chardet

Chardet is a character detection module written in pure JavaScript (TypeScript). Module uses occurrence analysis to determine the most probable encoding.

• Packed size is only 22 KB
• Works in all environments: Node / Browser / Native
• Works on all platforms: Linux / Mac / Windows
• No dependencies
• No native code / bindings
• 100% written in TypeScript
• Extensive code coverage

Installation

npm i chardet

Usage

To return the encoding with the highest confidence:

import chardet from 'chardet';

const encoding = chardet.detect(Buffer.from('hello there!'));
// or
const encoding = await chardet.detectFile('/path/to/file');
// or
const encoding = chardet.detectFileSync('/path/to/file');

To return the full list of possible encodings use analyse method.

import chardet from 'chardet';
chardet.analyse(Buffer.from('hello there!'));

Returned value is an array of objects sorted by confidence value in descending order

[
  { confidence: 90, name: 'UTF-8' },
  { confidence: 20, name: 'windows-1252', lang: 'fr' },
];

In browser, you can use Uint8Array instead of the Buffer:

import chardet from 'chardet';
chardet.analyse(new Uint8Array([0x68, 0x65, 0x6c, 0x6c, 0x6f]));

Working with large data sets

Sometimes, when data set is huge and you want to optimize performance (with a trade off of less accuracy), you can sample only the first N bytes of the buffer:

const encoding = await chardet.detectFile('/path/to/file', { sampleSize: 32 });

You can also specify where to begin reading from in the buffer:

const encoding = await chardet.detectFile('/path/to/file', {
  sampleSize: 32,
  offset: 128,
});

Working with strings

In both Node.js and browsers, all strings in memory are represented in UTF-16 encoding. This is a fundamental aspect of the JavaScript language specification. Therefore, you cannot use plain strings directly as input for chardet.analyse() or chardet.detect(). Instead, you need the original string data in the form of a Buffer or Uint8Array.

In other words, if you receive a piece of data over the network and want to detect its encoding, use the original data payload, not its string representation. By the time you convert data to a string, it will be in UTF-16 encoding.

Note on TextEncoder: By default, it returns a UTF-8 encoded buffer, which means the buffer will not be in the original encoding of the string.

Supported Encodings:

• UTF-8
• UTF-16 LE
• UTF-16 BE
• UTF-32 LE
• UTF-32 BE
• ISO-2022-JP
• ISO-2022-KR
• ISO-2022-CN
• Shift_JIS
• Big5
• EUC-JP
• EUC-KR
• GB18030
• ISO-8859-1
• ISO-8859-2
• ISO-8859-5
• ISO-8859-6
• ISO-8859-7
• ISO-8859-8
• ISO-8859-9
• windows-1250
• windows-1251
• windows-1252
• windows-1253
• windows-1254
• windows-1255
• windows-1256
• KOI8-R

Currently only these encodings are supported.

TypeScript?

Yes. Type definitions are included.

References

• ICU project http://site.icu-project.org/