@roundingwellos/libphonenumber-js

libphonenumber-js

A simpler (and smaller) rewrite of Google Android's famous libphonenumber library: easy phone number parsing and formatting in javascript.

See Demo

LibPhoneNumber

libphonenumber is a phone number formatting and parsing library released by Google, originally developed for (and currently used in) Google's Android mobile phone operating system. Implementing a rigorous phone number formatting and parsing library was crucial for the phone OS overall usability (back then, in the early 2000s, it was originally meant to be a phone after all, not just a SnapChat device).

libphonenumber-js is a simplified pure javascript port of the original libphonenumber library (written in C++ and Java because those are the programming languages used in Android OS). While libphonenumber has an official javascript port which is being maintained by Google, it is tightly coupled to Google's closure javascript utility framework. It still can be compiled into one big bundle which weighs 220 KiloBytes — quite a size for a phone number input component. It can be reduced to a specific set of countries only but that wouldn't be an option for a worldwide international solution.

One part of me was curious about how all this phone matching machinery worked, and another part of me was curious if there's a way to reduce those 220 KiloBytes to something more reasonable while also getting rid of the closure library and rewriting it all in pure javascript. So, that was my little hackathon for a couple of weeks, and seems that it succeeded. The resulting library does everything a modern web application needs while maintaining a much smaller size of about 75 KiloBytes.

Difference from Google's libphonenumber

• Pure javascript, doesn't require any 3rd party libraries
• Metadata size is just about 75 KiloBytes while the original libphonenumber metadata size is about 200 KiloBytes
• Better "as you type" formatting (and also more iPhone-alike style)
• Doesn't parse alphabetic phone numbers like 1-800-GOT-MILK as we don't use telephone sets in the XXIst century that much (and we have phonebooks in your mobile phones)
• Doesn't handle carrier codes: they're only used in Colombia and Brazil, and only when dialing within those countries from a mobile phone to a fixed line number (the locals surely already know those carrier codes by themselves)
• Assumes all phone numbers being formatted are internationally diallable, because that's the only type of phone numbers users are supposed to be inputting on websites (no one inputs short codes, emergency telephone numbers like 911, etc.)
• Doesn't parse phone numbers with extensions (again, this is not the type of phone numbers users should input on websites — they're supposed to input their personal mobile phone numbers, or home stationary phone numbers if they're living in an area where celltowers don't have a good signal, not their business/enterprise stationary phone numbers)
• Doesn't use possibleDigits data to speed up phone number pre-validation (it just skips to the regular expression check itself)
• Doesn't distinguish between fixed line, mobile, pager, voicemail, toll free and other XXth century bullsh*t
• Doesn't format phone numbers for "out of country dialing", e.g. 011 ... in the US (again, just use the +... notation accepted worldwide for mobile phones)
• Doesn't parse tel:... URIs (RFC 3966) because it's not relevant for user-facing web experience
• When formatting international numbers replaces all braces, dashes, etc with spaces (because that's the logical thing to do, and leaving braces in an international number isn't)

Installation

npm install libphonenumber-js --save

Usage

import { parse, format, asYouType } from 'libphonenumber-js'

parse('8 (800) 555 35 35', 'RU')
// { country: 'RU', phone: '8005553535' }

format('2133734253', 'US', 'International')
// '[phone]'

new asYouType().input('[phone]')
// '[phone]'
new asYouType('US').input('2133734')
// '(213) 373-4'

Country code definition

"Country code" means either a two-letter ISO country code (like US) or a special 001 country code used for non-geographical entities (as per Google's libphonenumber library). For example, [phone] phone number belongs to Russia so it has RU country code where as [phone] phone number could belong to any country so it has 001 country code.

API

parse(text, options)

options can be either an object

country:
{
  restrict — (country code)
             the phone number must be in this country

  default — (country code)
            default country to use for phone number parsing and validation
            (if no country code could be derived from the phone number)
}

or just a country code which is gonna be country.restrict.

Returns { country, phone } where

• country is a country code
• phone is a national (significant) number

If the phone number supplied isn't valid then an empty object {} is returned.

parse('[phone]') === { country: 'US', phone: '2133734253' }
parse('[phone]', 'US') === { country: 'US', phone: '2133734253' }

format(parsed_number, format)

Formats a phone number using one of the following formats:

• International — e.g. [phone]
• International_plaintext — (aka E.164) e.g. [phone]
• National — e.g. [phone]

parsed_number argument must be an already parse()d phone number (to strip national prefix from it). That means that first a phone number is parse()d and only then is it format()ted and there's no other way around it. For example, a phone number is parse()d before storing it in a database and then it is forrmat()ted each time it is read from the database. The parsed_number object argument can also be expanded into two string arguments (for those who prefer this kind of syntax):

format({ country: 'US', phone: '2133734253' }, 'International') === '[phone]'
format('2133734253', 'US', 'International') === '[phone]'

// The following won't work becase the phone number argument is invalid
// (has not been parsed previously and therefore contains the `0` national prefix)
format('017212345678', 'DE', 'International_plaintext') !== '[phone]'

getNumberType(parsed_number)

Determines phone number type (fixed line, mobile, toll free, etc). This function will work if --extended metadata is available (see Metadata section of this document). The regular expressions used to differentiate between various phone number types consume a lot of space (two thirds of the total size of the --extended library build) therefore they're not included in the bundle by default.

