ZeroDivisionError thrown by bx.intervals.intersection.IntervalNode.__cinit__

[borissim]

I create a lot of IntervalNode objects and eventually this error pops up

File "intersection.pyx", line 395, in bx.intervals.intersection.IntervalTree.insert (lib/bx/intervals/intersection.c:4902)
  File "intersection.pyx", line 127, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1657)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 127, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1657)
  File "intersection.pyx", line 118, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1567)
  File "intersection.pyx", line 127, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1657)
  File "intersection.pyx", line 127, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1657)
  File "intersection.pyx", line 120, in bx.intervals.intersection.IntervalNode.insert (lib/bx/intervals/intersection.c:1600)
  File "intersection.pyx", line 91, in bx.intervals.intersection.IntervalNode.__cinit__ (lib/bx/intervals/intersection.c:1330)
ZeroDivisionError: float division

The issue seems to be that at
self.priority = ceil(nlog * log(-1.0/(1.0 * rand()/RAND_MAX - 1)))
I assume rand() can produce 0 or RAND_MAX, at which point (1.0 * rand()/RAND_MAX - 1) equals 0, hence ZeroDivisionError is thrown

[rsharris]

Right. The comment above the failing line claims that "python's uniform is perfect since the upper limit is not inclusive". But rand() isn't from python, it's from C. And the C standard says rand() is inclusive, 0<=rand()<=RAND_MAX.

I think the question is whether the intent was to have (RAND_MAX+1) as the denominator instead of RAND_MAX. That comment makes me think python's random.random() used to be used here.

Secondary questions would be (a) why the C PRNG is being used instead of the python PRNG, since C rand() implementations are usually poor, and (b) why a binomial distribution is being used instead of uniform.

[welyt]

same error， any update？