vscode/src/vs/workbench/services/textfile/common/encoding.ts

/*---------------------------------------------------------------------------------------------
 *  Copyright (c) Microsoft Corporation. All rights reserved.
 *  Licensed under the MIT License. See License.txt in the project root for license information.
 *--------------------------------------------------------------------------------------------*/

import { Readable, ReadableStream, newWriteableStream, listenStream } from 'vs/base/common/stream';
import { VSBuffer, VSBufferReadable, VSBufferReadableStream } from 'vs/base/common/buffer';

export const UTF8 = 'utf8';
export const UTF8_with_bom = 'utf8bom';
export const UTF16be = 'utf16be';
export const UTF16le = 'utf16le';

export type UTF_ENCODING = typeof UTF8 | typeof UTF8_with_bom | typeof UTF16be | typeof UTF16le;

export function isUTFEncoding(encoding: string): encoding is UTF_ENCODING {
	return [UTF8, UTF8_with_bom, UTF16be, UTF16le].some(utfEncoding => utfEncoding === encoding);
}

export const UTF16be_BOM = [0xFE, 0xFF];
export const UTF16le_BOM = [0xFF, 0xFE];
export const UTF8_BOM = [0xEF, 0xBB, 0xBF];

const ZERO_BYTE_DETECTION_BUFFER_MAX_LEN = 512; 	// number of bytes to look at to decide about a file being binary or not
const NO_ENCODING_GUESS_MIN_BYTES = 512; 			// when not auto guessing the encoding, small number of bytes are enough
const AUTO_ENCODING_GUESS_MIN_BYTES = 512 * 8; 		// with auto guessing we want a lot more content to be read for guessing
const AUTO_ENCODING_GUESS_MAX_BYTES = 512 * 128; 	// set an upper limit for the number of bytes we pass on to jschardet

export interface IDecodeStreamOptions {
	guessEncoding: boolean;
	minBytesRequiredForDetection?: number;

	overwriteEncoding(detectedEncoding: string | null): Promise<string>;
}

export interface IDecodeStreamResult {
	stream: ReadableStream<string>;
	detected: IDetectedEncodingResult;
}

export interface IDecoderStream {
	write(buffer: Uint8Array): string;
	end(): string | undefined;
}

class DecoderStream implements IDecoderStream {

	/**
	 * This stream will only load iconv-lite lazily if the encoding
	 * is not UTF-8. This ensures that for most common cases we do
	 * not pay the price of loading the module from disk.
	 *
	 * We still need to be careful when converting UTF-8 to a string
	 * though because we read the file in chunks of Buffer and thus
	 * need to decode it via TextDecoder helper that is available
	 * in browser and node.js environments.
	 */
	static async create(encoding: string): Promise<DecoderStream> {
		let decoder: IDecoderStream | undefined = undefined;
		if (encoding !== UTF8) {
			const iconv = await import('iconv-lite-umd');
			decoder = iconv.getDecoder(toNodeEncoding(encoding));
		} else {
			const utf8TextDecoder = new TextDecoder();
			decoder = {
				write(buffer: Uint8Array): string {
					return utf8TextDecoder.decode(buffer, {
						// Signal to TextDecoder that potentially more data is coming
						// and that we are calling `decode` in the end to consume any
						// remainders
						stream: true
					});
				},

				end(): string | undefined {
					return utf8TextDecoder.decode();
				}
			};
		}

		return new DecoderStream(decoder);
	}

	private constructor(private iconvLiteDecoder: IDecoderStream) { }

	write(buffer: Uint8Array): string {
		return this.iconvLiteDecoder.write(buffer);
	}

	end(): string | undefined {
		return this.iconvLiteDecoder.end();
	}
}

export function toDecodeStream(source: VSBufferReadableStream, options: IDecodeStreamOptions): Promise<IDecodeStreamResult> {
	const minBytesRequiredForDetection = options.minBytesRequiredForDetection ?? options.guessEncoding ? AUTO_ENCODING_GUESS_MIN_BYTES : NO_ENCODING_GUESS_MIN_BYTES;

	return new Promise<IDecodeStreamResult>((resolve, reject) => {
		const target = newWriteableStream<string>(strings => strings.join(''));

		const bufferedChunks: VSBuffer[] = [];
		let bytesBuffered = 0;

		let decoder: IDecoderStream | undefined = undefined;

		const createDecoder = async () => {
			try {

				// detect encoding from buffer
				const detected = await detectEncodingFromBuffer({
					buffer: VSBuffer.concat(bufferedChunks),
					bytesRead: bytesBuffered
				}, options.guessEncoding);

