feat: refactor folder structure & add new codec parser and gen & add unit tests

packages/codecs/package.json

@@ -1,19 +0,0 @@
-{
-    "name": "konoplayer-codecs",
-    "version": "0.1.0",
-    "private": true,
-    "type": "module",
-    "scripts": {
-        "build": "rsbuild build",
-        "dev": "rsbuild dev",
-        "preview": "rsbuild preview"
-    },
-    "dependencies": {
-        "konoebml": "0.1.2-rc.5",
-        "lit": "^3.2.1"
-    },
-    "devDependencies": {
-        "@rsbuild/core": "^1.2.14",
-        "typescript": "^5.8.2"
-    }
-}

packages/core/package.json

@@ -0,0 +1,8 @@
+{
+    "name": "@konoplayer/core",
+    "version": "0.1.0",
+    "private": true,
+    "type": "module",
+    "scripts": {},
+    "dependencies": {}
+}

packages/core/src/codecs/audio-codecs.ts

@@ -0,0 +1,32 @@
+export enum AudioCodec {
+  Unknown = 0,
+  AAC = 1,
+  MP3 = 2,
+  PCM = 3,
+  Vorbis = 4,
+  FLAC = 5,
+  AMR_NB = 6,
+  AMR_WB = 7,
+  PCM_MULAW = 8,
+  GSM_MS = 9,
+  PCM_S16BE = 10,
+  PCM_S24BE = 11,
+  Opus = 12,
+  EAC3 = 13,
+  PCM_ALAW = 14,
+  ALAC = 15,
+  AC3 = 16,
+  MpegHAudio = 17,
+  DTS = 18,
+  DTSXP2 = 19,
+  DTSE = 20,
+  AC4 = 21,
+  IAMF = 22,
+  PCM_S32BE = 23,
+  PCM_S32LE = 24,
+  PCM_S24LE = 25,
+  PCM_S16LE = 26,
+  PCM_F32BE = 27,
+  PCM_F32LE = 28,
+  MaxValue = PCM_F32LE, // Must equal the last "real" codec above.
+}

packages/core/src/codecs/index.ts

@@ -0,0 +1,2 @@
+export { AudioCodec } from './audio-codecs';
+export { VideoCodec } from './video-codecs';

packages/core/src/codecs/video-codecs.ts

@@ -0,0 +1,97 @@
+export enum VideoCodec {
+  Unknown = 0,
+  H264 = 1,
+  VC1 = 2,
+  MPEG2 = 3,
+  MPEG4 = 4,
+  Theora = 5,
+  VP8 = 6,
+  VP9 = 7,
+  HEVC = 8,
+  DolbyVision = 9,
+  AV1 = 10,
+  MaxValue = AV1, // Must equal the last "real" codec above.
+}
+
+export enum VideoCodecProfile {
+  VIDEO_CODEC_PROFILE_UNKNOWN = -1,
+  VIDEO_CODEC_PROFILE_MIN = VIDEO_CODEC_PROFILE_UNKNOWN,
+  H264PROFILE_MIN = 0,
+  H264PROFILE_BASELINE = H264PROFILE_MIN,
+  H264PROFILE_MAIN = 1,
+  H264PROFILE_EXTENDED = 2,
+  H264PROFILE_HIGH = 3,
+  H264PROFILE_HIGH10PROFILE = 4,
+  H264PROFILE_HIGH422PROFILE = 5,
+  H264PROFILE_HIGH444PREDICTIVEPROFILE = 6,
+  H264PROFILE_SCALABLEBASELINE = 7,
+  H264PROFILE_SCALABLEHIGH = 8,
+  H264PROFILE_STEREOHIGH = 9,
+  H264PROFILE_MULTIVIEWHIGH = 10,
+  H264PROFILE_MAX = H264PROFILE_MULTIVIEWHIGH,
+  VP8PROFILE_MIN = 11,
+  VP8PROFILE_ANY = VP8PROFILE_MIN,
+  VP8PROFILE_MAX = VP8PROFILE_ANY,
+  VP9PROFILE_MIN = 12,
+  VP9PROFILE_PROFILE0 = VP9PROFILE_MIN,
+  VP9PROFILE_PROFILE1 = 13,
+  VP9PROFILE_PROFILE2 = 14,
+  VP9PROFILE_PROFILE3 = 15,
+  VP9PROFILE_MAX = VP9PROFILE_PROFILE3,
+  HEVCPROFILE_MIN = 16,
+  HEVCPROFILE_MAIN = HEVCPROFILE_MIN,
+  HEVCPROFILE_MAIN10 = 17,
+  HEVCPROFILE_MAIN_STILL_PICTURE = 18,
+  HEVCPROFILE_MAX = HEVCPROFILE_MAIN_STILL_PICTURE,
+  DOLBYVISION_PROFILE0 = 19,
+  // Deprecated: DOLBYVISION_PROFILE4 = 20,
+  DOLBYVISION_PROFILE5 = 21,
+  DOLBYVISION_PROFILE7 = 22,
+  THEORAPROFILE_MIN = 23,
+  THEORAPROFILE_ANY = THEORAPROFILE_MIN,
+  THEORAPROFILE_MAX = THEORAPROFILE_ANY,
+  AV1PROFILE_MIN = 24,
+  AV1PROFILE_PROFILE_MAIN = AV1PROFILE_MIN,
+  AV1PROFILE_PROFILE_HIGH = 25,
+  AV1PROFILE_PROFILE_PRO = 26,
+  AV1PROFILE_MAX = AV1PROFILE_PROFILE_PRO,
+  DOLBYVISION_PROFILE8 = 27,
+  DOLBYVISION_PROFILE9 = 28,
+  HEVCPROFILE_EXT_MIN = 29,
+  HEVCPROFILE_REXT = HEVCPROFILE_EXT_MIN,
+  HEVCPROFILE_HIGH_THROUGHPUT = 30,
+  HEVCPROFILE_MULTIVIEW_MAIN = 31,
+  HEVCPROFILE_SCALABLE_MAIN = 32,
+  HEVCPROFILE_3D_MAIN = 33,
+  HEVCPROFILE_SCREEN_EXTENDED = 34,
+  HEVCPROFILE_SCALABLE_REXT = 35,
+  HEVCPROFILE_HIGH_THROUGHPUT_SCREEN_EXTENDED = 36,
+  HEVCPROFILE_EXT_MAX = HEVCPROFILE_HIGH_THROUGHPUT_SCREEN_EXTENDED,
+  VVCPROFILE_MIN = 37,
+  VVCPROFILE_MAIN10 = VVCPROFILE_MIN,
+  VVCPROFILE_MAIN12 = 38,
+  VVCPROFILE_MAIN12_INTRA = 39,
+  VVCPROIFLE_MULTILAYER_MAIN10 = 40,
+  VVCPROFILE_MAIN10_444 = 41,
+  VVCPROFILE_MAIN12_444 = 42,
+  VVCPROFILE_MAIN16_444 = 43,
+  VVCPROFILE_MAIN12_444_INTRA = 44,
+  VVCPROFILE_MAIN16_444_INTRA = 45,
+  VVCPROFILE_MULTILAYER_MAIN10_444 = 46,
+  VVCPROFILE_MAIN10_STILL_PICTURE = 47,
+  VVCPROFILE_MAIN12_STILL_PICTURE = 48,
+  VVCPROFILE_MAIN10_444_STILL_PICTURE = 49,
+  VVCPROFILE_MAIN12_444_STILL_PICTURE = 50,
+  VVCPROFILE_MAIN16_444_STILL_PICTURE = 51,
+  VVCPROFILE_MAX = VVCPROFILE_MAIN16_444_STILL_PICTURE,
+  VIDEO_CODEC_PROFILE_MAX = VVCPROFILE_MAIN16_444_STILL_PICTURE,
+}
+
+export type VideoCodecLevel = number; // uint32
+export const NoVideoCodecLevel: VideoCodecLevel = 0;
+
+export type VideoCodecProfileLevel = {
+  codec: VideoCodec;
+  profile: VideoCodecProfile;
+  level: VideoCodecLevel;
+};

packages/core/src/data/bit.ts

@@ -0,0 +1,39 @@
+export class BitReader {
+  private data: Uint8Array;
+  private byteOffset = 0;
+  private bitOffset = 0;
+
+  constructor(data: Uint8Array) {
+    this.data = data;
+  }
+
+  readBits(numBits: number): number {
+    let value = 0;
+    for (let i = 0; i < numBits; i++) {
+      const bit = (this.data[this.byteOffset] >> (7 - this.bitOffset)) & 1;
+      value = (value << 1) | bit;
+      this.bitOffset++;
+      if (this.bitOffset === 8) {
+        this.bitOffset = 0;
+        this.byteOffset++;
+      }
+    }
+    return value;
+  }
+
+  skipBits(numBits: number): void {
+    this.bitOffset += numBits;
+    while (this.bitOffset >= 8) {
+      this.bitOffset -= 8;
+      this.byteOffset++;
+    }
+  }
+
+  hasData(): boolean {
+    return this.byteOffset < this.data.length;
+  }
+
+  getRemainingBytes(): Uint8Array {
+    return this.data.slice(this.byteOffset);
+  }
+}

packages/core/src/data/fetch.ts

@@ -0,0 +1,65 @@
+export interface RangedStream {
+  controller: AbortController;
+  response: Response;
+  body: ReadableStream;
+  totalSize?: number;
+}
+
+export interface CreateRangedStreamOptions {
+  url: string;
+  byteStart?: number;
+  byteEnd?: number;
+}
+
+export async function createRangedStream({
+  url,
+  byteStart = 0,
+  byteEnd,
+}: CreateRangedStreamOptions) {
+  const controller = new AbortController();
+  const signal = controller.signal;
+  const headers = new Headers();
+  headers.append(
+    'Range',
+    typeof byteEnd === 'number'
+      ? `bytes=${byteStart}-${byteEnd}`
+      : `bytes=${byteStart}-`
+  );
+
+  const response = await fetch(url, { signal, headers });
+
+  if (!response.ok) {
+    throw new Error('fetch video stream failed');
+  }
+
+  const acceptRanges = response.headers.get('Accept-Ranges');
+
+  if (acceptRanges !== 'bytes') {
+    throw new Error('video server does not support byte ranges');
+  }
+
+  const body = response.body;
+
+  if (!(body instanceof ReadableStream)) {
+    throw new Error('can not get readable stream from response.body');
+  }
+
+  const contentRange = response.headers.get('Content-Range');
+
+  //
+  // Content-Range Header Syntax:
+  // Content-Range: <unit> <range-start>-<range-end>/<size>
+  // Content-Range: <unit> <range-start>-<range-end>/*
+  // Content-Range: <unit> */<size>
+  //
+  const totalSize = contentRange
+    ? Number.parseInt(contentRange.split('/')[1], 10)
+    : undefined;
+
+  return {
+    controller,
+    response,
+    body,
+    totalSize,
+  };
+}

