A small library that parses comma-delimited integer ranges (such as "1-3,8-10") and manipulates such range data. This type of data is commonly used to specify which lines to highlight or which pages to print.

Key features:

• Addition (aka union, e.g., 1-2,6 + 3-5 → 1-6)
• Subtraction (e.g., 1-10 − 5-9 → 1-4,10)
• Inclusion check (e.g., 3,7-9 ⊂ 1-10)
• Intersection (e.g., 1-5 ∩ 2-8 → 2-5)
• Unbounded ranges (aka infinite ranges, e.g., 5-, meaning "all integers ≥ 5")
• Ranges including negative integers or zero
• ES6 iterator (for ... of, spread operator)
• Array building ("flatten")

The range data are always sorted and normalized to the smallest possible representation.

🚨 Note: The following README is for the 5.x release, whose API has changed drastically. For the docs of the 4.x release, see this.

Install via npm or yarn:

npm install multi-integer-range

Version 5 is a hybrid package; it provides both a CommonJS version and an ES Module version, built from the same TypeScript source. Bundlers such as Webpack can automatically pick the ESM version and perform tree-shaking. This package has no external dependencies nor does it use any Node-specific API.

🚨 The API style has changed drastically in version 5. The new API is slightly more verbose, but is simpler and tree-shakable 🌲. For example, if you don't use the default parser, your bundle will not include it. See the CHANGELOG and the docs for version 4.

import * as mr from 'multi-integer-range';

const ranges1 = mr.parse('1-6,9-12'); // [[1, 6], [9, 12]]
const ranges2 = mr.parse('7-10, 100'); // [[7, 10], [100, 100]]
const ranges3 = mr.normalize([1, 5, 6, [4, 2]]); // [[1, 6]]

const sum = mr.append(ranges1, ranges2); // [[1, 12], [100, 100]]
const diff = mr.subtract(ranges1, ranges2); // [[1, 6], [11, 12]]
const commonValues = mr.intersect(ranges1, ranges2); // [[9, 10]]

const str = mr.stringify(sum); // "1-12,100"
const bool = mr.has(ranges1, ranges3); // true
const isSame = mr.equals(ranges1, ranges2); // false
const array = mr.flatten(diff); // [1, 2, 3, 4, 5, 6, 11, 12]
const len = mr.length(ranges1); // 10

The fundamental data structure of this package is a normalized array of [min, max] tuples, as shown below. Here, 'normalized' means the range data is in the smallest possible representation and is sorted in ascending order. You can denote an unbounded (aka infinite) range using the JavaScript constant Infinity.

type Range = readonly [min: number, max: number];
type MultiIntegerRange = readonly Range[];

// Examples of normalized MultiIntegerRanges
[[1, 3], [5, 6], [9, 12]] // 1-3,5-6,9-12
[[-Infinity, 4], [7, 7], [10, Infinity]] // -4,7,10-
[[-Infinity, Infinity]] // all integers
[] // empty

// These are NOT normalized. Don't pass them to append() and such!
[[3, 1]] // min is larger than max
[[7, 9], [1, 4]] // not in the ascending order
[[1, 5], [3, 7]] // there is an overlap of ranges
[[1, 2], [3, 4]] // the two ranges can be combined to "1-4"
[[Infinity, Infinity]] // makes no sense

Most functions take one or two normalized MultiIntegerRanges as shown above to work correctly. To produce a valid normalized MultiIntegerRange, you can use normalize(), parse() or initialize(). You can write a normalized MultiIntgerRange by hand as shown above, too.

normalize(data?: number | (number | Range)[]) creates a normalized MultiIntegerRange from a single integer or an unsorted array of integers/Ranges. This and initialize are the only functions that can safely take an unsorted array. Do not pass unnormalized range data to other functions.

console.log(mr.normalize(10)); // [[10, 10]]
console.log(mr.normalize([3, 1, 2, 4, 5])); // [[1, 5]]
console.log(mr.normalize([5, [2, 0], 6, 4])); // [[0, 2], [4, 6]]
console.log(mr.normalize([7, 7, 10, 7, 7])); // [[7, 7], [10, 10]]
console.log(mr.normalize()); // []