				// ensure to respect overwrite of encoding
				detected.encoding = await options.overwriteEncoding(detected.encoding);

				// decode and write buffered content
				decoder = await DecoderStream.create(detected.encoding);
				const decoded = decoder.write(VSBuffer.concat(bufferedChunks).buffer);
				target.write(decoded);

				bufferedChunks.length = 0;
				bytesBuffered = 0;

				// signal to the outside our detected encoding and final decoder stream
				resolve({
					stream: target,
					detected
				});
			} catch (error) {
				reject(error);
			}
		};

		listenStream(source, {
			onData: async chunk => {

				// if the decoder is ready, we just write directly
				if (decoder) {
					target.write(decoder.write(chunk.buffer));
				}

				// otherwise we need to buffer the data until the stream is ready
				else {
					bufferedChunks.push(chunk);
					bytesBuffered += chunk.byteLength;

					// buffered enough data for encoding detection, create stream
					if (bytesBuffered >= minBytesRequiredForDetection) {

						// pause stream here until the decoder is ready
						source.pause();

						await createDecoder();

						// resume stream now that decoder is ready but
						// outside of this stack to reduce recursion
						setTimeout(() => source.resume());
					}
				}
			},
			onError: error => target.error(error), // simply forward to target
			onEnd: async () => {

				// we were still waiting for data to do the encoding
				// detection. thus, wrap up starting the stream even
				// without all the data to get things going
				if (!decoder) {
					await createDecoder();
				}

				// end the target with the remainders of the decoder
				target.end(decoder?.end());
			}
		});
	});
}

export async function toEncodeReadable(readable: Readable<string>, encoding: string, options?: { addBOM?: boolean }): Promise<VSBufferReadable> {
	const iconv = await import('iconv-lite-umd');
	const encoder = iconv.getEncoder(toNodeEncoding(encoding), options);

	let bytesWritten = false;
	let done = false;

	return {
		read() {
			if (done) {
				return null;
			}

			const chunk = readable.read();
			if (typeof chunk !== 'string') {
				done = true;

				// If we are instructed to add a BOM but we detect that no
				// bytes have been written, we must ensure to return the BOM
				// ourselves so that we comply with the contract.
				if (!bytesWritten && options?.addBOM) {
					switch (encoding) {
						case UTF8:
						case UTF8_with_bom:
							return VSBuffer.wrap(Uint8Array.from(UTF8_BOM));
						case UTF16be:
							return VSBuffer.wrap(Uint8Array.from(UTF16be_BOM));
						case UTF16le:
							return VSBuffer.wrap(Uint8Array.from(UTF16le_BOM));
					}
				}

				const leftovers = encoder.end();
				if (leftovers && leftovers.length > 0) {
					bytesWritten = true;

					return VSBuffer.wrap(leftovers);
				}

				return null;
			}

			bytesWritten = true;

			return VSBuffer.wrap(encoder.write(chunk));
		}
	};
}

export async function encodingExists(encoding: string): Promise<boolean> {
	const iconv = await import('iconv-lite-umd');

	return iconv.encodingExists(toNodeEncoding(encoding));
}

export function toNodeEncoding(enc: string | null): string {
	if (enc === UTF8_with_bom || enc === null) {
		return UTF8; // iconv does not distinguish UTF 8 with or without BOM, so we need to help it
	}

	return enc;
}

export function detectEncodingByBOMFromBuffer(buffer: VSBuffer | null, bytesRead: number): typeof UTF8_with_bom | typeof UTF16le | typeof UTF16be | null {
	if (!buffer || bytesRead < UTF16be_BOM.length) {
		return null;
	}

	const b0 = buffer.readUInt8(0);
	const b1 = buffer.readUInt8(1);

	// UTF-16 BE
	if (b0 === UTF16be_BOM[0] && b1 === UTF16be_BOM[1]) {
		return UTF16be;
	}

	// UTF-16 LE
	if (b0 === UTF16le_BOM[0] && b1 === UTF16le_BOM[1]) {
		return UTF16le;
	}

	if (bytesRead < UTF8_BOM.length) {
		return null;
	}

	const b2 = buffer.readUInt8(2);