packages/core/src/data/index.ts

@@ -0,0 +1,6 @@
+export {
+  type RangedStream,
+  type CreateRangedStreamOptions,
+  createRangedStream,
+} from './fetch';
+export { BitReader } from './bit';

packages/core/src/errors.ts

@@ -0,0 +1,25 @@
+export class UnsupportedCodecError extends Error {
+  constructor(codec: string, context: string) {
+    super(`codec ${codec} is not supported in ${context} context`);
+  }
+}
+
+export class ParseCodecError extends Error {
+  constructor(codec: string, detail: string) {
+    super(`code ${codec} private parse failed: ${detail}`);
+  }
+}
+
+export class UnreachableOrLogicError extends Error {
+  constructor(detail: string) {
+    super(`unreachable or logic error: ${detail}`);
+  }
+}
+
+export class ParseCodecErrors extends Error {
+  cause: Error[] = [];
+
+  constructor() {
+    super('failed to parse codecs');
+  }
+}

packages/core/tsconfig.json

@@ -0,0 +1,10 @@
+{
+    "extends": "../../tsconfig.base.json",
+    "compilerOptions": {
+        "composite": true,
+        "outDir": "./dist"
+    },
+    "include": [
+        "src"
+    ]
+}

packages/ffmpeg/Cargo.toml

@@ -1,5 +1,5 @@
 [package]
-name = "konoplayer-codecs"
+name = "konoplayer-ffmpeg"
 version = "0.1.0"
 edition = "2024"
 

packages/matroska/package.json

@@ -0,0 +1,11 @@
+{
+    "name": "@konoplayer/matroska",
+    "version": "0.1.0",
+    "private": true,
+    "type": "module",
+    "scripts": {},
+    "dependencies": {
+        "@konoplayer/core": "workspace:*",
+        "konoebml": "^0.1.2"
+    }
+}

packages/matroska/src/codecs/aac.ts

@@ -0,0 +1,110 @@
+import { ParseCodecError } from '@konoplayer/core/errors';
+import { ArkErrors, type } from 'arktype';
+
+export const AAC_CODEC_TYPE = 'AAC';
+
+export const AudioObjectTypeSchema = type('1 | 2 | 3 | 4 | 5 | 29 | 67');
+
+export const SamplingFrequencyIndexSchema = type('1|2|3|4|5|6|7|8|9|10|11|12');
+
+export const ChannelConfigurationSchema = type('1 | 2 | 3 | 4 | 5 | 6 | 7');
+
+export const AudioSpecificConfigSchema = type({
+  audioObjectType: AudioObjectTypeSchema, // AAC profiles: Main, LC, SSR, LTP, HE, HE v2
+  samplingFrequencyIndex: SamplingFrequencyIndexSchema.optional(), // Sampling rate index
+  channelConfiguration: ChannelConfigurationSchema, // Channel config (1-7)
+  sbrPresent: type.boolean.optional(), // Optional: Indicates SBR presence
+  psPresent: type.boolean.optional(), // Optional: Indicates PS presence (for HE-AAC v2)
+});
+
+export type AudioSpecificConfigType = typeof AudioSpecificConfigSchema.infer;
+
+/**
+ * Parse AudioSpecificConfig from codec_private Uint8Array
+ * @param codecPrivate - Uint8Array containing codec_private data
+ * @returns Parsed AudioSpecificConfig or throws an error if invalid
+ */
+export function parseAudioSpecificConfig(
+  codecPrivate: Uint8Array
+): AudioSpecificConfigType {
+  if (codecPrivate.length < 2) {
+    throw new ParseCodecError(AAC_CODEC_TYPE, 'codec_private data too short');
+  }
+
+  // Create a DataView for bit-level manipulation
+  const view = new DataView(
+    codecPrivate.buffer,
+    codecPrivate.byteOffset,
+    codecPrivate.byteLength
+  );
+  let byteOffset = 0;
+  let bitOffset = 0;
+
+  // Helper function to read specific number of bits
+  function readBits(bits: number): number {
+    let value = 0;
+    for (let i = 0; i < bits; i++) {
+      const byte = view.getUint8(byteOffset);
+      const bit = (byte >> (7 - bitOffset)) & 1;
+      value = (value << 1) | bit;
+      bitOffset++;
+      if (bitOffset === 8) {
+        bitOffset = 0;
+        byteOffset++;
+      }
+    }
+    return value;
+  }
+
+  // Read 5 bits for audioObjectType
+  const audioObjectType = readBits(5);
+
+  // Read 4 bits for samplingFrequencyIndex
+  const samplingFrequencyIndex = readBits(4);
+
+  // Read 4 bits for channelConfiguration
+  const channelConfiguration = readBits(4);
+
+  // Check for SBR/PS extension (if audioObjectType indicates HE-AAC)
+  let sbrPresent = false;
+  let psPresent = false;
+  if (audioObjectType === 5 || audioObjectType === 29) {
+    sbrPresent = true;
+    if (audioObjectType === 29) {
+      psPresent = true; // HE-AAC v2 includes Parametric Stereo
+    }
+    // Skip extension-specific bits if present (simplified here)
+    // In real cases, additional parsing may be needed
+  }
+
+  // Construct the result object
+  const config: AudioSpecificConfigType = {
+    audioObjectType:
+      audioObjectType as AudioSpecificConfigType['audioObjectType'],
+    samplingFrequencyIndex:
+      samplingFrequencyIndex as AudioSpecificConfigType['samplingFrequencyIndex'],
+    channelConfiguration:
+      channelConfiguration as AudioSpecificConfigType['channelConfiguration'],
+    ...(sbrPresent && { sbrPresent }),
+    ...(psPresent && { psPresent }),
+  };
+
+  // Validate with arktype
+  const validation = AudioSpecificConfigSchema(config);
+  if (validation instanceof ArkErrors) {
+    const error = new ParseCodecError(
+      AAC_CODEC_TYPE,
+      'Invalid AudioSpecificConfig'
+    );
+    error.cause = validation;
+    throw error;
+  }
+
+  return config;
+}
+
+export function genCodecIdByAudioSpecificConfig(
+  config: AudioSpecificConfigType
+) {
+  return `mp4a.40.${config.audioObjectType}`;
+}

packages/matroska/src/codecs/av1.ts

@@ -0,0 +1,167 @@
+import { BitReader } from '@konoplayer/core/data';
+import { type } from 'arktype';
+import type { TrackEntryType } from '../schema';
+import { ParseCodecError } from '@konoplayer/core/errors';
+
+export const AV1_CODEC_TYPE = 'AV1';
+
+export const AV1DecoderConfigurationRecordSchema = type({
+  marker: type.number, // 1 bit, must be 1
+  version: type.number, // 7 bits, must be 1
+  seqProfile: type.number, // 3 bits, seq profile (0-7)
+  seqLevelIdx0: type.number, // 5 bits, seq level (0-31)
+  seqTier0: type.number, // 1 bit, tier (0 or 1)
+  highBitdepth: type.number, // 1 bit, high or low
+  twelveBit: type.number, // 1 bit, if 12-bit
+  monochrome: type.number, // 1 bit, if mono chrome
+  chromaSubsamplingX: type.number, // 1 bit, sub sampling X
+  chromaSubsamplingY: type.number, // 1 bit, sub sampling Y
+  chromaSamplePosition: type.number, // 2 bits
+  initialPresentationDelayPresent: type.number, // 1 bit
+  initialPresentationDelayMinus1: type.number.optional(), // 4 bits, optoinal
+  configOBUs: type.instanceOf(Uint8Array<ArrayBufferLike>), // remain OBU data
+});
+
+export type AV1DecoderConfigurationRecordType =
+  typeof AV1DecoderConfigurationRecordSchema.infer;
+
+/**
+ * [webkit impl](https://github.com/movableink/webkit/blob/7e43fe7000b319ce68334c09eed1031642099726/Source/WebCore/platform/graphics/AV1Utilities.cpp#L48)
+ */
+export function parseAV1DecoderConfigurationRecord(
+  track: TrackEntryType
+): AV1DecoderConfigurationRecordType {
+  const codecPrivate = track.CodecPrivate;
+
+  if (!codecPrivate) {
+    throw new ParseCodecError(
+      AV1_CODEC_TYPE,
+      'CodecPrivate of AVC Track is missing'
+    );
+  }
+
+  if (codecPrivate.length < 4) {
+    throw new ParseCodecError(
+      AV1_CODEC_TYPE,
+      'Input data too short for AV1DecoderConfigurationRecord'
+    );
+  }
+
+  const reader = new BitReader(codecPrivate);
+
+  // Byte 0
+  const marker = reader.readBits(1);
+  const version = reader.readBits(7);
+  if (marker !== 1 || version !== 1) {
+    throw new ParseCodecError(
+      AV1_CODEC_TYPE,
+      `Invalid marker (${marker}) or version (${version})`
+    );
+  }
+
+  const seqProfile = reader.readBits(3);
+  const seqLevelIdx0 = reader.readBits(5);
+
+  // Byte 1
+  const seqTier0 = reader.readBits(1);
+  const highBitdepth = reader.readBits(1);
+  const twelveBit = reader.readBits(1);
+  const monochrome = reader.readBits(1);
+  const chromaSubsamplingX = reader.readBits(1);
+  const chromaSubsamplingY = reader.readBits(1);
+  const chromaSamplePosition = reader.readBits(2);
+
+  // Byte 2
+  const reserved1 = reader.readBits(3);
+  if (reserved1 !== 0) {
+    throw new ParseCodecError(
+      AV1_CODEC_TYPE,
+      `Reserved bits must be 0, got ${reserved1}`
+    );
+  }
+  const initialPresentationDelayPresent = reader.readBits(1);
+  let initialPresentationDelayMinus1: number | undefined;
+  if (initialPresentationDelayPresent) {
+    initialPresentationDelayMinus1 = reader.readBits(4);
+  } else {
+    const reserved2 = reader.readBits(4);
+    if (reserved2 !== 0) {
+      throw new ParseCodecError(
+        AV1_CODEC_TYPE,
+        `Reserved bits must be 0, got ${reserved2}`
+      );
+    }
+  }
+
+  // remain bytes as configOBUs
+  const configOBUs = reader.getRemainingBytes();
+
+  return {
+    marker,
+    version,
+    seqProfile,
+    seqLevelIdx0,
+    seqTier0,
+    highBitdepth,
+    twelveBit,
+    monochrome,
+    chromaSubsamplingX,
+    chromaSubsamplingY,
+    chromaSamplePosition,
+    initialPresentationDelayPresent,
+    initialPresentationDelayMinus1,
+    configOBUs,
+  };
+}
+
+/**
+ * [webkit impl](https://github.com/movableink/webkit/blob/7e43fe7000b319ce68334c09eed1031642099726/Source/WebCore/platform/graphics/AV1Utilities.cpp#L197)
+ */
+export function genCodecStringByAV1DecoderConfigurationRecord(
+  config: AV1DecoderConfigurationRecordType
+): string {
+  const parts: string[] = [];
+
+  // Prefix
+  parts.push('av01');
+
+  // Profile
+  parts.push(config.seqProfile.toString());
+
+  // Level and Tier
+  const levelStr = config.seqLevelIdx0.toString().padStart(2, '0');
+  const tierStr = config.seqTier0 === 0 ? 'M' : 'H';
+  parts.push(`${levelStr}${tierStr}`);
+
+  // Bit Depth
+  let bitDepthStr: string;
+  if (config.highBitdepth === 0) {
+    bitDepthStr = '08'; // 8-bit
+  } else if (config.twelveBit === 0) {
+    bitDepthStr = '10'; // 10-bit
+  } else {
+    bitDepthStr = '12'; // 12-bit
+  }
+  parts.push(bitDepthStr);
+
+  // Monochrome
+  parts.push(config.monochrome.toString());
+
+  // Chroma Subsampling
+  const chromaSubsampling = `${config.chromaSubsamplingX}${config.chromaSubsamplingY}${config.chromaSamplePosition}`;
+  parts.push(chromaSubsampling);
+
+  // Initial Presentation Delay（optional）
+  if (
+    config.initialPresentationDelayPresent === 1 &&
+    config.initialPresentationDelayMinus1 !== undefined
+  ) {
+    const delay = (config.initialPresentationDelayMinus1 + 1)
+      .toString()
+      .padStart(2, '0');
+    parts.push(delay);
+  }
+
+  // joined
+  return parts.join('.');
+}

