What determines equivalence between Intervals?

[daler]

What determines equivalence between Intervals?

[brentp]

still thinking about this. But once you have str(self), you can just test for equality, no need to hash.
So you could just have

def __hash__(self):
    return str(self)

in which case, all you really need is to test if str(self) == str(other) in __richcmp__

all that said, i kinda like the idea of 2 intervals being equal if their starts and stops are the same.

[daler]

good points. str(self) == str(other) is easy to do if you want it, but testing chrom/start/stop is more annoying, and therefore better encapsulated in __richcmp__.

two somewhat trickier questions though: 1) should strand continue to be ignored, and 2) should > and < have any meaning?

[brentp]

1. seems intuitive that strand should be taken into account.

2. maybe:
   <: strictly to the left of (not touching)
   >: strictly to the right of (not touching)
   <=: not >
>=: not <

[daler]

sounds good. i guess as long as it's well documented, there shouldn't be any surprises.

[daler]

This has been implemented for a while now, using "profiles" of feature1/feature2 start/stop comparisons -- see
http://packages.python.org/pybedtools/topical-comparisons.html

Strand still isn't considered . . . just haven't needed it yet. Also, the semantics are currently different than what you suggest because I wasn't thinking about your comment when I was writing the code. That is, slightly overlapping features are both <= as well as <. But with the "profiles" this should be easy to change if needed -- see

pybedtools/pybedtools/cbedtools.pyx

Line 25 in b494b6b

# Keys are tuples of start/start, stop/stop, start/stop, stop/start.

.