	// UTF-8
	if (b0 === UTF8_BOM[0] && b1 === UTF8_BOM[1] && b2 === UTF8_BOM[2]) {
		return UTF8_with_bom;
	}

	return null;
}

// we explicitly ignore a specific set of encodings from auto guessing
// - ASCII: we never want this encoding (most UTF-8 files would happily detect as
//          ASCII files and then you could not type non-ASCII characters anymore)
// - UTF-16: we have our own detection logic for UTF-16
// - UTF-32: we do not support this encoding in VSCode
const IGNORE_ENCODINGS = ['ascii', 'utf-16', 'utf-32'];

/**
 * Guesses the encoding from buffer.
 */
async function guessEncodingByBuffer(buffer: VSBuffer): Promise<string | null> {
	const jschardet = await import('jschardet');

	// ensure to limit buffer for guessing due to https://github.com/aadsm/jschardet/issues/53
	const limitedBuffer = buffer.slice(0, AUTO_ENCODING_GUESS_MAX_BYTES);

	// before guessing jschardet calls toString('binary') on input if it is a Buffer,
	// since we are using it inside browser environment as well we do conversion ourselves
	// https://github.com/aadsm/jschardet/blob/v2.1.1/src/index.js#L36-L40
	const binaryString = encodeLatin1(limitedBuffer.buffer);

	const guessed = jschardet.detect(binaryString);
	if (!guessed || !guessed.encoding) {
		return null;
	}

	const enc = guessed.encoding.toLowerCase();
	if (0 <= IGNORE_ENCODINGS.indexOf(enc)) {
		return null; // see comment above why we ignore some encodings
	}

	return toIconvLiteEncoding(guessed.encoding);
}

const JSCHARDET_TO_ICONV_ENCODINGS: { [name: string]: string } = {
	'ibm866': 'cp866',
	'big5': 'cp950'
};

function toIconvLiteEncoding(encodingName: string): string {
	const normalizedEncodingName = encodingName.replace(/[^a-zA-Z0-9]/g, '').toLowerCase();
	const mapped = JSCHARDET_TO_ICONV_ENCODINGS[normalizedEncodingName];

	return mapped || normalizedEncodingName;
}

function encodeLatin1(buffer: Uint8Array): string {
	let result = '';
	for (let i = 0; i < buffer.length; i++) {
		result += String.fromCharCode(buffer[i]);
	}

	return result;
}

/**
 * The encodings that are allowed in a settings file don't match the canonical encoding labels specified by WHATWG.
 * See https://encoding.spec.whatwg.org/#names-and-labels
 * Iconv-lite strips all non-alphanumeric characters, but ripgrep doesn't. For backcompat, allow these labels.
 */
export function toCanonicalName(enc: string): string {
	switch (enc) {
		case 'shiftjis':
			return 'shift-jis';
		case 'utf16le':
			return 'utf-16le';
		case 'utf16be':
			return 'utf-16be';
		case 'big5hkscs':
			return 'big5-hkscs';
		case 'eucjp':
			return 'euc-jp';
		case 'euckr':
			return 'euc-kr';
		case 'koi8r':
			return 'koi8-r';
		case 'koi8u':
			return 'koi8-u';
		case 'macroman':
			return 'x-mac-roman';
		case 'utf8bom':
			return 'utf8';
		default:
			const m = enc.match(/windows(\d+)/);
			if (m) {
				return 'windows-' + m[1];
			}

			return enc;
	}
}

export interface IDetectedEncodingResult {
	encoding: string | null;
	seemsBinary: boolean;
}

export interface IReadResult {
	buffer: VSBuffer | null;
	bytesRead: number;
}

export function detectEncodingFromBuffer(readResult: IReadResult, autoGuessEncoding?: false): IDetectedEncodingResult;
export function detectEncodingFromBuffer(readResult: IReadResult, autoGuessEncoding?: boolean): Promise<IDetectedEncodingResult>;
export function detectEncodingFromBuffer({ buffer, bytesRead }: IReadResult, autoGuessEncoding?: boolean): Promise<IDetectedEncodingResult> | IDetectedEncodingResult {

	// Always first check for BOM to find out about encoding
	let encoding = detectEncodingByBOMFromBuffer(buffer, bytesRead);

	// Detect 0 bytes to see if file is binary or UTF-16 LE/BE
	// unless we already know that this file has a UTF-16 encoding
	let seemsBinary = false;
	if (encoding !== UTF16be && encoding !== UTF16le && buffer) {
		let couldBeUTF16LE = true; // e.g. 0xAA 0x00
		let couldBeUTF16BE = true; // e.g. 0x00 0xAA
		let containsZeroByte = false;