packages/matroska/src/codecs/avc.ts

@@ -0,0 +1,148 @@
+import { ParseCodecError } from '@konoplayer/core/errors';
+import { type } from 'arktype';
+import type { TrackEntryType } from '../schema';
+
+export const AVC_CODEC_TYPE = 'h264(AVC)';
+
+export const AVCDecoderConfigurationRecordSchema = type({
+  configurationVersion: type.number, // Configuration version, typically 1
+  avcProfileIndication: type.number, // AVC profile
+  profileCompatibility: type.number, // Profile compatibility
+  avcLevelIndication: type.number, // AVC level
+  lengthSizeMinusOne: type.number, // NAL unit length field size minus 1
+  sps: type
+    .instanceOf(Uint8Array<ArrayBufferLike>)
+    .array()
+    .atLeastLength(1), // Sequence Parameter Sets (SPS)
+  pps: type
+    .instanceOf(Uint8Array<ArrayBufferLike>)
+    .array()
+    .atLeastLength(1), // Picture Parameter Sets (PPS)
+});
+
+export type AVCDecoderConfigurationRecordType =
+  typeof AVCDecoderConfigurationRecordSchema.infer;
+
+/**
+ *
+ * @see [webkit](https://github.com/movableink/webkit/blob/7e43fe7000b319ce68334c09eed1031642099726/Source/WebCore/platform/graphics/HEVCUtilities.cpp#L84)
+ */
+export function parseAVCDecoderConfigurationRecord(
+  track: TrackEntryType
+): AVCDecoderConfigurationRecordType {
+  // ISO/IEC 14496-10:2014
+  // 7.3.2.1.1 Sequence parameter set data syntax
+  const codecPrivate = track.CodecPrivate;
+
+  if (!codecPrivate) {
+    throw new ParseCodecError(
+      AVC_CODEC_TYPE,
+      'CodecPrivate of AVC Track is missing'
+    );
+  }
+
+  // AVCDecoderConfigurationRecord is at a minimum 24 bytes long
+  if (codecPrivate.length < 24) {
+    throw new ParseCodecError(
+      AVC_CODEC_TYPE,
+      'Input data too short for AVCDecoderConfigurationRecord'
+    );
+  }
+
+  const view = new DataView(codecPrivate.buffer);
+  let offset = 0;
+
+  const readUint8 = (move: boolean) => {
+    const result = view.getUint8(offset);
+    if (move) {
+      offset += 1;
+    }
+    return result;
+  };
+
+  const readUint16 = (move: boolean) => {
+    const result = view.getUint16(offset, false);
+    if (move) {
+      offset += 2;
+    }
+    return result;
+  };
+
+  const configurationVersion = readUint8(true);
+  const avcProfileIndication = readUint8(true);
+  const profileCompatibility = readUint8(true);
+  const avcLevelIndication = readUint8(true);
+
+  // Read lengthSizeMinusOne (first 6 bits are reserved, typically 0xFF, last 2 bits are the value)
+  const lengthSizeMinusOne = readUint8(true) & 0x03;
+
+  // Read number of SPS (first 3 bits are reserved, typically 0xE0, last 5 bits are SPS count)
+  const numOfSPS = readUint8(true) & 0x1f;
+  const sps: Uint8Array[] = [];
+
+  // Parse SPS
+  for (let i = 0; i < numOfSPS; i++) {
+    if (offset + 2 > codecPrivate.length) {
+      throw new ParseCodecError(AVC_CODEC_TYPE, 'Invalid SPS length');
+    }
+
+    const spsLength = readUint16(true);
+
+    if (offset + spsLength > codecPrivate.length) {
+      throw new ParseCodecError(
+        AVC_CODEC_TYPE,
+        'SPS data exceeds buffer length'
+      );
+    }
+
+    sps.push(codecPrivate.subarray(offset, offset + spsLength));
+    offset += spsLength;
+  }
+
+  // Read number of PPS
+  if (offset >= codecPrivate.length) {
+    throw new ParseCodecError(AVC_CODEC_TYPE, 'No space for PPS count');
+  }
+  const numOfPPS = readUint8(true);
+  const pps: Uint8Array[] = [];
+
+  // Parse PPS
+  for (let i = 0; i < numOfPPS; i++) {
+    if (offset + 2 > codecPrivate.length) {
+      throw new ParseCodecError(AVC_CODEC_TYPE, 'Invalid PPS length');
+    }
+
+    const ppsLength = readUint16(true);
+
+    if (offset + ppsLength > codecPrivate.length) {
+      throw new ParseCodecError(
+        AVC_CODEC_TYPE,
+        'PPS data exceeds buffer length'
+      );
+    }
+
+    pps.push(codecPrivate.subarray(offset, offset + ppsLength));
+    offset += ppsLength;
+  }
+
+  return {
+    configurationVersion,
+    avcProfileIndication,
+    profileCompatibility,
+    avcLevelIndication,
+    lengthSizeMinusOne,
+    sps,
+    pps,
+  };
+}
+
+export function genCodecStringByAVCDecoderConfigurationRecord(
+  config: AVCDecoderConfigurationRecordType
+): string {
+  const profileHex = config.avcProfileIndication.toString(16).padStart(2, '0');
+  const profileCompatHex = config.profileCompatibility
+    .toString(16)
+    .padStart(2, '0');
+  const levelHex = config.avcLevelIndication.toString(16).padStart(2, '0');
+  return `avc1.${profileHex}${profileCompatHex}${levelHex}`;
+}

