UNPKG

@gmod/cram

Version:

read CRAM files with pure Javascript

281 lines 12.4 kB
import QuickLRU from '@jbrowse/quick-lru'; import crc32 from 'crc/calculators/crc32'; import { CramMalformedError, CramUnimplementedError } from "../errors.js"; import * as htscodecs from "../htscodecs/index.js"; import { open } from "../io.js"; import { parseHeaderText } from "../sam.js"; import { decodeUtf8, parseItem } from "./util.js"; import { unzip } from "../unzip.js"; import CramContainer from "./container/index.js"; import { cramFileDefinition, getSectionParsers, } from "./sectionParsers.js"; import { xzDecompress } from "../xz-decompress/xz-decompress.js"; // source: https://abdulapopoola.com/2019/01/20/check-endianness-with-javascript/ let isLittleEndian; function checkLittleEndian() { if (isLittleEndian === undefined) { isLittleEndian = new Uint8Array(new Uint32Array([0x11223344]).buffer)[0] === 0x44; } return isLittleEndian; } export default class CramFile { file; validateChecksums; fetchReferenceSequenceCallback; options; featureCache; header; _sectionParsers; _definitionResult; _samHeaderResult; constructor(args) { this.file = open(args.url, args.path, args.filehandle); this.validateChecksums = args.validateChecksums ?? false; this.fetchReferenceSequenceCallback = args.seqFetch; this.options = { checkSequenceMD5: args.checkSequenceMD5, cacheSize: args.cacheSize ?? 20000, }; // cache of features in a slice, keyed by the slice offset. caches all of // the features in a slice, or none. the cache is actually used by the // slice object, it's just kept here at the level of the file this.featureCache = new QuickLRU({ maxSize: this.options.cacheSize, }); if (!checkLittleEndian()) { throw new Error('Detected big-endian machine, may be unable to run'); } } read(length, position) { return this.file.read(length, position); } async _getSectionParsers() { if (!this._sectionParsers) { const { majorVersion } = await this.getDefinition(); this._sectionParsers = getSectionParsers(majorVersion); } return this._sectionParsers; } async getDefinition() { if (this._definitionResult === undefined) { this._definitionResult = this._fetchDefinition(); this._definitionResult.catch(() => { this._definitionResult = undefined; }); } return this._definitionResult; } async _fetchDefinition() { const { maxLength, parser } = cramFileDefinition(); const headbytes = await this.file.read(maxLength, 0); const definition = parser(headbytes).value; if (definition.magic !== 'CRAM') { throw new Error('Not a CRAM file, does not match magic string'); } else if (definition.majorVersion !== 2 && definition.majorVersion !== 3) { throw new CramUnimplementedError(`CRAM version ${definition.majorVersion} not supported`); } else { return definition; } } async getSamHeader() { if (this._samHeaderResult === undefined) { this._samHeaderResult = this._fetchSamHeader(); this._samHeaderResult.catch(() => { this._samHeaderResult = undefined; }); } return this._samHeaderResult; } async _fetchSamHeader() { const firstContainer = await this.getContainerById(0); if (!firstContainer) { throw new CramMalformedError('file contains no containers'); } const firstBlock = await firstContainer.getFirstBlock(); const content = firstBlock.content; const dataView = new DataView(content.buffer, content.byteOffset, content.byteLength); const headerLength = dataView.getInt32(0, true); const textStart = 4; const text = decodeUtf8(content.subarray(textStart, textStart + headerLength)); this.header = text; return parseHeaderText(text); } async getHeaderText() { await this.getSamHeader(); return this.header; } // Walk containers from the start of the file. Yields each container along // with its parsed header. The first container's length is recomputed by // reading all of its blocks because the recorded length cannot be trusted // (htslib bug); subsequent containers use header._size + header.length. async *iterContainers() { const sectionParsers = await this._getSectionParsers(); let position = sectionParsers.cramFileDefinition.maxLength; let i = 0; // eslint-disable-next-line @typescript-eslint/no-unnecessary-condition while (true) { const container = this.getContainerAtPosition(position); const header = await container.getHeader(); yield container; if (i === 0) { position = header._endPosition; for (let j = 0; j < header.numBlocks; j++) { const block = await this.readBlock(position); position = block._endPosition; } } else { position += header._size + header.length; } i++; } } async getContainerById(containerNumber) { let i = 0; for await (const container of this.iterContainers()) { if (i === containerNumber) { return container; } i++; } return undefined; } async checkCrc32(position, length, recordedCrc32, description) { const b = await this.file.read(length, position); // this shift >>> 0 is equivalent to crc32(b).unsigned but uses the // internal calculator of crc32 to avoid accidentally importing buffer // https://github.com/alexgorbatchev/crc/blob/31fc3853e417b5fb5ec83335428805842575f699/src/define_crc.ts#L5 const calculatedCrc32 = crc32(b) >>> 0; if (calculatedCrc32 !