		// This is a simplified guess to detect UTF-16 BE or LE by just checking if
		// the first 512 bytes have the 0-byte at a specific location. For UTF-16 LE
		// this would be the odd byte index and for UTF-16 BE the even one.
		// Note: this can produce false positives (a binary file that uses a 2-byte
		// encoding of the same format as UTF-16) and false negatives (a UTF-16 file
		// that is using 4 bytes to encode a character).
		for (let i = 0; i < bytesRead && i < ZERO_BYTE_DETECTION_BUFFER_MAX_LEN; i++) {
			const isEndian = (i % 2 === 1); // assume 2-byte sequences typical for UTF-16
			const isZeroByte = (buffer.readUInt8(i) === 0);

			if (isZeroByte) {
				containsZeroByte = true;
			}

			// UTF-16 LE: expect e.g. 0xAA 0x00
			if (couldBeUTF16LE && (isEndian && !isZeroByte || !isEndian && isZeroByte)) {
				couldBeUTF16LE = false;
			}

			// UTF-16 BE: expect e.g. 0x00 0xAA
			if (couldBeUTF16BE && (isEndian && isZeroByte || !isEndian && !isZeroByte)) {
				couldBeUTF16BE = false;
			}

			// Return if this is neither UTF16-LE nor UTF16-BE and thus treat as binary
			if (isZeroByte && !couldBeUTF16LE && !couldBeUTF16BE) {
				break;
			}
		}

		// Handle case of 0-byte included
		if (containsZeroByte) {
			if (couldBeUTF16LE) {
				encoding = UTF16le;
			} else if (couldBeUTF16BE) {
				encoding = UTF16be;
			} else {
				seemsBinary = true;
			}
		}
	}

	// Auto guess encoding if configured
	if (autoGuessEncoding && !seemsBinary && !encoding && buffer) {
		return guessEncodingByBuffer(buffer.slice(0, bytesRead)).then(guessedEncoding => {
			return {
				seemsBinary: false,
				encoding: guessedEncoding
			};
		});
	}