packages/matroska/src/codecs/hevc.ts

@@ -0,0 +1,214 @@
+import { ParseCodecError } from '@konoplayer/core/errors';
+import { ArkErrors, type } from 'arktype';
+import type { TrackEntryType } from '../schema';
+
+export const HEVC_CODEC_TYPE = 'h265(HEVC)';
+
+export const HEVCDecoderConfigurationRecordArraySchema = type({
+  arrayCompleteness: type.number,
+  nalUnitType: type.number,
+  numNalus: type.number,
+  nalUnit: type.instanceOf(Uint8Array<ArrayBufferLike>).array(),
+});
+
+export type HEVCDecoderConfigurationRecordArrayType =
+  typeof HEVCDecoderConfigurationRecordArraySchema.infer;
+
+// Define the schema for HEVCDecoderConfigurationRecord
+export const HEVCDecoderConfigurationRecordSchema = type({
+  configurationVersion: type.number, // Must be 1
+  generalProfileSpace: type.number,
+  generalTierFlag: type.number,
+  generalProfileIdc: type.number,
+  generalProfileCompatibilityFlags: type.number,
+  generalConstraintIndicatorFlags: type.number,
+  generalLevelIdc: type.number,
+  minSpatialSegmentationIdc: type.number,
+  parallelismType: type.number,
+  chromaFormat: type.number,
+  bitDepthLumaMinus8: type.number,
+  bitDepthChromaMinus8: type.number,
+  avgFrameRate: type.number,
+  constantFrameRate: type.number,
+  numTemporalLayers: type.number,
+  temporalIdNested: type.number,
+  lengthSizeMinusOne: type.number,
+  numOfArrays: type.number,
+  nalUnits: HEVCDecoderConfigurationRecordArraySchema.array(),
+});
+
+export type HEVCDecoderConfigurationRecordType =
+  typeof HEVCDecoderConfigurationRecordSchema.infer;
+
+export function parseHEVCDecoderConfigurationRecord(
+  track: TrackEntryType
+): HEVCDecoderConfigurationRecordType {
+  const codecPrivate = track.CodecPrivate;
+  if (!codecPrivate) {
+    throw new ParseCodecError(
+      HEVC_CODEC_TYPE,
+      'CodecPrivate of HEVC Track is missing'
+    );
+  }
+  const view = new DataView(codecPrivate.buffer);
+  let offset = 0;
+
+  const readUint8 = (move: boolean) => {
+    const result = view.getUint8(offset);
+    if (move) {
+      offset += 1;
+    }
+    return result;
+  };
+
+  const readUint16 = (move: boolean) => {
+    const result = view.getUint16(offset, false);
+    if (move) {
+      offset += 2;
+    }
+    return result;
+  };
+
+  const readUint48 = (move: boolean) => {
+    const result =
+      view.getUint16(offset, false) * 2 ** 32 +
+      view.getUint32(offset + 2, false);
+    if (move) {
+      offset += 6;
+    }
+    return result;
+  };
+
+  const readUint32 = (move: boolean) => {
+    const result = view.getUint32(offset, false);
+    if (move) {
+      offset += 4;
+    }
+    return result;
+  };
+
+  // Read and validate basic fields
+  const config: HEVCDecoderConfigurationRecordType = {
+    configurationVersion: readUint8(true),
+    generalProfileSpace: (readUint8(false) & 0xc0) >> 6,
+    generalTierFlag: (readUint8(false) & 0x20) >> 5,
+    generalProfileIdc: readUint8(true) & 0x1f,
+    generalProfileCompatibilityFlags: readUint32(true),
+    generalConstraintIndicatorFlags: readUint48(true),
+    generalLevelIdc: readUint8(true),
+    minSpatialSegmentationIdc: readUint16(true) & 0x0fff,
+    parallelismType: readUint8(true) & 0x03,
+    chromaFormat: readUint8(true) & 0x03,
+    bitDepthLumaMinus8: readUint8(true) & 0x07,
+    bitDepthChromaMinus8: readUint8(true) & 0x07,
+    avgFrameRate: readUint16(true),
+    constantFrameRate: (readUint8(false) & 0xc0) >> 6,
+    numTemporalLayers: (readUint8(false) & 0x38) >> 3,
+    temporalIdNested: (readUint8(false) & 0x04) >> 2,
+    lengthSizeMinusOne: readUint8(true) & 0x03,
+    numOfArrays: readUint8(true),
+    nalUnits: [],
+  };
+
+  // Parse NAL unit arrays
+  const arrays = config.nalUnits;
+
+  for (let i = 0; i < config.numOfArrays; i++) {
+    const array: HEVCDecoderConfigurationRecordArrayType = {
+      arrayCompleteness: (readUint8(false) & 0x80) >> 7,
+      nalUnitType: readUint8(true) & 0x3f,
+      numNalus: readUint16(true),
+      nalUnit: [] as Uint8Array<ArrayBufferLike>[],
+    };
+
+    for (let j = 0; j < array.numNalus; j++) {
+      const nalUnitLength = readUint16(true);
+      array.nalUnit.push(codecPrivate.subarray(offset, offset + nalUnitLength));
+      offset += nalUnitLength;
+    }
+    arrays.push(array);
+  }
+
+  // Validate using arktype
+  const validation = HEVCDecoderConfigurationRecordSchema(config);
+  if (validation instanceof ArkErrors) {
+    const error = new ParseCodecError(
+      HEVC_CODEC_TYPE,
+      'Invalid HEVC configuration record'
+    );
+    error.cause = validation;
+    throw error;
+  }
+
+  return validation;
+}
+
+function reverseBits32(value: number): number {
+  let result = 0;
+  for (let i = 0; i < 32; i++) {
+    result = (result << 1) | ((value >> i) & 1);
+  }
+  return result;
+}
+
+/**
+ * @see[webkit implementation](https://github.com/movableink/webkit/blob/7e43fe7000b319ce68334c09eed1031642099726/Source/WebCore/platform/graphics/HEVCUtilities.cpp#L204)
+ */
+export function genCodecStringByHEVCDecoderConfigurationRecord(
+  config: HEVCDecoderConfigurationRecordType
+) {
+  const result: string[] = [];
+
+  // prefix
+  result.push(`hev${config.configurationVersion}`);
+
+  // Profile Space
+  if (config.generalProfileSpace > 0) {
+    const profileSpaceChar = String.fromCharCode(
+      'A'.charCodeAt(0) + config.generalProfileSpace - 1
+    );
+    result.push(profileSpaceChar + config.generalProfileIdc.toString());
+  } else {
+    result.push(config.generalProfileIdc.toString());
+  }
+
+  // Profile Compatibility Flags
+  const compatFlags = reverseBits32(config.generalProfileCompatibilityFlags)
+    .toString(16)
+    .toUpperCase();
+  result.push(compatFlags);
+
+  // Tier Flag and Level IDC
+  const tierPrefix = config.generalTierFlag ? 'H' : 'L';
+  result.push(tierPrefix + config.generalLevelIdc.toString());
+
+  // Constraint Indicator Flags
+  let constraintBytes: number[];
+  if (Array.isArray(config.generalConstraintIndicatorFlags)) {
+    constraintBytes = config.generalConstraintIndicatorFlags as number[];
+  } else {
+    // split 48 bit integer into 6 byte
+    const flags = BigInt(config.generalConstraintIndicatorFlags);
+    constraintBytes = [];
+    for (let i = 5; i >= 0; i--) {
+      constraintBytes.push(Number((flags >> BigInt(8 * i)) & BigInt(0xff)));
+    }
+  }
+
+  // find last non-zero byte
+  const lastNonZeroIndex = constraintBytes.reduce(
+    (last, byte, i) => (byte ? i : last),
+    -1
+  );
+  if (lastNonZeroIndex >= 0) {
+    for (let i = 0; i <= lastNonZeroIndex; i++) {
+      const byteHex = constraintBytes[i]
+        .toString(16)
+        .padStart(2, '0')
+        .toUpperCase();
+      result.push(byteHex);
+    }
+  }
+
+  return result.join('.');
+}

