Reputation: 1161
I have two ordered sets of intervals ((
means "start" and )
means "stop"):
1: ( ) ( )
0---1------3------5---6---7------9---10----> time
2: ( )( ) ( )
In two list it would look like:
intervals1 = [(0,3), (7,10)]
intervals2 = [(0,1), (1,5), (6,9)]
The further evaluation of the time series will be an integration over time of both. For that, I would like to keep the interval character but as common intervals. In the given example, the time series and corresponding list would look like:
common: ( )( )( ) ( )( )( )
0---1------3------5---6---7------9---10----> time
intervals = [(0,1), (1,3), (3,5), (6,7), (7,9), (9,10)]
How can I combine these two time series efficiently?
Upvotes: 1
Views: 681
Reputation: 104712
I think you can efficiently solve this issue with a stack based algorithm, since you know that no more than two intervals can overlap at any given position:
def merge_intervals(a, b):
stack = sorted(a+b, reverse=True)
while len(stack) > 1:
first = stack.pop()
second = stack.pop()
if first == second: # identical intervals can be merged
yield first
elif first[1] <= second[0]: # no overlapping, yield first interval, put back second
yield first
stack.append(second)
elif first[0] == second[0]: # overlap at start, yield shorter, put back rest of longer
if first[1] > second[1]:
first, second = second, first
yield first
stack.append((first[1], second[1]))
elif first[1] < second[1]: # partial overlap, yield first two parts, put back rest
yield first[0], second[0]
yield second[0], first[1]
stack.append((first[1], second[1]))
else: # first[1] >= second[1] # total envelopment
yield first[0], second[0]
yield second
if first[1] != second[1]:
stack.append((second[1], first[1]))
yield from stack # there may or may not be one element left over
This is a generator, so you'll get your desired output with:
intervals = list(merge_intervals(intervals1, intervals2))
Upvotes: 2