Numeric Calculator

Calculate addition, subtraction, multiplication, division precisely, for tackling the precision issue

---

Installation

It is available as an npm package.

// with npm
npm install @rinxun/numeric-calculator

// with yarn
yarn add @rinxun/numeric-calculator

License

This project is licensed under the terms of the MIT license.

Why do we need this

0.1 + 0.2 = 0.30000000000000004; 
0.1 + 0.2 - 0.3 = 5.551115123125783e-17;

Constructor

• new NC(operand: IOperand, config: IConfig)

  • number: Initial numeric

    • optional
    • type IOperand = instance | number | string

  • config: Configurations

    • optional

    • config = {
        precision?: number, // global default precision, default to 15, should be in the range 0 - 20, effects `toPrecision` when `toPrecision` has no args
        fractionDigits?: number, // global default decimal length, default to 2, should be in the range 0 - 20, effects `toFixed` when `toFixed` has no args
        enableCheckBoundary?: boolean // if true, it will check if the value is out of the safe boundary
      }

Methods

The instance supports chain operations.

• plus(operands: IOperand[]) => instance
• minus(operands: IOperand[]) => instance
• times(operands: IOperand[]) => instance
• divide(operands: IOperand[]) => instance
• round(fractionDigits?: number) => number
• toPrecision(precision?: number) => number
• toFixed(fractionDigits?: number) => string

Usage

Import

// CommonJS
const { default: NC } = require('@rinxun/numeric-calculator');

// ES6 
import NC from '@rinxun/numeric-calculator';

Instantiate

const inst1 = new NC(); 
// or with initial numeric
const inst2 = new NC(10); 
const inst3 = new NC('10'); 
const inst4 = new NC(new NC(10)); 
// or with config
const inst5 = new NC(10, { enableCheckBoundary: true }); 

inst.plus(operands: IOperand[])

const result = new NC().plus(2.134).toPrecision(); // 2.134
const result = new NC().plus(2.1341, new NC(1.09).plus(2), '0.65').toPrecision(); // => 2.1341+(1.09+2)+0.65 = 5.8741
const result = new NC(10).plus(2.134).toPrecision(); // => 10+2.134 = 12.134
const result = new NC(10).plus(2.1341, '3.09', 0.65).toPrecision(); // => 10+2.1341+3.09+0.65 = 15.8741

inst.minus(operands: IOperand[])

const result = new NC().minus(0.4).toPrecision(); // 0.4
const result = new NC().minus(0.25, new NC(2.5).minus(1), '3.498').toPrecision(); // => 0.25-(2.5-1)-3.498 -4.748
const result = new NC(2).minus(0.4).toPrecision(); // => 2-0.4 = 1.6
const result = new NC(-2).minus(0.25, 1.5, 3.498).toPrecision(); // -2-0.25-1.5-3.498 = -7.248

inst.times(operands: IOperand[])

const result = new NC().times(0.12).toPrecision(); // 0.12
const result = new NC().times(0.12, new NC(8).times(12.5), '5').toPrecision(); // => 0.12*(8*12.5)*5 = 60
const result = new NC(-4).times(0.12).toPrecision(); // => -4*0.12 = -0.48
const result = new NC(3.2).times(0.12, '100', 5).toPrecision(); // => 3.2*0.12*100*5 = 192

inst.divide(operands: IOperand[])

const result = new NC().divide(3).toPrecision(); // 3
const result = new NC().divide(100, new NC(60).divide(3), '2').toPrecision(); // => 100/(60/3)/2 = 2.5
const result = new NC(36.12).divide(3).toPrecision(); // => 36.12/3 = 12.04
const result = new NC(-1000).divide(100, '20', 2).toPrecision(); // -1000/100/20/2 = -0.25

inst.round(fractionDigits?: number, options?: { snap?: boolean })

const result = new NC().round(3); // 0
const result = new NC(5.255).round(2); // 5.26
const result = new NC(7.65).times(16.5).round(2); // => 7.65 * 16.5 = 126.225 ≈ 126.23
const result = new NC(26.05).times(16.5).round(2); // => 26.05 * 16.5 = 429.825 ≈ 429.83

options.snap (default false) makes rounding tolerant to binary floating-point representation dust. After plain float arithmetic upstream, a true 95.015 can arrive as 95.01499999999942, and exact-decimal rounding of the represented value then rounds down, disagreeing with the true value. With snap, the amplified value is nudged by a few ULPs relative to its own magnitude before Math.round, so only values sitting exactly on a rounding boundary are affected — everything beyond a few ULPs of the boundary keeps its exact-decimal result, and there is no magic number to tune:

const dust = 2959.9649999999997; // true value 2959.965, distorted by float subtraction upstream
const result = new NC(dust).round(2); // 2959.96 — exact decimal of the represented value
const result = new NC(dust).round(2, { snap: true }); // 2959.97 — matches the true half-up value
const result = new NC(719.6446).round(2, { snap: true }); // 719.64 — values beyond a few ULPs of the boundary are untouched

inst.toPrecision(precision?: number)

const result = new NC(1.2345).toPrecision(); // 1.2345
const result = new NC(1.2345).toPrecision(4); // 1.234
const result = new NC(1.2e-2).toPrecision(); // 0.012

inst.toFixed(fractionDigits?: number)

Follow the Banker's Rounding rules

const result = new NC(1.2345).toFixed(); // '1.23'
const result = new NC(1.2345).toFixed(3); // '1.234'
const result = new NC(1.2346).toFixed(3); // '1.235'
const result = new NC(1.2e-2).toFixed(); // '0.01'

chain operations

const result = new NC(10).plus(2).minus(4).times(2).divide(8).toPrecision(); // => (10+2-4)*2/8 = 2
const result = new NC(10).plus(new NC().times(2, 4), '6').divide(new NC(12).minus(4)).toPrecision(); // => (10+(2*4)+6)/(12-4) = 3

Changelog

If you have recently updated, please read the changelog for details of what has changed.