projectIntoRange speedups

coverage.py

"""
Find coverage in [winStart, winEnd) implied by tStart, tEnd
vectors.

Original from rangeQueries, and two attempts at speeding it up.

In the common case I could imagine projectIntoRangeFast2 being fastest,
but on my amplicons test case projectIntoRangeFast1 wins.
"""

def projectIntoRange(tStart, tEnd, rangeStart, rangeEnd):
    assert(len(tStart) == len(tEnd))

    res = n.array([0] * (rangeEnd - rangeStart))

    for i in range(0, len(tStart)):
        s = max(rangeStart, tStart[i]) - rangeStart
        e = min(rangeEnd, tEnd[i]) - rangeStart

        if (e > s):
            res[s:e] = res[s:e] + 1
    return(res)


def projectIntoRangeFast1(tStart, tEnd, winStart, winEnd):

    # An attempt at speeding up RQ.projectIntoRange
    assert(len(tStart) == len(tEnd))
    res = np.zeros(shape=winEnd-winStart, dtype=np.uint)
    # Clip to window and translate
    tStart_ = np.maximum(winStart, tStart) - winStart
    tEnd_   = np.minimum(winEnd, tEnd) - winStart
    for (s, e) in zip(tStart_, tEnd_):
        res[s:e] += 1
    return res


def projectIntoRangeFast2(tStart, tEnd, winStart, winEnd):
    # FASTER
    # Make the first differences, then integrate
    assert(len(tStart) == len(tEnd))
    deltas = np.zeros(shape=winEnd-winStart, dtype=np.int)
    # Clip to window and translate
    tStart_ = np.maximum(winStart, tStart) - winStart
    tEnd_   = np.minimum(winEnd, tEnd) - winStart
    for (s, e) in zip(tStart_, tEnd_):
        deltas[s] += 1
        deltas[e-1] -= 1
    return np.cumsum(deltas)