	return { seemsBinary, encoding };
}

export const SUPPORTED_ENCODINGS: { [encoding: string]: { labelLong: string; labelShort: string; order: number; encodeOnly?: boolean; alias?: string } } = {
	utf8: {
		labelLong: 'UTF-8',
		labelShort: 'UTF-8',
		order: 1,
		alias: 'utf8bom'
	},
	utf8bom: {
		labelLong: 'UTF-8 with BOM',
		labelShort: 'UTF-8 with BOM',
		encodeOnly: true,
		order: 2,
		alias: 'utf8'
	},
	utf16le: {
		labelLong: 'UTF-16 LE',
		labelShort: 'UTF-16 LE',
		order: 3
	},
	utf16be: {
		labelLong: 'UTF-16 BE',
		labelShort: 'UTF-16 BE',
		order: 4
	},
	windows1252: {
		labelLong: 'Western (Windows 1252)',
		labelShort: 'Windows 1252',
		order: 5
	},
	iso88591: {
		labelLong: 'Western (ISO 8859-1)',
		labelShort: 'ISO 8859-1',
		order: 6
	},
	iso88593: {
		labelLong: 'Western (ISO 8859-3)',
		labelShort: 'ISO 8859-3',
		order: 7
	},
	iso885915: {
		labelLong: 'Western (ISO 8859-15)',
		labelShort: 'ISO 8859-15',
		order: 8
	},
	macroman: {
		labelLong: 'Western (Mac Roman)',
		labelShort: 'Mac Roman',
		order: 9
	},
	cp437: {
		labelLong: 'DOS (CP 437)',
		labelShort: 'CP437',
		order: 10
	},
	windows1256: {
		labelLong: 'Arabic (Windows 1256)',
		labelShort: 'Windows 1256',
		order: 11
	},
	iso88596: {
		labelLong: 'Arabic (ISO 8859-6)',
		labelShort: 'ISO 8859-6',
		order: 12
	},
	windows1257: {
		labelLong: 'Baltic (Windows 1257)',
		labelShort: 'Windows 1257',
		order: 13
	},
	iso88594: {
		labelLong: 'Baltic (ISO 8859-4)',
		labelShort: 'ISO 8859-4',
		order: 14
	},
	iso885914: {
		labelLong: 'Celtic (ISO 8859-14)',
		labelShort: 'ISO 8859-14',
		order: 15
	},
	windows1250: {
		labelLong: 'Central European (Windows 1250)',
		labelShort: 'Windows 1250',
		order: 16
	},
	iso88592: {
		labelLong: 'Central European (ISO 8859-2)',
		labelShort: 'ISO 8859-2',
		order: 17
	},
	cp852: {
		labelLong: 'Central European (CP 852)',
		labelShort: 'CP 852',
		order: 18
	},
	windows1251: {
		labelLong: 'Cyrillic (Windows 1251)',
		labelShort: 'Windows 1251',
		order: 19
	},
	cp866: {
		labelLong: 'Cyrillic (CP 866)',
		labelShort: 'CP 866',
		order: 20
	},
	iso88595: {
		labelLong: 'Cyrillic (ISO 8859-5)',
		labelShort: 'ISO 8859-5',
		order: 21
	},
	koi8r: {
		labelLong: 'Cyrillic (KOI8-R)',
		labelShort: 'KOI8-R',
		order: 22
	},
	koi8u: {
		labelLong: 'Cyrillic (KOI8-U)',
		labelShort: 'KOI8-U',
		order: 23
	},
	iso885913: {
		labelLong: 'Estonian (ISO 8859-13)',
		labelShort: 'ISO 8859-13',
		order: 24
	},
	windows1253: {
		labelLong: 'Greek (Windows 1253)',
		labelShort: 'Windows 1253',
		order: 25
	},
	iso88597: {
		labelLong: 'Greek (ISO 8859-7)',
		labelShort: 'ISO 8859-7',
		order: 26
	},
	windows1255: {
		labelLong: 'Hebrew (Windows 1255)',
		labelShort: 'Windows 1255',
		order: 27
	},
	iso88598: {
		labelLong: 'Hebrew (ISO 8859-8)',
		labelShort: 'ISO 8859-8',
		order: 28
	},
	iso885910: {
		labelLong: 'Nordic (ISO 8859-10)',
		labelShort: 'ISO 8859-10',
		order: 29
	},
	iso885916: {
		labelLong: 'Romanian (ISO 8859-16)',
		labelShort: 'ISO 8859-16',
		order: 30
	},
	windows1254: {
		labelLong: 'Turkish (Windows 1254)',
		labelShort: 'Windows 1254',
		order: 31
	},
	iso88599: {
		labelLong: 'Turkish (ISO 8859-9)',
		labelShort: 'ISO 8859-9',
		order: 32
	},
	windows1258: {
		labelLong: 'Vietnamese (Windows 1258)',
		labelShort: 'Windows 1258',
		order: 33
	},
	gbk: {
		labelLong: 'Simplified Chinese (GBK)',
		labelShort: 'GBK',
		order: 34
	},
	gb18030: {
		labelLong: 'Simplified Chinese (GB18030)',
		labelShort: 'GB18030',
		order: 35
	},
	cp950: {
		labelLong: 'Traditional Chinese (Big5)',
		labelShort: 'Big5',
		order: 36
	},
	big5hkscs: {
		labelLong: 'Traditional Chinese (Big5-HKSCS)',
		labelShort: 'Big5-HKSCS',
		order: 37
	},
	shiftjis: {
		labelLong: 'Japanese (Shift JIS)',
		labelShort: 'Shift JIS',
		order: 38
	},
	eucjp: {
		labelLong: 'Japanese (EUC-JP)',
		labelShort: 'EUC-JP',
		order: 39
	},
	euckr: {
		labelLong: 'Korean (EUC-KR)',
		labelShort: 'EUC-KR',
		order: 40
	},
	windows874: {
		labelLong: 'Thai (Windows 874)',
		labelShort: 'Windows 874',
		order: 41
	},
	iso885911: {
		labelLong: 'Latin/Thai (ISO 8859-11)',
		labelShort: 'ISO 8859-11',
		order: 42
	},
	koi8ru: {
		labelLong: 'Cyrillic (KOI8-RU)',
		labelShort: 'KOI8-RU',
		order: 43
	},
	koi8t: {
		labelLong: 'Tajik (KOI8-T)',
		labelShort: 'KOI8-T',
		order: 44
	},
	gb2312: {
		labelLong: 'Simplified Chinese (GB 2312)',
		labelShort: 'GB 2312',
		order: 45
	},
	cp865: {
		labelLong: 'Nordic DOS (CP 865)',
		labelShort: 'CP 865',
		order: 46
	},
	cp850: {
		labelLong: 'Western European DOS (CP 850)',
		labelShort: 'CP 850',
		order: 47
	}
};