packages/matroska/src/codecs/index.ts

@@ -0,0 +1,347 @@
+import { UnsupportedCodecError } from '@konoplayer/core/errors';
+import { VideoCodec, AudioCodec } from '@konoplayer/core/codecs';
+import type { TrackEntryType } from '../schema';
+import {
+  genCodecIdByAudioSpecificConfig,
+  parseAudioSpecificConfig,
+} from './aac';
+import {
+  genCodecStringByAVCDecoderConfigurationRecord,
+  parseAVCDecoderConfigurationRecord,
+} from './avc';
+import {
+  genCodecStringByAV1DecoderConfigurationRecord,
+  parseAV1DecoderConfigurationRecord,
+} from './av1.ts';
+import {
+  genCodecStringByHEVCDecoderConfigurationRecord,
+  parseHEVCDecoderConfigurationRecord,
+} from './hevc.ts';
+import {
+  genCodecStringByVP9DecoderConfigurationRecord,
+  parseVP9DecoderConfigurationRecord,
+} from './vp9.ts';
+
+export const VideoCodecId = {
+  VCM: 'V_MS/VFW/FOURCC',
+  UNCOMPRESSED: 'V_UNCOMPRESSED',
+  MPEG4_ISO_SP: 'V_MPEG4/ISO/SP',
+  MPEG4_ISO_ASP: 'V_MPEG4/ISO/ASP',
+  MPEG4_ISO_AP: 'V_MPEG4/ISO/AP',
+  MPEG4_MS_V3: 'V_MPEG4/MS/V3',
+  MPEG1: 'V_MPEG1',
+  MPEG2: 'V_MPEG2',
+  H264: 'V_MPEG4/ISO/AVC',
+  HEVC: 'V_MPEGH/ISO/HEVC',
+  AVS2: 'V_AVS2',
+  AVS3: 'V_AVS3',
+  RV10: 'V_REAL/RV10',
+  RV20: 'V_REAL/RV20',
+  RV30: 'V_REAL/RV30',
+  RV40: 'V_REAL/RV40',
+  QUICKTIME: 'V_QUICKTIME',
+  THEORA: 'V_THEORA',
+  PROPRES: 'V_PRORES',
+  VP8: 'V_VP8',
+  VP9: 'V_VP9',
+  FFV1: 'V_FFV1',
+  AV1: 'V_AV1',
+} as const;
+
+export type VideoCodecIdType =
+  | `${(typeof VideoCodecId)[keyof typeof VideoCodecId]}`
+  | string;
+
+export const AudioCodecId = {
+  MPEG_L3: 'A_MPEG/L3',
+  MPEG_L2: 'A_MPEG/L2',
+  MPEG_L1: 'A_MPEG/L1',
+  PCM_INT_BIG: 'A_PCM/INT/BIG',
+  PCM_INT_LIT: 'A_PCM/INT/LIT',
+  PCM_FLOAT_IEEE: 'A_PCM/FLOAT/IEEE',
+  MPC: 'A_MPC',
+  AC3: 'A_AC3',
+  AC3_BSID9: 'A_AC3/BSID9',
+  AC3_BSID10: 'A_AC3/BSID10',
+  ALAC: 'A_ALAC',
+  DTS: 'A_DTS',
+  DTS_EXPRESS: 'A_DTS/EXPRESS',
+  DTS_LOSSLESS: 'A_DTS/LOSSLESS',
+  VORBIS: 'A_VORBIS',
+  OPUS: 'A_OPUS',
+  FLAC: 'A_FLAC',
+  EAC3: 'A_EAC3',
+  REAL_14_4: 'A_REAL/14_4',
+  REAL_28_8: 'A_REAL/28_8',
+  REAL_COOK: 'A_REAL/COOK',
+  REAL_SIPR: 'A_REAL/SIPR',
+  REAL_RALF: 'A_REAL/RALF',
+  REAL_ATRC: 'A_REAL/ATRC',
+  MS_ACM: 'A_MS/ACM',
+  AAC: 'A_AAC',
+  AAC_MPEG2_MAIN: 'A_AAC/MPEG2/MAIN',
+  AAC_MPEG2_LC: 'A_AAC/MPEG2/LC',
+  AAC_MPEG2_LC_SBR: 'A_AAC/MPEG2/LC/SBR',
+  AAC_MPEG2_SSR: 'A_AAC/MPEG2/SSR',
+  AAC_MPEG4_MAIN: 'A_AAC/MPEG4/MAIN',
+  AAC_MPEG4_LC: 'A_AAC/MPEG4/LC',
+  AAC_MPEG4_SBR: 'A_AAC/MPEG4/LC/SBR',
+  AAC_MPEG4_SSR: 'A_AAC/MPEG4/SSR',
+  AAC_MPEG4_LTP: 'A_AAC/MPEG4/LTP',
+  QUICKTIME: 'A_QUICKTIME',
+  QDMC: 'A_QUICKTIME/QDMC',
+  QDM2: 'A_QUICKTIME/QDM2',
+  TTA1: 'A_TTA1',
+  WAVEPACK4: 'A_WAVPACK4',
+  ATRAC: 'A_ATRAC/AT1',
+} as const;
+
+export type AudioCodecIdType =
+  | `${(typeof AudioCodecId)[keyof typeof AudioCodecId]}`
+  | string;
+
+export const SubtitleCodecId = {
+  UTF8: 'S_TEXT/UTF8',
+  SSA: 'S_TEXT/SSA',
+  ASS: 'S_TEXT/ASS',
+  WEBVTT: 'S_TEXT/WEBVTT',
+  BMP: 'S_IMAGE/BMP',
+  DVBSUB: 'S_DVBSUB',
+  VOBSUB: 'S_VOBSUB',
+  HDMV_PGS: 'S_HDMV/PGS',
+  HDMV_TEXTST: 'S_HDMV/TEXTST',
+  KATE: 'S_KATE',
+  ARIBSUB: 'S_ARIBSUB',
+} as const;
+
+export type SubtitleCodecIdType =
+  | `${(typeof SubtitleCodecId)[keyof typeof SubtitleCodecId]}`
+  | string;
+
+export interface VideoDecoderConfigExt extends VideoDecoderConfig {
+  codecType: VideoCodec;
+}
+
+export function videoCodecIdToWebCodecs(
+  track: TrackEntryType,
+  keyframe: Uint8Array
+): VideoDecoderConfigExt {
+  const codecId = track.CodecID;
+  const codecPrivate = track.CodecPrivate;
+  const shareOptions = {
+    description: codecPrivate,
+  };
+  switch (codecId) {
+    case VideoCodecId.HEVC:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.HEVC,
+        codec: genCodecStringByHEVCDecoderConfigurationRecord(
+          parseHEVCDecoderConfigurationRecord(track)
+        ),
+      };
+    case VideoCodecId.VP9:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.VP9,
+        codec: genCodecStringByVP9DecoderConfigurationRecord(
+          parseVP9DecoderConfigurationRecord(track, keyframe)
+        ),
+      };
+    case VideoCodecId.AV1:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.AV1,
+        codec: genCodecStringByAV1DecoderConfigurationRecord(
+          parseAV1DecoderConfigurationRecord(track)
+        ),
+      };
+    case VideoCodecId.H264:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.H264,
+        codec: genCodecStringByAVCDecoderConfigurationRecord(
+          parseAVCDecoderConfigurationRecord(track)
+        ),
+      };
+    case VideoCodecId.THEORA:
+      return { ...shareOptions, codecType: VideoCodec.Theora, codec: 'theora' };
+    case VideoCodecId.VP8:
+      return { ...shareOptions, codecType: VideoCodec.VP8, codec: 'vp8' };
+    case VideoCodecId.MPEG4_ISO_SP:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.MPEG4,
+        codec: 'mp4v.01.3',
+      };
+    case VideoCodecId.MPEG4_ISO_ASP:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.MPEG4,
+        codec: 'mp4v.20.9',
+      };
+    case VideoCodecId.MPEG4_ISO_AP:
+      return {
+        ...shareOptions,
+        codecType: VideoCodec.MPEG4,
+        codec: 'mp4v.20.9',
+      };
+    default:
+      throw new UnsupportedCodecError(codecId, 'web codecs video decoder');
+  }
+}
+
+export interface AudioDecoderConfigExt extends AudioDecoderConfig {
+  codecType: AudioCodec;
+}
+
+export function audioCodecIdToWebCodecs(
+  track: TrackEntryType
+): AudioDecoderConfigExt {
+  const codecId = track.CodecID;
+  const codecPrivate = track.CodecPrivate;
+  const bitDepth = track.Audio?.BitDepth;
+  const numberOfChannels = Number(track.Audio?.Channels);
+  const sampleRate = Number(track.Audio?.SamplingFrequency);
+
+  const shareOptions = {
+    numberOfChannels,
+    sampleRate,
+    description: codecPrivate,
+  };
+
+  switch (track.CodecID) {
+    case AudioCodecId.AAC_MPEG4_MAIN:
+    case AudioCodecId.AAC_MPEG2_MAIN:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: 'mp4a.40.1',
+      };
+    case AudioCodecId.AAC_MPEG2_LC:
+    case AudioCodecId.AAC_MPEG4_LC:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: 'mp4a.40.2',
+      };
+    case AudioCodecId.AAC_MPEG2_SSR:
+    case AudioCodecId.AAC_MPEG4_SSR:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: 'mp4a.40.3',
+      };
+    case AudioCodecId.AAC_MPEG4_LTP:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: 'mp4a.40.4',
+      };
+    case AudioCodecId.AAC_MPEG2_LC_SBR:
+    case AudioCodecId.AAC_MPEG4_SBR:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: 'mp4a.40.5',
+      };
+    case AudioCodecId.AAC:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AAC,
+        codec: codecPrivate
+          ? genCodecIdByAudioSpecificConfig(
+              parseAudioSpecificConfig(codecPrivate)
+            )
+          : 'mp4a.40.2',
+      };
+    case AudioCodecId.AC3:
+    case AudioCodecId.AC3_BSID9:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.AC3,
+        codec: 'ac-3',
+      };
+    case AudioCodecId.EAC3:
+    case AudioCodecId.AC3_BSID10:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.EAC3,
+        codec: 'ec-3',
+      };
+    case AudioCodecId.MPEG_L3:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.MP3,
+        codec: 'mp3',
+      };
+    case AudioCodecId.VORBIS:
+      return { ...shareOptions, codecType: AudioCodec.Vorbis, codec: 'vorbis' };
+    case AudioCodecId.FLAC:
+      return { ...shareOptions, codecType: AudioCodec.FLAC, codec: 'flac' };
+    case AudioCodecId.OPUS:
+      return { ...shareOptions, codecType: AudioCodec.Opus, codec: 'opus' };
+    case AudioCodecId.ALAC:
+      return { ...shareOptions, codecType: AudioCodec.ALAC, codec: 'alac' };
+    case AudioCodecId.PCM_INT_BIG:
+      if (bitDepth === 16) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S16BE,
+          codec: 'pcm-s16be',
+        };
+      }
+      if (bitDepth === 24) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S24BE,
+          codec: 'pcm-s24be',
+        };
+      }
+      if (bitDepth === 32) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S32BE,
+          codec: 'pcm-s32be',
+        };
+      }
+      throw new UnsupportedCodecError(
+        `${codecId}(${bitDepth}b)`,
+        'web codecs audio decoder'
+      );
+    case AudioCodecId.PCM_INT_LIT:
+      if (bitDepth === 16) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S16LE,
+          codec: 'pcm-s16le',
+        };
+      }
+      if (bitDepth === 24) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S24LE,
+          codec: 'pcm-s24le',
+        };
+      }
+      if (bitDepth === 32) {
+        return {
+          ...shareOptions,
+          codecType: AudioCodec.PCM_S32LE,
+          codec: 'pcm-s32le',
+        };
+      }
+      throw new UnsupportedCodecError(
+        `${codecId}(${bitDepth}b)`,
+        'web codecs audio decoder'
+      );
+    case AudioCodecId.PCM_FLOAT_IEEE:
+      return {
+        ...shareOptions,
+        codecType: AudioCodec.PCM_F32LE,
+        codec: 'pcm-f32le',
+      };
+    default:
+      throw new UnsupportedCodecError(codecId, 'web codecs audio decoder');
+  }
+}

