A Range and RangeSet implementation based on search trees.
This library is a replacement for array-based range/interval handling libraries drange
and range-set designed with performance on large sets in mind.
It offers better search complexity, O(log(n))
vs O(n)
, and a better average update performance, with the worst case
scenario matching the former's O(n)
, see Performance. The performance improvements happen at a cost of higher memory consumption,
depending on an underlying search tree implementation.
This library's design was inspired by Guava's RangeSet.
This library comes with Typescript typings built in, so you only need to include the package itself:
npm i tree-range-set
yarn add tree-range-set
This library was built and tested with Node.js in mind, but there should be no reason for it not work in a modern browser. That being said, handling large sets of ranges in a browser might be suboptimal, so you might want to look at drange for that instead.
If you're planning on using it with a custom type, implementing a RangeSpec for your type might be required (see below for details).
These are some of the basic use cases:
import { RangeSet, Range } from "tree-range-set";
// Create a range of natural numbers
const naturalRange = RangeSet.numeric().add(Range.closeOpen(1, Infinity));
naturalRange.contains(2); // yes, a natural number
// All the rational numbers?
const rationalRange = RangeSet.numeric().add(Range.open(-Infinity, Infinity));
naturalRange.contains(rationalRange); // false, but it works the other way around:
rationalRange.contains(naturalRange); // yes!
// What if we only want non-natural numbers?
const nonNaturalRange = rationalRange.subtract(naturalRange);
nonNaturalRange.contains(-1); // yes
nonNaturalRange.contains(1); // no, 1 is a natural number
import { RangeSet, Range, AbstractRangeSpec, RangeSpec } from "tree-range-set";
class DateSpec extends AbstractRangeSpec<Date> implements RangeSpec<Date> {
get comparator(): (a: Date, b: Date) => number {
return (a, b) => a.getTime() - b.getTime();
}
// Infinity is optional, it provides special handling for closed ranges enclosed by infinity
isInfinity(value: Date): boolean {
return false;
}
}
// Instantiate the spec
const spec = new DateSpec();
const todayRange = Range.closeOpen(
new Date("2020-07-01T00:00:00Z"),
new Date("2020-07-02T00:00:00Z"),
spec);
const workingDaysRange = RangeSet.of(spec)
.add(todayRange)
.add(Range.closeOpen(
new Date("2020-07-02T00:00:00Z"), new Date("2020-07-02T00:00:00Z"), spec)
)
.add(Range.closeOpen(
new Date("2020-07-05T00:00:00Z"), new Date("2020-07-05T00:00:00Z"), spec)
);
// Now we can query them:
todayRange.contains(new Date("2020-07-01T11:00:00Z")); // yes
workingDaysRange.contains(new Date("2020-07-01T11:00:00Z")); // yes, it's a working day
// We now can set up convenience functions to check dates based on our ranges:
const canScheduleMeeting = (desiredTime: Range<Date>) => workingDaysRange.contains(desiredTime);
canScheduleMeeting(
Range.closeOpen(new Date("2020-07-02T12:00:00Z"), new Date("2020-07-02T13:00:00Z"), spec)
); // yes
canScheduleMeeting(
Range.closeOpen(new Date("2020-07-04T12:00:00Z"), new Date("2020-07-04T13:00:00Z"), spec)
); // no, we don't work on the 4th.
While it is possible to provide an alternative tree implementation, bintrees
work fine, and this README is
too narrow to cover such an advanced topic. This
should provide a good starting point for such a task.
The below figures were measured by Benchmark.js with this suite. I did my best to isolate potential bias, i.e. in creating relevant configs, but I'm open to future suggestions on how to improve this benchmark.
drange x 14.42 ops/sec ±2.31% (40 runs sampled)
range-set x 1.68 ops/sec ±17.26% (9 runs sampled)
tree-range-set x 67.37 ops/sec ±7.37% (58 runs sampled)
The numbers represent adding and subtracting around 8,000 ranges in a single set, in separate installments of 8 (add 4, subtract 4).