The arguments can be

• either the result of the parse() function call: { country, phone }
• or a pair of arguments (phone, country_code) in which case it is passed to the parse() function right away and the resulting { country, phone } object is used

getNumberType('8005553535', 'RU') === 'MOBILE'

isValidNumber(parsed_number)

(aka is_valid_number)

Checks if a phone number is valid.

The arguments can be

• either the result of the parse() function call: { country, phone }
• or a pair of arguments (phone, country_code) in which case it is passed to the parse() function right away and the resulting { country, phone } object is used

isValidNumber('[phone]') === true
isValidNumber('+1-213-373') === false

isValidNumber('[phone]', 'US') === true
isValidNumber('(213) 37', 'US') === false

isValidNumber({ phone: '2133734253', country: 'US' }) === true

The difference between using parse() and isValidNumber() for phone number validation is that isValidNumber() also checks the precise regular expressions of possible phone numbers for a country. For example, for Germany parse('123456', 'DE') would return { country: 'DE', phone: '123456' } because this phone number matches the general phone number rules for Germany. But, if the metadata is compiled with --extended flag (see below) and the precise regular expressions for possible phone numbers are included in the metadata then isValidNumber() is gonna use those precise regular expressions for validation and isValid('123456', 'DE') will return false because the phone number 123456 doesn't actually exist in Germany.

So, the general phone number rules for a country are mainly for phone number formatting: they dictate how different phone numbers (matching those general regular expressions) should be formatted. And parse() uses only those general regular expressions (as per the reference Google's libphonenumber implementation) to perform basic phone number validation. isValidNumber(), on the other hand, is all about validation, so it digs deeper into precise regular expressions (if they're included in metadata) for possible phone numbers in a given country. And that's the difference between them: parse() parses phone numbers and loosely validates them while isValidNumber() validates phone number precisely (provided the precise regular expressions are included in metadata).

By default those precise regular expressions aren't included in metadata at all because that would cause metadata to grow twice in its size (the complete metadata would be about 200 KiloBytes). If anyone needs to generate custom metadata then it's very easy to do so: just follow the instructions provided in the Customizing metadata section of this document (the option to look for is --extended).

class asYouType(default_country_code)

(aka as_you_type)

Creates a formatter for partially entered phone number. The two-letter default_country_code is optional and, if specified, is gonna be the default country for the phone number being input (in case it's not an international one). The instance of this class has two methods:

• input(text) — takes any text and appends it to the input; returns the formatted phone number
• reset() — resets the input

The instance of this class has also these fields:

• country — a country code of the country this phone belongs to
• country_phone_code — a phone code of the country
• national_number — national number part (so far)
• template — currently used phone number formatting template, where digits (and the plus sign, if present) are denoted by x-es

new asYouType().input('[phone]') === '[phone]'
new asYouType('US').input('2133734') === '(213) 373-4'

const formatter = new asYouType()
formatter.input('[phone]') === '[phone]'
formatter.country === 'US'
formatter.country_phone_code = '1'
formatter.template === 'xx xxx xxx xxxx'

getPhoneCode(country_code)

There have been requests for a function returning a phone code by country code.

getPhoneCode('RU') === '7'
getPhoneCode('IL') === '972'

Metadata

Metadata is generated from Google's original PhoneNumberMetadata.xml by transforming XML into JSON and removing unnecessary fields.

Currently I have an npm script for monitoring changes to PhoneNumberMetadata.xml in Google's repo and automatically creating a Pull Request in this repo with the fresh metadata when it is updated. What's left is to test this script and schedule it to run daily on my machine. So this project's metadata is supposed to be up-to-date. Still, in case the automatic metadata update script malfunctions some day, anyone can request metadata update via a Pull Request here on GitHub:

• Fork this repo
• npm install
• npm run metadata:update
• Submit a Pull Request to this repo from the update-metadata branch of your fork

npm run metadata:update command creates a new update-metadata branch, downloads the new PhoneNumberMetadata.xml into the project folder replacing the old one, generates JSON metadata out of the XML one, checks if the metadata has changed, runs the tests, commits the new metadata and pushes the commit to the remote update-metadata branch of your fork.