packages/matroska/src/codecs/vp9.ts

@@ -0,0 +1,232 @@
+import { type } from 'arktype';
+import type { TrackEntryType } from '../schema';
+import { BitReader } from '@konoplayer/core/data';
+import { ParseCodecError } from '@konoplayer/core/errors';
+
+export const VP9_CODEC_TYPE = 'vp9';
+
+export enum VP9ColorSpaceEnum {
+  UNKNOWN = 0,
+  BT_601 = 1, // eq bt_470bg
+  BT_709 = 2,
+  SMPTE_170 = 3,
+  SMPTE_240 = 4,
+  BT_2020 = 5,
+  RESERVED = 6,
+  SRGB = 7,
+}
+
+export enum VP9YUVRange {
+  STUDIO_SWING = 0,
+  FULL_SWING = 1,
+}
+
+export enum VP9Subsampling {
+  UNKNOWN = 0,
+  YUV420 = 1,
+  YUV422 = 2,
+  YUV440 = 3,
+  YUV444 = 4,
+}
+
+export const VP9PerformenceLevel = [
+  { level: '10', maxSampleRate: 829440, maxResolution: 36864 }, // Level 1
+  { level: '11', maxSampleRate: 2764800, maxResolution: 73728 }, // Level 1
+  { level: '20', maxSampleRate: 4608000, maxResolution: 122880 }, // Level 2
+  { level: '21', maxSampleRate: 9216000, maxResolution: 245760 }, // Level 2.1
+  { level: '30', maxSampleRate: 20736000, maxResolution: 552960 }, // Level 3
+  { level: '31', maxSampleRate: 36864000, maxResolution: 983040 }, // Level 3.1
+  { level: '40', maxSampleRate: 83558400, maxResolution: 2228224 }, // Level 4
+  { level: '41', maxSampleRate: 160432128, maxResolution: 2228224 }, // Level 4.1
+  { level: '50', maxSampleRate: 311951360, maxResolution: 8912896 }, // Level 5
+  { level: '51', maxSampleRate: 588251136, maxResolution: 8912896 }, // Level 5.1
+  { level: '52', maxSampleRate: 1176502272, maxResolution: 8912896 }, // Level 5.2
+  { level: '60', maxSampleRate: 1176502272, maxResolution: 35651584 }, // Level 6
+  { level: '61', maxSampleRate: 2353004544, maxResolution: 35651584 }, // Level 6.1
+  { level: '62', maxSampleRate: 4706009088, maxResolution: 35651584 }, // Level 6.2
+];
+
+export const VP9DecoderConfigurationRecordSchema = type({
+  profile: type.number, // 0 | 1 | 2 | 3,
+  bitDepth: type.number, // 8 | 10 | 12
+  colorSpace: type.number,
+  subsampling: type.number, // 420 | 422 | 444
+  width: type.number,
+  height: type.number,
+  yuvRangeFlag: type.number.optional(),
+  hasScaling: type.boolean,
+  renderWidth: type.number,
+  renderHeight: type.number,
+  frameRate: type.number, // frame per second
+  estimateLevel: type.string,
+});
+
+export type VP9DecoderConfigurationRecordType =
+  typeof VP9DecoderConfigurationRecordSchema.infer;
+
+export function parseVP9DecoderConfigurationRecord(
+  track: TrackEntryType,
+  keyframe: Uint8Array
+): VP9DecoderConfigurationRecordType {
+  const reader = new BitReader(keyframe);
+  const frameRate = 1_000_000_000 / Number(track.DefaultDuration) || 30;
+
+  // Frame Marker: 2 bits, must be 0b10
+  const frameMarker = reader.readBits(2);
+  if (frameMarker !== 2) {
+    throw new ParseCodecError(VP9_CODEC_TYPE, 'invalid frame marker');
+  }
+
+  // Profile: 2 bits
+  const version = reader.readBits(1);
+  const high = reader.readBits(1);
+
+  const profile = (high << 1) + version;
+
+  let reservedZero = 0;
+  if (profile === 3) {
+    reservedZero = reader.readBits(1);
+    if (reservedZero !== 0) {
+      throw new ParseCodecError(
+        VP9_CODEC_TYPE,
+        'Invalid reserved zero bit for profile 3'
+      );
+    }
+  }
+
+  // Show Existing Frame: 1 bit
+  const showExistingFrame = reader.readBits(1);
+  if (showExistingFrame === 1) {
+    throw new ParseCodecError(VP9_CODEC_TYPE, 'not a keyframe to parse');
+  }
+
+  // Frame Type: 1 bit (0 = keyframe)
+  const frameType = reader.readBits(1);
+  if (frameType !== 0) {
+    throw new ParseCodecError(VP9_CODEC_TYPE, 'not a keyframe to parse');
+  }
+
+  // Show Frame and Error Resilient
+  const _showFrame = reader.readBits(1);
+  const _errorResilient = reader.readBits(1);
+
+  // Sync Code: 3 bytes (0x49, 0x83, 0x42)
+  const syncCode =
+    (reader.readBits(8) << 16) | (reader.readBits(8) << 8) | reader.readBits(8);
+  if (syncCode !== 0x498342) {
+    throw new ParseCodecError(VP9_CODEC_TYPE, 'Invalid sync code');
+  }
+
+  // Bit Depth
+  let bitDepth: number;
+  if (profile >= 2) {
+    const tenOrTwelveBit = reader.readBits(1);
+    bitDepth = tenOrTwelveBit === 0 ? 10 : 12;
+  } else {
+    bitDepth = 8;
+  }
+
+  const colorSpace = reader.readBits(3);
+
+  let subsamplingX: number;
+  let subsamplingY: number;
+  let yuvRangeFlag: number | undefined;
+  if (colorSpace !== VP9ColorSpaceEnum.SRGB) {
+    yuvRangeFlag = reader.readBits(1);
+    if (profile === 1 || profile === 3) {
+      subsamplingX = reader.readBits(1);
+      subsamplingY = reader.readBits(1);
+      reservedZero = reader.readBits(1);
+    } else {
+      subsamplingX = 1;
+      subsamplingY = 1;
+    }
+  } else {
+    if (profile !== 1 && profile !== 3) {
+      throw new ParseCodecError(
+        VP9_CODEC_TYPE,
+        'VP9 profile with sRGB ColorSpace must be 1 or 3'
+      );
+    }
+    subsamplingX = 0;
+    subsamplingY = 0;
+    reservedZero = reader.readBits(1);
+  }
+
+  let subsampling: VP9Subsampling;
+
+  if (!subsamplingX && subsamplingY) {
+    subsampling = VP9Subsampling.YUV440;
+  } else if (subsamplingX && !subsamplingY) {
+    subsampling = VP9Subsampling.YUV422;
+  } else if (subsamplingX && subsamplingY) {
+    subsampling = VP9Subsampling.YUV420;
+  } else if (!subsamplingX && !subsamplingY) {
+    subsampling = VP9Subsampling.YUV444;
+  } else {
+    subsampling = VP9Subsampling.UNKNOWN;
+  }
+
+  // Frame Size (resolution)
+  const widthMinus1 = reader.readBits(16);
+  const heightMinus1 = reader.readBits(16);
+  const hasScaling = !!reader.readBits(1);
+  let renderWidthMinus1 = widthMinus1;
+  let renderHeightMinus1 = heightMinus1;
+  if (hasScaling) {
+    renderWidthMinus1 = reader.readBits(16);
+    renderHeightMinus1 = reader.readBits(16);
+  }
+
+  const width = widthMinus1 + 1;
+  const height = heightMinus1 + 1;
+
+  const sampleRate = width * height * frameRate;
+  const resolution = width * height;
+
+  let estimateLevel = '62';
+  for (const { level, maxSampleRate, maxResolution } of VP9PerformenceLevel) {
+    if (sampleRate <= maxSampleRate && resolution <= maxResolution) {
+      // 检查 profile 和 bitDepth 的额外要求
+      if (profile >= 2 && bitDepth > 8 && Number.parseFloat(level) < 20) {
+        continue;
+      }
+      estimateLevel = level;
+      break;
+    }
+  }
+
+  return {
+    profile,
+    bitDepth,
+    colorSpace,
+    subsampling,
+    yuvRangeFlag,
+    width,
+    height,
+    hasScaling,
+    renderWidth: renderWidthMinus1 + 1,
+    renderHeight: renderHeightMinus1 + 1,
+    frameRate,
+    estimateLevel,
+  };
+}
+
+// The format of the 'vp09' codec string is specified in the webm GitHub repo:
+// <https://github.com/webmproject/vp9-dash/blob/master/VPCodecISOMediaFileFormatBinding.md#codecs-parameter-string>
+//
+// The codecs parameter string for the VP codec family is as follows:
+//   <sample entry 4CC>.<profile>.<level>.<bitDepth>.<chromaSubsampling>.
+//   <colourPrimaries>.<transferCharacteristics>.<matrixCoefficients>.
+//   <videoFullRangeFlag>
+//  All parameter values are expressed as double-digit decimals.
+//  sample entry 4CC, profile, level, and bitDepth are all mandatory fields.
+export function genCodecStringByVP9DecoderConfigurationRecord(
+  config: VP9DecoderConfigurationRecordType
+): string {
+  const profileStr = config.profile.toString().padStart(2, '0');
+  const bitDepthStr = config.bitDepth.toString().padStart(2, '0');
+  const levelStr = config.estimateLevel;
+
+  return `vp09.${profileStr}.${levelStr}.${bitDepthStr}`;
+}

