Segment trees are a data structure that can be used for a variety of problems. Usually, I’ve seen them used for range queries, but they can also be used for finding overlapping intervals.
When applied to the problem of answering range queries (e.g. sum on an interval), the key idea with segment trees is to split the problem space into smaller parts during preprocessing, so that during query we can visit only a small part of the problem space. Exactly the same idea is used for finding overlapping intervals.
One important remark is that we’re considering fixed intervals, i.g. we know all intervals in advance before we do any preprocessing, and we can’t insert/remove intervals after preprocessing.
Let’s consider 10 random intervals. We can generate them using the following code:
function randomIntervals(n) {
let intervals = [];
for (let i = 0; i < n; i++) {
let start = random(0, containerWidth)
let end = start + random(0, (containerWidth - start) / 2)
intervals.push([start, end, i])
}
return intervals;
}
and they look like this:
Instead of working directly with $n$ float coordinates, we’ll convert intervals to an array of $N = 2n$ endpoints, which we will also sort by their value. On the picture above the endpoints are marked with magenta vertical lines.
It’s important to both sort the endpoints by value and keep the reference to the original interval index. We don’t need to know whether that inteval is a start or an end, but we’ll need indices later.
function Endpoint(v, ix) {
this.v = v
this.ix = ix
}
function buildEndpoints(intervals) {
eps = []
for (const [ix, [start, end]] of intervals.entries()) {
eps.push(new Endpoint(start, ix))
eps.push(new Endpoint(end, ix))
}
eps.sort((a, b) => a.v - b.v)
return eps
}
The next step is to build the auxilary nodes in the tree.
function Node(b, e) {
this.b = b
this.e = e
this.left = null
this.right = null
this.key = 0
this.aux = []
}
function build(eps, level, s, t) {
var v = new Node(s, t)
mark(eps, level, s, t, cBuilt, 20, 2)
if (s+1 == t) { return v }
const m = Math.floor((s+t)/2)
v.key = m
v.left = build(eps, level+1, s, m)
v.right = build(eps, level+1, m, t)
return v
}
More content