Adding tests revealing 0-length interval nonsense - #1320
Conversation
Adding tests that test intersection/abutment using 0-length intervals. Many are failing... I've added an ascii diagram in the test to show my belief of what the various intervals represent. Zero length intervals don't intersect themselves. Should they? The equivalent methods in Interval and Locatable have slightly different failure modes.
| {chr1_22_30, zeroLength_chr1_22_21, true, false}, | ||
| {chr1_23_30, zeroLength_chr1_22_21, false, true}, | ||
| {chr1_24_30, zeroLength_chr1_22_21, false, false}, | ||
| {zeroLength_chr1_22_21, zeroLength_chr1_22_21, true, false}, |
There was a problem hiding this comment.
hmmm. to answer his question we need a definition of "intersects" and "abuts"
Intersects: There exists at least one base in the reference that is present in both intervals.
Abuts: Two intervals that do not intersect (c.f.) , however, the union of bases they cover can be written as one interval.
According to this definition, two empty intervals in the same position do not intersect, but they do abut.
There was a problem hiding this comment.
we have both intersects and overlaps? huh?
There was a problem hiding this comment.
it's the same function, implemented twice!
There was a problem hiding this comment.
We have more in other places I'm sure. Because of fear or weird details like this we never merged the implementations of a lot of these.
4 participants
This PR adds failing tests that test intersection/abutment using 0-length intervals.
Many are failing...
I've added an ascii diagram in the test to show my belief of what the various intervals represent.
The equivalent methods in
IntervalandLocatablehave slightly different failure modes.We should decide what the right answer is, and then fix the various methods to implement those.
We should also check related methods like
enclosesandoverlapinCoordMath.I'm sick of this issue and want it to be resolved so we never have to think about it again.