packages/matroska/src/model.ts

@@ -0,0 +1,469 @@
+import {
+  type EbmlClusterTagType,
+  type EbmlCuePointTagType,
+  type EbmlCuesTagType,
+  type EbmlInfoTagType,
+  type EbmlMasterTagType,
+  type EbmlSeekHeadTagType,
+  type EbmlSegmentTagType,
+  EbmlTagIdEnum,
+  EbmlTagPosition,
+  type EbmlTagsTagType,
+  type EbmlTagTagType,
+  type EbmlTagType,
+  type EbmlTrackEntryTagType,
+  type EbmlTracksTagType,
+} from 'konoebml';
+import { convertEbmlTagToComponent, type InferType } from './util';
+import { isEqual, maxBy } from 'lodash-es';
+import { ArkErrors, type Type } from 'arktype';
+import {
+  ClusterSchema,
+  type ClusterType,
+  CuePointSchema,
+  type CuePointType,
+  type CueTrackPositionsType,
+  InfoSchema,
+  type InfoType,
+  SeekHeadSchema,
+  type SeekHeadType,
+  TagSchema,
+  type TagType,
+  TrackEntrySchema,
+  type TrackEntryType,
+  TrackTypeRestrictionEnum,
+} from './schema';
+import { concatBufs } from 'konoebml/lib/tools';
+import {
+  ParseCodecErrors,
+  UnreachableOrLogicError,
+  UnsupportedCodecError,
+} from '@konoplayer/core/errors';
+import { audioCodecIdToWebCodecs, videoCodecIdToWebCodecs } from './codecs';
+import { Queue } from 'mnemonist';
+import { BehaviorSubject } from 'rxjs';
+
+export const SEEK_ID_KAX_INFO = new Uint8Array([0x15, 0x49, 0xa9, 0x66]);
+export const SEEK_ID_KAX_TRACKS = new Uint8Array([0x16, 0x54, 0xae, 0x6b]);
+export const SEEK_ID_KAX_CUES = new Uint8Array([0x1c, 0x53, 0xbb, 0x6b]);
+export const SEEK_ID_KAX_TAGS = new Uint8Array([0x12, 0x54, 0xc3, 0x67]);
+
+export class SegmentSystem {
+  startTag: EbmlSegmentTagType;
+  headTags: EbmlTagType[] = [];
+  firstCluster: EbmlClusterTagType | undefined;
+
+  cue: CueSystem;
+  cluster: ClusterSystem;
+  seek: SeekSystem;
+  info: InfoSystem;
+  track: TrackSystem;
+  tag: TagSystem;
+
+  constructor(startNode: EbmlSegmentTagType) {
+    this.startTag = startNode;
+    this.cue = new CueSystem(this);
+    this.cluster = new ClusterSystem(this);
+    this.seek = new SeekSystem(this);
+    this.info = new InfoSystem(this);
+    this.track = new TrackSystem(this);
+    this.tag = new TagSystem(this);
+  }
+
+  get contentStartOffset() {
+    return this.startTag.startOffset + this.startTag.headerLength;
+  }
+
+  private seekLocal() {
+    const infoTag = this.seek.seekTagBySeekId(SEEK_ID_KAX_INFO);
+    const tracksTag = this.seek.seekTagBySeekId(SEEK_ID_KAX_TRACKS);
+    const cuesTag = this.seek.seekTagBySeekId(SEEK_ID_KAX_CUES);
+    const tagsTag = this.seek.seekTagBySeekId(SEEK_ID_KAX_TAGS);
+
+    if (cuesTag?.id === EbmlTagIdEnum.Cues) {
+      this.cue.prepareCuesWithTag(cuesTag);
+    }
+    if (infoTag?.id === EbmlTagIdEnum.Info) {
+      this.info.prepareWithInfoTag(infoTag);
+    }
+    if (tracksTag?.id === EbmlTagIdEnum.Tracks) {
+      this.track.prepareTracksWithTag(tracksTag);
+    }
+    if (tagsTag?.id === EbmlTagIdEnum.Tags) {
+      this.tag.prepareTagsWithTag(tagsTag);
+    }
+  }
+
+  scanMeta(tag: EbmlTagType) {
+    if (
+      tag.id === EbmlTagIdEnum.SeekHead &&
+      tag.position === EbmlTagPosition.End
+    ) {
+      this.seek.addSeekHeadTag(tag);
+    }
+    this.headTags.push(tag);
+    this.seek.memoTag(tag);
+    if (tag.id === EbmlTagIdEnum.Cluster && !this.firstCluster) {
+      this.firstCluster = tag;
+      this.seekLocal();
+    }
+    return this;
+  }
+
+  async completeMeta() {
+    this.seekLocal();
+
+    await this.parseCodes();
+
+    return this;
+  }
+
+  async parseCodes() {
+    const candidates = this.track.tracks.filter(
+      (c) =>
+        c.TrackType === TrackTypeRestrictionEnum.AUDIO ||
+        c.TrackType === TrackTypeRestrictionEnum.VIDEO
+    );
+    const parseErrors = new ParseCodecErrors();
+
+    for (const t of candidates) {
+      try {
+        await this.track.initTrack(t, this.);
+      } catch (e) {
+        parseErrors.cause.push(e as Error);
+      }
+    }
+    if (parseErrors.cause.length > 0) {
+      console.error(parseErrors);
+    }
+  }
+}
+
+export type SegmentComponent<T> = T & {
+  get segment(): SegmentSystem;
+};
+
+export function withSegment<T extends object>(
+  component: T,
+  segment: SegmentSystem
+): SegmentComponent<T> {
+  const component_ = component as T & { segment: SegmentSystem };
+  component_.segment = segment;
+  return component_;
+}
+
+export class SegmentComponentSystemTrait<
+  E extends EbmlMasterTagType,
+  S extends Type<any>,
+> {
+  segment: SegmentSystem;
+
+  get schema(): S {
+    throw new Error('unimplemented!');
+  }
+
+  constructor(segment: SegmentSystem) {
+    this.segment = segment;
+  }
+
+  componentFromTag(tag: E): SegmentComponent<InferType<S>> {
+    const extracted = convertEbmlTagToComponent(tag);
+    const result = this.schema(extracted) as
+      | (InferType<S> & { segment: SegmentSystem })
+      | ArkErrors;
+    if (result instanceof ArkErrors) {
+      const errors = result;
+      console.error(
+        'Parse component from tag error:',
+        tag.toDebugRecord(),
+        errors.flatProblemsByPath
+      );
+      throw errors;
+    }
+    result.segment = this.segment;
+    return result;
+  }
+}
+
+export class SeekSystem extends SegmentComponentSystemTrait<
+  EbmlSeekHeadTagType,
+  typeof SeekHeadSchema
+> {
+  override get schema() {
+    return SeekHeadSchema;
+  }
+
+  seekHeads: SeekHeadType[] = [];
+  private offsetToTagMemo: Map<number, EbmlTagType> = new Map();
+
+  memoTag(tag: EbmlTagType) {
+    this.offsetToTagMemo.set(tag.startOffset, tag);
+  }
+
+  addSeekHeadTag(tag: EbmlSeekHeadTagType) {
+    const seekHead = this.componentFromTag(tag);
+    this.seekHeads.push(seekHead);
+    return seekHead;
+  }
+
+  offsetFromSeekPosition(position: number): number {
+    return position + this.segment.contentStartOffset;
+  }
+
+  seekTagByStartOffset(
+    startOffset: number | undefined
+  ): EbmlTagType | undefined {
+    return startOffset! >= 0
+      ? this.offsetToTagMemo.get(startOffset!)
+      : undefined;
+  }
+
+  seekOffsetBySeekId(seekId: Uint8Array): number | undefined {
+    const seekPosition = this.seekHeads[0]?.Seek?.find((c) =>
+      isEqual(c.SeekID, seekId)
+    )?.SeekPosition;
+    return seekPosition! >= 0
+      ? this.offsetFromSeekPosition(seekPosition! as number)
+      : undefined;
+  }
+
+  seekTagBySeekId(seekId: Uint8Array): EbmlTagType | undefined {
+    return this.seekTagByStartOffset(this.seekOffsetBySeekId(seekId));
+  }
+
+  get firstClusterOffset() {
+    if (!this.segment.firstCluster) {
+      throw new UnreachableOrLogicError('first cluster not found');
+    }
+    return this.segment.firstCluster.startOffset;
+  }
+}
+
+export class InfoSystem extends SegmentComponentSystemTrait<
+  EbmlInfoTagType,
+  typeof InfoSchema
+> {
+  override get schema() {
+    return InfoSchema;
+  }
+
+  info!: SegmentComponent<InfoType>;
+
+  prepareWithInfoTag(tag: EbmlInfoTagType) {
+    this.info = this.componentFromTag(tag);
+    return this;
+  }
+}
+
+export class ClusterSystem extends SegmentComponentSystemTrait<
+  EbmlClusterTagType,
+  typeof ClusterSchema
+> {
+  override get schema() {
+    return ClusterSchema;
+  }
+
+  clustersBuffer: SegmentComponent<ClusterType>[] = [];
+
+  addClusterWithTag(tag: EbmlClusterTagType) {
+    const cluster = this.componentFromTag(tag);
+    this.clustersBuffer.push(cluster);
+    return cluster;
+  }
+}
+
+export interface GetTrackEntryOptions {
+  priority?: (v: SegmentComponent<TrackEntryType>) => number;
+  predicate?: (v: SegmentComponent<TrackEntryType>) => boolean;
+}
+
+export interface TrackState<Decoder, Config, Frame> {
+  decoder: Decoder;
+  configuration?: Config;
+  frameBuffer$: BehaviorSubject<Queue<Frame>>;
+}
+
+export class TrackSystem extends SegmentComponentSystemTrait<
+  EbmlTrackEntryTagType,
+  typeof TrackEntrySchema
+> {
+  override get schema() {
+    return TrackEntrySchema;
+  }
+
+  tracks: SegmentComponent<TrackEntryType>[] = [];
+  videoTrackState = new WeakMap<
+    TrackEntryType,
+    TrackState<VideoDecoder, VideoDecoderConfig, VideoFrame>
+  >();
+  audioTrackState = new WeakMap<
+    TrackEntryType,
+    TrackState<AudioDecoder, AudioDecoderConfig, AudioData>
+  >();
+
+  getTrackEntry({
+    priority = (track) =>
+      (Number(!!track.FlagForced) << 4) + Number(!!track.FlagDefault),
+    predicate = (track) => track.FlagEnabled !== 0,
+  }: GetTrackEntryOptions) {
+    return this.tracks
+      .filter(predicate)
+      .toSorted((a, b) => priority(b) - priority(a))
+      .at(0);
+  }
+
+  prepareTracksWithTag(tag: EbmlTracksTagType) {
+    this.tracks = tag.children
+      .filter((c) => c.id === EbmlTagIdEnum.TrackEntry)
+      .map((c) => this.componentFromTag(c));
+    return this;
+  }
+
+  async initTrack(track: TrackEntryType) {
+    if (track.TrackType === TrackTypeRestrictionEnum.AUDIO) {
+      const configuration = audioCodecIdToWebCodecs(track);
+      if (await AudioDecoder.isConfigSupported(configuration)) {
+        throw new UnsupportedCodecError(configuration.codec, 'audio decoder');
+      }
+
+      const queue$ = new BehaviorSubject(new Queue<AudioData>());
+      this.audioTrackState.set(track, {
+        configuration,
+        decoder: new AudioDecoder({
+          output: (audioData) => {
+            const queue = queue$.getValue();
+            queue.enqueue(audioData);
+            queue$.next(queue);
+          },
+          error: (e) => {
+            queue$.error(e);
+          },
+        }),
+        frameBuffer$: queue$,
+      });
+    } else if (track.TrackType === TrackTypeRestrictionEnum.VIDEO) {
+      const configuration = videoCodecIdToWebCodecs(track, this.keyframe);
+      if (await VideoDecoder.isConfigSupported(configuration)) {
+        throw new UnsupportedCodecError(configuration.codec, 'audio decoder');
+      }
+
+      const queue$ = new BehaviorSubject(new Queue<VideoFrame>());
+      this.videoTrackState.set(track, {
+        configuration,
+        decoder: new VideoDecoder({
+          output: (audioData) => {
+            const queue = queue$.getValue();
+            queue.enqueue(audioData);
+            queue$.next(queue);
+          },
+          error: (e) => {
+            queue$.error(e);
+          },
+        }),
+        frameBuffer$: queue$,
+      });
+    }
+  }
+}
+
+export class CueSystem extends SegmentComponentSystemTrait<
+  EbmlCuePointTagType,
+  typeof CuePointSchema
+> {
+  override get schema() {
+    return CuePointSchema;
+  }
+
+  cues: SegmentComponent<CuePointType>[] = [];
+
+  prepareCuesWithTag(tag: EbmlCuesTagType) {
+    this.cues = tag.children
+      .filter((c) => c.id === EbmlTagIdEnum.CuePoint)
+      .map(this.componentFromTag.bind(this));
+    return this;
+  }
+
+  findClosestCue(seekTime: number): CuePointType | undefined {
+    const cues = this.cues;
+    if (!cues || cues.length === 0) {
+      return undefined;
+    }
+
+    let left = 0;
+    let right = cues.length - 1;
+
+    if (seekTime <= cues[0].CueTime) {
+      return cues[0];
+    }
+
+    if (seekTime >= cues[right].CueTime) {
+      return cues[right];
+    }
+
+    while (left <= right) {
+      const mid = Math.floor((left + right) / 2);
+
+      if (cues[mid].CueTime === seekTime) {
+        return cues[mid];
+      }
+
+      if (cues[mid].CueTime < seekTime) {
+        left = mid + 1;
+      } else {
+        right = mid - 1;
+      }
+    }
+
+    const before = cues[right];
+    const after = cues[left];
+    return Math.abs((before.CueTime as number) - seekTime) <
+      Math.abs((after.CueTime as number) - seekTime)
+      ? before
+      : after;
+  }
+
+  getCueTrackPositions(
+    cuePoint: CuePointType,
+    track?: number
+  ): CueTrackPositionsType {
+    let cueTrackPositions: CueTrackPositionsType | undefined;
+    if (track! >= 0) {
+      cueTrackPositions = cuePoint.CueTrackPositions.find(
+        (c) => c.CueTrack === track
+      );
+    }
+    if (!cueTrackPositions) {
+      cueTrackPositions = maxBy(
+        cuePoint.CueTrackPositions,
+        (c) => c.CueClusterPosition
+      )!;
+    }
+    return cueTrackPositions;
+  }
+
+  get prepared(): boolean {
+    return this.cues.length > 0;
+  }
+}
+
+export class TagSystem extends SegmentComponentSystemTrait<
+  EbmlTagTagType,
+  typeof TagSchema
+> {
+  override get schema() {
+    return TagSchema;
+  }
+
+  tags: SegmentComponent<TagType>[] = [];
+
+  prepareTagsWithTag(tag: EbmlTagsTagType) {
+    this.tags = tag.children
+      .filter((c) => c.id === EbmlTagIdEnum.Tag)
+      .map((c) => this.componentFromTag(c));
+    return this;
+  }
+
+  get prepared(): boolean {
+    return this.tags.length > 0;
+  }
+}