== recordedCrc32) { throw new CramMalformedError(`crc mismatch in ${description}: recorded CRC32 = ${recordedCrc32}, but calculated CRC32 = ${calculatedCrc32}`); } } /** * @returns {Promise[number]} the number of containers in the file * * note: this is currently used only in unit tests, and after removing file * length check, relies on a try catch to read return an error to break */ async containerCount() { let containerCount = 0; try { // eslint-disable-next-line @typescript-eslint/no-unused-vars for await (const _container of this.iterContainers()) { containerCount += 1; } } catch (e) { containerCount--; } return containerCount; } getContainerAtPosition(position) { return new CramContainer(this, position); } async readBlockHeader(position) { const { cramBlockHeader } = await this._getSectionParsers(); const buffer = await this.file.read(cramBlockHeader.maxLength, position); return parseItem(buffer, cramBlockHeader.parser, 0, position); } async _parseSection(section, position, size = section.maxLength, preReadBuffer) { const buffer = preReadBuffer ?? (await this.file.read(size, position)); const data = parseItem(buffer, section.parser, 0, position); if (data._size !== size) { throw new CramMalformedError(`section read error: requested size ${size} does not equal parsed size ${data._size}`); } return data; } async _uncompress(compressionMethod, inputBuffer, uncompressedSize) { let buf; if (compressionMethod === 'gzip') { buf = await unzip(inputBuffer); } else if (compressionMethod === 'bzip2') { buf = await htscodecs.bz2_uncompress(inputBuffer, uncompressedSize); } else if (compressionMethod === 'lzma') { buf = await xzDecompress(inputBuffer); } else if (compressionMethod === 'rans') { buf = await htscodecs.rans_uncompress(inputBuffer); } else if (compressionMethod === 'rans4x16') { buf = await htscodecs.r4x16_uncompress(inputBuffer); } else if (compressionMethod === 'arith') { buf = await htscodecs.arith_uncompress(inputBuffer); } else if (compressionMethod === 'fqzcomp') { buf = await htscodecs.fqzcomp_uncompress(inputBuffer); } else if (compressionMethod === 'tok3') { buf = await htscodecs.tok3_uncompress(inputBuffer); } else { throw new CramUnimplementedError(`${compressionMethod} decompression not yet implemented`); } if (buf.length !== uncompressedSize) { throw new CramMalformedError(`${compressionMethod} decompression produced ${buf.length} bytes, expected ${uncompressedSize}`); } return buf; } async readBlock(position) { const { majorVersion } = await this.getDefinition(); const { cramBlockHeader, cramBlockCrc32 } = await this._getSectionParsers(); const headerBuf = await this.file.read(cramBlockHeader.maxLength, position); const blockHeader = parseItem(headerBuf, cramBlockHeader.parser, 0, position); const totalSize = blockHeader._size + blockHeader.compressedSize + (majorVersion >= 3 ? cramBlockCrc32.maxLength : 0); const fullBuffer = await this.file.read(totalSize, position); return this.readBlockFromBuffer(fullBuffer, 0, position); } async readBlockFromBuffer(buffer, bufferOffset, filePosition) { const { majorVersion } = await this.getDefinition(); const sectionParsers = await this._getSectionParsers(); const { cramBlockHeader } = sectionParsers; const headerBytes = buffer.subarray(bufferOffset, bufferOffset + cramBlockHeader.maxLength); const blockHeader = parseItem(headerBytes, cramBlockHeader.parser, 0, filePosition); const blockContentPosition = blockHeader._endPosition; const contentOffset = bufferOffset + blockHeader._size; const d = buffer.subarray(contentOffset, contentOffset + blockHeader.compressedSize); // Per CRAM spec (PR #681), blocks with uncompressed size 0 are treated as // empty regardless of the method byte — htsjdk has produced invalid empty // RANS blocks that would otherwise fail to decompress. const uncompressedData = blockHeader.uncompressedSize === 0 ? new Uint8Array(0) : blockHeader.compressionMethod !== 'raw' ? await this._uncompress(blockHeader.compressionMethod, d, blockHeader.uncompressedSize) : d; const block = { ...blockHeader, _endPosition: blockContentPosition, contentPosition: blockContentPosition, content: uncompressedData, }; if (majorVersion >= 3) { const crcOffset = contentOffset + blockHeader.compressedSize; const crcBytes = buffer.subarray(crcOffset, crcOffset + sectionParsers.cramBlockCrc32.maxLength); const crc = parseItem(crcBytes, sectionParsers.cramBlockCrc32.parser, 0, blockContentPosition + blockHeader.compressedSize); block.crc32 = crc.crc32; if (this.validateChecksums) { const blockData = buffer.subarray(bufferOffset, bufferOffset + blockHeader._size + blockHeader.compressedSize); const calculatedCrc32 = crc32(blockData) >>> 0; if (calculatedCrc32 !== crc.crc32) { throw new CramMalformedError(`crc mismatch in block data: recorded CRC32 = ${crc.crc32}, but calculated CRC32 = ${calculatedCrc32}`); } } block._endPosition = crc._endPosition; block._size = block.compressedSize + sectionParsers.cramBlockCrc32.maxLength; } else { block._endPosition = blockContentPosition + block.compressedSize; block._size = block.compressedSize; } return block; } } //# sourceMappingURL=file.js.map