Adding new intervals to a contig after querying for intersections does not work.

[TedBrookings]

The OverlapDetector has an issue that its intervaltree must be indexed before any overlaps can be checked, however the cgranges object it is based on will only index once (it checks if it's already indexed and skips a second call).

This results in a bug where adding new intervals to the tree does not result in them being returned for overlap checks.

detector = OverlapDetector()
detector.add(Interval("chr1", 100, 200, name="foo"))
detector.get_overlaps(Interval("chr1", 100, 200))
# yields "foo"
detector.add(Interval("chr1", 101, 201, name="bar"))
detector.get_overlaps(Interval("chr1", 100, 200))
# still only yields "foo"

[TedBrookings]

I can confirm that superintervals does not have this problem. IntervalSet (supposedly, see below) must be re-indexed when new intervals are added, but pybedlite OverlapDetector already does this. superintervals.IntervalSet actually performs the reindex. It looks like a relatively easy fix to replace cgranges:

• IntervalSet add intervals with a start, end, and value, and the value could be set to the index in refname_to_intervals. Then the logic is very similar.
• IntervalSet uses closed intervals, so I think we'd want to add 1 to start in order to maintain backwards compatibility

example:

iset = superintervals.IntervalSet()
iset.add(100, 200, 0)
iset.index()
iset.find_overlaps(100, 200)
# returns [0]
iset.add(100, 200, 1)
iset.index()
iset.find_overlaps(100, 200)
# returns [1, 0]

In my testing this worked even without reindexing. Since in the README it claims indexing is required, I would keep it just to be safe.