Alternatively, a developer may wish to update metadata urgently, without waiting for a pull request approval. In this case just perform the steps described in the Customizing metadata section of this document.

React

There's also a React component utilizing this library: react-phone-number-input

Bug reporting

If any inconsistencies with the original Google's libphonenumber are spot then an issue can be created in this repo. Validation bugs should be reported only if they appear after full metadata is generated because with the default metadata the validation of this library is looser than the original Google libphonenumber's.

Webpack

If you're using Webpack 1 (which you most likely are) then make sure that

• You have json-loader set up for *.json files in Webpack configuration (Webpack 2 has json-loader set up by default)
• json-loader doesn't exclude /node_modules/
• If you override resolve.extensions in Webpack configuration then make sure .json extension is present in the list

Webpack 2 sets up json-loader by default so there's no need for any special configuration. So better upgrade to Webpack 2 instead.

Standalone

For those who aren't using bundlers for some reason there's a way to build a standalone version of the library

• git clone https://github.com/halt-hammerzeit/libphonenumber-js.git
• npm install
• npm run browser-build
• See the bundle folder for libphonenumber-js.min.js

<script src="/scripts/libphonenumber-js.min.js"></script>
<script>
  alert(new libphonenumber.asYouType('US').input('213-373-4253'))
</script>

Customizing metadata

If only a specific set of countries is needed in a project, and a developer really wants to reduce the resulting bundle size, say, by 50 KiloBytes, then he can generate custom metadata and pass it as an extra argument to this library's functions. Or, say, if a developer wants to use the complete metadata (which is about 200 KiloBytes) for precise phone number validation then he can also generate such complete metadata set.

First, add metadata generation script to your project's package.json

{
  "scripts": {
    "libphonenumber-metadata": "libphonenumber-generate-metadata metadata.min.json --countries RU,DE --extended",
  }
}

And then run it like npm run libphonenumber-metadata.

The first argument is the output metadata file path. --countries argument is a comma-separated list of the required countries (if --countries is omitted then all countries are included). --extended argument may be passed to increase the precision of phone number validation but at the same time it will enlarge the resulting metadata size approximately twice.

Then use the generated metadata.min.json with the exported "custom" functions.

For a "tree-shaking" ES6-capable bundler (e.g. Webpack 2) that would be

import { parseCustom, formatCustom, isValidNumberCustom, asYouTypeCustom } from 'libphonenumber-js'
import metadata from './metadata.min.json'

export const parse = (...args) => parseCustom(...args, metadata)
export const format = (...args) => formatCustom(...args, metadata)
export const isValidNumber = (...args) => isValidNumberCustom(...args, metadata)

export class asYouType extends asYouTypeCustom {
  constructor(country) {
    super(country, metadata)
  }
}

And for Common.js environment that would be

var custom = require('libphonenumber-js/custom')
var metadata = require('./metadata.min.json')

exports.parse = function parse() {
  var parameters = Array.prototype.slice.call(arguments)
  parameters.push(metadata)
  return custom.parse.apply(this, parameters)
}

exports.format = function format() {
  var parameters = Array.prototype.slice.call(arguments)
  parameters.push(metadata)
  return custom.format.apply(this, parameters)
}

exports.isValidNumber = function isValidNumber() {
  var parameters = Array.prototype.slice.call(arguments)
  parameters.push(metadata)
  return custom.isValidNumber.apply(this, parameters)
}

exports.asYouType = function asYouType(country) {
  custom.asYouType.call(this, country, metadata)
}

exports.asYouType.prototype = Object.create(custom.asYouType.prototype, {})
exports.asYouType.prototype.constructor = exports.asYouType

ES6 "tree-shaking" is a non-trivial thing and at the moment of writing it's not guaranteed that a given ES6-aware bundler will actually be intelligent enough to tree-shake unused code, so there's always another option for those cases (if they arise): using libphonenumber-js/custom Common.js export.

import { parse as parseCustom } from 'libphonenumber-js/custom'
import metadata from './metadata.min.json'

export const parse = (...args) => parseCustom(...args, metadata)

Contributing

After cloning this repo, ensure dependencies are installed by running:

npm install

This module is written in ES6 and uses Babel for ES5 transpilation. Widely consumable JavaScript can be produced by running:

npm run build

Once npm run build has run, you may import or require() directly from node.

After developing, the full test suite can be evaluated by running:

npm test

When you're ready to test your new functionality on a real project, you can run

npm pack

It will build, test and then create a .tgz archive which you can then install in your project folder

npm install [module name with version].tar.gz

License

MIT