// Do not directly pass an unnormalized array
// to functions other than normalize().
const unsorted = [[3, 1], [2, 8]];
const wrong = mr.length(unsorted); // This won't work!
const correct = mr.length(mr.normalize(unsorted)); // 8

parse(data: string, options?: Options) creates a normalized MultiIntegerRange from a string. The string parser is permissive and accepts space characters before/after comma/hyphens. It calls normalize() under the hood, so the order is not important, and overlapped numbers are silently ignored.

console.log(mr.parse('1-3,10')); // [[1, 3], [10, 10]]
console.log(mr.parse('3,\t8-3,2,3,\n10, 9 - 7 ')); // [[2, 10]]

By default, the string parser does not try to parse unbounded ranges or negative integers. You need to pass an options object to modify the parsing behavior. To avoid ambiguity, all negative integers must always be enclosed in parentheses. If you don't like the default parse(), you can always create and use your custom parsing function instead, as long as it returns a normalized MultiIntegerRange.

console.log(mr.parse('7-')); // throws a SyntaxError

console.log(mr.parse('7-', { parseUnbounded: true })); // [[7, Infinity]]
console.log(mr.parse('(-7)-(-1)', { parseNegative: true })); // [[-7, -1]]
console.log(
  mr.parse('0-,(-6)-(-2),-(-100)', {
    parseUnbounded: true,
    parseNegative: true
  })
); // [[-Infinity, -100], [-6, -2], [0, Infinity]]

See api-reference.md.

Since a MultiIntegerRange is just an array of Ranges, if you naively iterate over it (e.g., in a for-of loop), you'll simply get each Range tuple one by one. To iterate each integer contained in the MultiIntegerRange instead, use iterate() like so:

const ranges = mr.parse('2,5-7');

for (const page of ranges) {
  console.log(page);
} // prints 2 items: [2, 2] and [5, 7]

for (const page of mr.iterate(ranges)) {
  console.log(page);
} // prints 4 items: 2, 5, 6 and 7

// array spreading (alternative of flatten())
const arr1 = [...mr.iterate(ranges)]; //=> [2, 5, 6, 7]
const arr2 = Array.from(mr.iterate(ranges)); //=> [2, 5, 6, 7]

Intersection is especially useful to "trim" unbounded ranges.

const userInput = '-5,15-';
const pagesInMyDoc = [[1, 20]]; // 1-20
const pagesToPrint = mr.intersect(
  mr.parse(userInput, { parseUnbounded: true }),
  pagesInMyDoc
); // [[1, 5], [15, 20]]
for (const page of mr.iterate(pagesToPrint)) await printPage(page);

For compatibility purposes, version 5 exports the MultiRange class and multirange function, which is mostly compatible with the 4.x API but has been rewritten to use the new functional API under the hood. See the 4.x documentation for the usage. The use of this compatibility layer is discouraged because it is not tree-shakable and has no performance merit. Use this only during migration. These may be removed in the future.

Performance Considerations: This library works efficiently for large ranges as long as they're mostly continuous (e.g., 1-10240000,20480000-50960000). However, this library is not intended to be efficient with a heavily fragmented set of integers that are scarcely continuous (e.g., random 10000 integers between 1 to 1000000).

No Integer Type Checks: Make sure you are not passing floating-point numbers to this library. For example, don't do normalize(3.14). For performance reasons, the library does not check if a passed number is an integer. Passing a float will result in unexpected and unrecoverable behavior.

range-parser specializes in parsing range requests in HTTP headers as defined in RFC 7233, and it behaves in a way that is usually inappropriate for other purposes. For example, '-5' means "last 5 bytes".

parse-numeric-range is fine for small ranges, but it always builds a "flat" array, which makes it very inefficient for large ranges such as byte ranges. Also, whether you like it or not, it handles overlapping or descending ranges as-is, without normalization. For example, '4-2,1-3' results in [4, 3, 2, 1, 2, 3].

multi-integer-range is a general-purpose library for handling this type of data structure. It has a default parser that is intuitive enough for many purposes, but you can also use a custom parser. Its real value lies in its ability to treat normalized ranges as intermediate forms, allowing for a variety of mathematical operations. See the API reference.

¹: With combine option. ²: With parseUnbounded option. ³: When size is 10000.

To test:

npm ci
npm test

To generate CJS and ESM builds:

npm ci
npm run build

Please report bugs and suggestions using GitHub issues.

See CHANGELOG.md.

Soichiro Miki (https://github.com/smikitky)

MIT