time.coerce/to-date-time "2019-05-25T02:30:00.000-00:00")
      b-start (clj-time.coerce/to-date-time "2019-05-25T02:30:00.000-00:00")
      b-end   (clj-time.coerce/to-date-time "2019-05-25T04:30:00.000-00:00")
      a-iter  (clj-time.core/interval a-start a-end)
      b-inter (clj-time.core/interval b-start b-end)]
  {:overlaps-interval? (clj-time.core/overlaps? a-iter b-inter)
   :overlaps-dates?    (clj-time.core/overlaps? a-start a-end b-start b-end)
   :abuts?             (clj-time.core/abuts? a-iter b-inter)
   })
=> {:overlaps-interval? false, :overlaps-dates? true, :abuts? true}

Note, I'm aware that clj-time is not actively maintained due to Joda Time being replaced with Java Time, However, I suspect that like myself many developers still rely on clj-time because of the existence of cljs-time

[seancorfield]

Good catch. There are two tests for this and they're both wrong.