Awesome

melanite

Convert a user-agent to a normalised device.

Installation

npm install --save melanite

Usage

Getting started

In order to use melanite, you need to provide one or more "matchers"; a matcher represents a device you wish to identify. Below is an example matcher that could be used to identify a Microsoft Xbox One device:

{
  "brand": "microsoft",
  "model": "xbox-one",
  "invariants": [
    "Xbox One"
  ],
  "disallowed": [],
  "fuzzy": "Mozilla/5.0(Windows NT 10.0; Win64; x64; Xbox; Xbox One) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.2743.116 Safari/537.36 Edge/15.15063",
  "type": "tv"
}

Let's take a look at each component of a matcher (all of which are mandatory).

brand

The brand of the device; it serves no purpose for identification and is merely a friendly name for a group of several different devices (e.g. by manufacturer)

model

The model of the device; like brand, it serves no purpose for identification and is instead a friendly name for a specific device.

invariants

invariants is an array of strings (each of these is referred to as an invariant); in order for a user agent to be matched to this matcher, it must contain every invariant (similar to an allowlist).

disallowed

disallowed is an array of strings, it is in the opposite of invariants. In order for a user agent to be matched to this matcher, it must not contain any of these strings (similar to a denylist).

In the above example, we haven't specified any disallowed items. The disallowed property is most useful when you have two or more matchers that are very similar to each other.

fuzzy

fuzzy can be thought of as an example user agent. When a group of matchers have been filtered using invariants and disallowed, melanite calculates the Levenshtein distance between the fuzzy and the user-agent - the matcher with the lowest Levenshtein distance is the one returned by melanite.

type

The type of the device; like brand and model, it serves no purpose for identification.

Identifying devices

Now that we have a matcher, let's use it to identify some user agents:

const melanite = require('melanite')

const userAgents = [
  'Mozilla/5.0(Windows NT 10.0; Win64; x64; Xbox; Xbox One) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.713.12 Safari/57.36 Edge/15.4063',
  'Some very strange user agent that we do not know about'
]

const matchers = [
  {
    "brand": "microsoft",
    "model": "xbox-one",
    "invariants": [
	  "Xbox One"
    ],
    "disallowed": [],
    "fuzzy": "Mozilla/5.0(Windows NT 10.0; Win64; x64; Xbox; Xbox One) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.2743.116 Safari/537.36 Edge/15.15063",
    "type": "tv"
  }
]

const identifyDevice = melanite.match(matchers)

const devices = userAgents.map((userAgent) => identifyDevice(userAgent))

console.log(devices)
/*
[
  { brand: 'microsoft', model: 'xbox-one', type: 'tv' },
  { brand: 'generic', model: 'device', type: 'unknown' }
]
*/

For a further example, please see example.js.

Snapshot data

We have provided a snasphot of the BBC's devices in production in the /data directory, if you wish to use these matchers your self.