@agrora/decimal

Decimal JS

Arbitrary precision floating point arithmetic library based on PHP's bcmath extension.

It's essentially a port of bcmath, but written from scratch to provide a neat and solid JavaScript/TypeScript API. Like, bcmath for JavaScript, but with a nice API and not just a braindead C-port.

Installation

npm i @agrora/decimal

TypeScript is supported out of the box. You don't need to install any typings.

Usage

import Decimal from '@agrora/decimal';

// Create a new decimal from a string
const decimal = Decimal.from('0.2');

// Use add() to add another decimal to it
const sum = decimal.add('0.1');

// Log the result as a string
console.log(sum.toString()); // "0.3"

Methods can be chained in an immutable way to perform multiple operations at once

const decimal = Decimal.from('2');

const sum = decimal.add('5');

console.log(sum.toString()); // "7"

const result = sum
    .multiply('10')
    .multiply('2')
    .divide('5');

console.log(result.toString()); // "20"
console.log(sum.toString()); // "7" (still!)

Decimals can be created from strings, numbers and other decimal-ish objects!

We call all these possible values DecimalLike in this library.

Decimal.from(5);

Decimal.from(2.2);

Decimal.from('18446744073709551616'); // Higher than a 64bit unsigned int!

Decimal.from('3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196442881097566593344612847564823378678316527...');

Decimal.from(Decimal.from(Decimal.from(Decimal.from(3))));

// If you're feeling hacky
Decimal.from({
    length: 3, 
    scale: 2, 
    sign: '-', 
    value: new Uint8Array([6, 0, 0, 2, 5])
});

API

The Decimal Class

The decimal class acts as a wrapper class around all the functions this library exposes. This is what you want to use for the most part.

import Decimal from '@agrora/decimal';

class Decimal {
    // Factory methods
    static from(value: DecimalLike): Decimal;
    static fromString(value: string): Decimal;
    static fromNumber(value: number): Decimal;
    static fromInfo(value: DecimalInfo): Decimal;
    
    static isDecimal(value: any): value is Decimal;
    static isDecimalLike(value: any): value is DecimalLike;
    
    static min(...values: DecimalLike[]): Decimal;
    static max(...values: DecimalLike[]): Decimal;
    
    // Basic arithmetic
    add(value: DecimalLike, scale?: number): Decimal;
    subtract(value: DecimalLike, scale?: number): Decimal;
    multiply(value: DecimalLike, scale?: number): Decimal;
    divide(value: DecimalLike, scale?: number): Decimal;
    
    modulo(value: DecimalLike, scale?: number): Decimal;
    divideModulo(value: DecimalLike, scale?: number): [Decimal, Decimal];
    
    raise(value: DecimalLike, scale?: number): Decimal;
    
    // Comparisons
    compareTo(value: DecimalLike): 1|0|-1;
    isEqualTo(value: DecimalLike): boolean;
    isGreaterThan(value: DecimalLike): boolean;
    isGreaterThanOrEqualTo(value: DecimalLike): boolean;
    isLowerThan(value: DecimalLike): boolean;
    isLowerThanOrEqualTo(value: DecimalLike): boolean;
    isZero(): boolean;
    isOne(): boolean;
    isMinusOne(): boolean;
    
    negate(): Decimal;
    isNegative(): boolean;
    isPositive(): boolean;
    
    // Conversions
    toString(): string;
    toInt(): number;
    toFloat(): number;
    toFixed(scale: number): number;
}

The rest of the functions are considered internal for now, but feel free to check out the source code.

The Problem it's trying to solve

JavaScript has floating point precision problems due to the way floating point numbers are represented. In fact, this is the case for many if not most languages.

To see the problem, open a browser, press F12, go to "Console" and enter the following

>> 0.1 + 0.2

and hit enter. You will notice it will not result in 0.3 as you might've assumed, but rather in

< 0.30000000000000004

This is a limitation of how floating point numbers can be represented binary.

Obviously this can lead to huge problems in financial and precision-critical applications.

Most languages solve this problem by having a float/double implementation as well as a decimal implementation. While floats/doubles use the floating point processing unit of your CPU, decimals are calculated as strings and numbers below our decimal radix (10), like in typical school maths where you use your hands and carries you remember.

This has two advantages:

• Floating point numbers are calculated precisely and never loose numbers
• Numbers can be bigger than 32 or even 64 bit. In fact, they can be as long as your RAM can store

It also comes at a disadvantage, which is the reason why e.g. graphic applications, games etc. don't worry about precision all too much:

It's expensive. It costs more performance than binary floating point operations. In todays applications, we don't need to worry about this too much. But before using this library, try to understand the difference between a float/double and a decimal and understand when to apply which.

If precision is neglectable (e.g. latitude/longitude, geometry, ray casting etc.), rather use a normal number type. For currencies, weights and other precision-critical numbers use a decimal.

Contributing

Before contributing, check out the Contribution Guidelines

Requires: npm

// Pull project
git clone https://github.com/Agrora/decimal-js

// Enter project directory
cd decimal-js

// Install development dependencies
npm install

// ... make your changes ...

// Run tests
npm run test

// Lint
npm run lint

// Fix linting problems
npm run lint:fix

// Build
npm run build

// ... create branch, commit, push, merge request etc. ...

Future Scope

• Decimal.prototype.getSquareRoot()
• Refactoring so that DecimalInfo.length is not needed anymore (based on DecimalInfo.value.length)
• Refactoring of arithmetic methods to reduce cyclomatic complexity

Credits

Credits of this library and its algorithms go to:

• tjhei (This search result)
• kvz (This search result)
• The original bcmath author Andy Gutmans (Source Code Credits)

Thank you for your initial work to make this library here possible.