packages/matroska/src/reactive.ts

@@ -0,0 +1,430 @@
+import {
+  EbmlStreamDecoder,
+  EbmlTagIdEnum,
+  EbmlTagPosition,
+  type EbmlTagType,
+} from 'konoebml';
+import {
+  defer,
+  EMPTY,
+  filter,
+  finalize,
+  from,
+  isEmpty,
+  map,
+  merge,
+  Observable,
+  of,
+  reduce,
+  scan,
+  share,
+  shareReplay,
+  switchMap,
+  take,
+  takeWhile,
+  withLatestFrom,
+} from 'rxjs';
+import {
+  createRangedStream,
+  type CreateRangedStreamOptions,
+} from '@konoplayer/core/data';
+import {
+  type CueSystem,
+  SEEK_ID_KAX_CUES,
+  SEEK_ID_KAX_TAGS,
+  type SegmentComponent,
+  SegmentSystem,
+} from './model';
+import { isTagIdPos, waitTick } from './util';
+import type { ClusterType } from './schema';
+
+export interface CreateRangedEbmlStreamOptions
+  extends CreateRangedStreamOptions {
+  tee?: boolean;
+}
+
+export function createRangedEbmlStream({
+  url,
+  byteStart = 0,
+  byteEnd,
+  tee = false,
+}: CreateRangedEbmlStreamOptions): Observable<{
+  ebml$: Observable<EbmlTagType>;
+  totalSize?: number;
+  response: Response;
+  body: ReadableStream<Uint8Array>;
+  controller: AbortController;
+  teeBody: ReadableStream<Uint8Array> | undefined;
+}> {
+  const stream$ = from(createRangedStream({ url, byteStart, byteEnd }));
+
+  return stream$.pipe(
+    switchMap(({ controller, body, totalSize, response }) => {
+      let requestCompleted = false;
+      let teeStream: ReadableStream<Uint8Array> | undefined;
+
+      let stream: ReadableStream<Uint8Array>;
+
+      if (tee) {
+        [stream, teeStream] = body.tee();
+      } else {
+        stream = body;
+      }
+
+      const originRequest$ = new Observable<EbmlTagType>((subscriber) => {
+        stream
+          .pipeThrough(
+            new EbmlStreamDecoder({
+              streamStartOffset: byteStart,
+              collectChild: (child) => child.id !== EbmlTagIdEnum.Cluster,
+              backpressure: {
+                eventLoop: waitTick,
+              },
+            })
+          )
+          .pipeTo(
+            new WritableStream({
+              write: async (tag) => {
+                await waitTick();
+                subscriber.next(tag);
+              },
+              close: () => {
+                if (!requestCompleted) {
+                  requestCompleted = true;
+                  subscriber.complete();
+                }
+              },
+            })
+          )
+          .catch((error) => {
+            if (requestCompleted && error?.name === 'AbortError') {
+              return;
+            }
+            requestCompleted = true;
+            subscriber.error(error);
+          });
+
+        return () => {
+          requestCompleted = true;
+          controller.abort();
+        };
+      }).pipe(
+        share({
+          resetOnComplete: false,
+          resetOnError: false,
+          resetOnRefCountZero: true,
+        })
+      );
+
+      const ebml$ = defer(() =>
+        requestCompleted ? EMPTY : originRequest$
+      ).pipe(
+        share({
+          resetOnError: false,
+          resetOnComplete: true,
+          resetOnRefCountZero: true,
+        })
+      );
+
+      return of({
+        ebml$,
+        totalSize,
+        response,
+        body: stream,
+        teeBody: teeStream,
+        controller,
+      });
+    })
+  );
+}
+
+export interface CreateEbmlControllerOptions
+  extends Omit<CreateRangedEbmlStreamOptions, 'byteStart' | 'byteEnd'> {}
+
+export function createEbmlController({
+  url,
+  ...options
+}: CreateEbmlControllerOptions) {
+  const metaRequest$ = createRangedEbmlStream({
+    ...options,
+    url,
+    byteStart: 0,
+    tee: true,
+  });
+
+  const controller$ = metaRequest$.pipe(
+    map(({ totalSize, ebml$, response, controller, teeBody }) => {
+      const head$ = ebml$.pipe(
+        filter(isTagIdPos(EbmlTagIdEnum.EBML, EbmlTagPosition.End)),
+        take(1),
+        shareReplay(1)
+      );
+
+      console.debug(
+        `stream of video "${url}" created, total size is ${totalSize ?? 'unknown'}`
+      );
+
+      const segmentStart$ = ebml$.pipe(
+        filter(isTagIdPos(EbmlTagIdEnum.Segment, EbmlTagPosition.Start))
+      );
+
+      /**
+       * while [matroska v4](https://www.matroska.org/technical/elements.html) doc tell that there is only one segment in a file
+       * some mkv generated by strange tools will emit several
+       */
+      const segments$ = segmentStart$.pipe(
+        map((startTag) => {
+          const segment = new SegmentSystem(startTag, teeBody!);
+          const clusterSystem = segment.cluster;
+          const seekSystem = segment.seek;
+
+          const meta$ = ebml$.pipe(
+            scan(
+              (acc, tag) => {
+                // avoid object recreation
+                acc.hasKeyframe =
+                  acc.hasKeyframe ||
+                  (tag.id === EbmlTagIdEnum.SimpleBlock && tag.keyframe) ||
+                  (tag.id === EbmlTagIdEnum.BlockGroup &&
+                    tag.children.every(
+                      (c) => c.id !== EbmlTagIdEnum.ReferenceBlock
+                    ));
+                acc.tag = tag;
+                return acc;
+              },
+              { hasKeyframe: false, tag: undefined as unknown as EbmlTagType }
+            ),
+            takeWhile(({ tag, hasKeyframe }) => {
+              return (
+                !isTagIdPos(EbmlTagIdEnum.Segment, EbmlTagPosition.End)(tag) &&
+                !(
+                  isTagIdPos(EbmlTagIdEnum.Cluster, EbmlTagPosition.End)(tag) &&
+                  hasKeyframe
+                )
+              );
+            }, true),
+            map(({ tag }) => tag),
+            share({
+              resetOnComplete: false,
+              resetOnError: false,
+              resetOnRefCountZero: true,
+            })
+          );
+
+          const withMeta$ = meta$.pipe(
+            reduce((segment, meta) => segment.scanMeta(meta), segment),
+            switchMap(() => segment.completeMeta()),
+            take(1),
+            shareReplay(1)
+          );
+
+          const withRemoteCues$ = withMeta$.pipe(
+            switchMap((s) => {
+              const cueSystem = s.cue;
+              const seekSystem = s.seek;
+              if (cueSystem.prepared) {
+                return EMPTY;
+              }
+              const remoteCuesTagStartOffset =
+                seekSystem.seekOffsetBySeekId(SEEK_ID_KAX_CUES);
+              if (remoteCuesTagStartOffset! >= 0) {
+                return createRangedEbmlStream({
+                  ...options,
+                  url,
+                  byteStart: remoteCuesTagStartOffset,
+                }).pipe(
+                  switchMap((req) => req.ebml$),
+                  filter(isTagIdPos(EbmlTagIdEnum.Cues, EbmlTagPosition.End)),
+                  withLatestFrom(withMeta$),
+                  map(([cues, withMeta]) => {
+                    withMeta.cue.prepareCuesWithTag(cues);
+                    return withMeta;
+                  })
+                );
+              }
+              return EMPTY;
+            }),
+            take(1),
+            shareReplay(1)
+          );
+
+          const withLocalCues$ = withMeta$.pipe(
+            switchMap((s) => (s.cue.prepared ? of(s) : EMPTY)),
+            shareReplay(1)
+          );
+
+          const withRemoteTags$ = withMeta$.pipe(
+            switchMap((s) => {
+              const tagSystem = s.tag;
+              const seekSystem = s.seek;
+              if (tagSystem.prepared) {
+                return EMPTY;
+              }
+
+              const remoteTagsTagStartOffset =
+                seekSystem.seekOffsetBySeekId(SEEK_ID_KAX_TAGS);
+              if (remoteTagsTagStartOffset! >= 0) {
+                return createRangedEbmlStream({
+                  ...options,
+                  url,
+                  byteStart: remoteTagsTagStartOffset,
+                }).pipe(
+                  switchMap((req) => req.ebml$),
+                  filter(isTagIdPos(EbmlTagIdEnum.Tags, EbmlTagPosition.End)),
+                  withLatestFrom(withMeta$),
+                  map(([tags, withMeta]) => {
+                    withMeta.tag.prepareTagsWithTag(tags);
+                    return withMeta;
+                  })
+                );
+              }
+              return EMPTY;
+            }),
+            take(1),
+            shareReplay(1)
+          );
+
+          const withLocalTags$ = withMeta$.pipe(
+            switchMap((s) => (s.tag.prepared ? of(s) : EMPTY)),
+            shareReplay(1)
+          );
+
+          const withCues$ = merge(withLocalCues$, withRemoteCues$).pipe(
+            take(1)
+          );
+
+          const withoutCues$ = withCues$.pipe(
+            isEmpty(),
+            switchMap((empty) => (empty ? withMeta$ : EMPTY))
+          );
+
+          const withTags$ = merge(withLocalTags$, withRemoteTags$).pipe(
+            take(1)
+          );
+
+          const withoutTags$ = withTags$.pipe(
+            isEmpty(),
+            switchMap((empty) => (empty ? withMeta$ : EMPTY))
+          );
+
+          const seekWithoutCues = (
+            seekTime: number
+          ): Observable<SegmentComponent<ClusterType>> => {
+            const request$ = withMeta$.pipe(
+              switchMap(() =>
+                createRangedEbmlStream({
+                  ...options,
+                  url,
+                  byteStart: seekSystem.firstClusterOffset,
+                })
+              )
+            );
+            const cluster$ = request$.pipe(
+              switchMap((req) => req.ebml$),
+              filter(isTagIdPos(EbmlTagIdEnum.Cluster, EbmlTagPosition.End)),
+              map((tag) => clusterSystem.addClusterWithTag(tag))
+            );
+
+            if (seekTime === 0) {
+              return cluster$;
+            }
+
+            return cluster$.pipe(
+              scan(
+                (acc, curr) => {
+                  // avoid object recreation
+                  acc.prev = acc.next;
+                  acc.next = curr;
+                  return acc;
+                },
+                {
+                  prev: undefined as SegmentComponent<ClusterType> | undefined,
+                  next: undefined as SegmentComponent<ClusterType> | undefined,
+                }
+              ),
+              filter((c) => c.next?.Timestamp! > seekTime),
+              map((c) => c.prev ?? c.next!)
+            );
+          };
+
+          const seekWithCues = (
+            cueSystem: CueSystem,
+            seekTime: number
+          ): Observable<SegmentComponent<ClusterType>> => {
+            if (seekTime === 0) {
+              return seekWithoutCues(seekTime);
+            }
+
+            const cuePoint = cueSystem.findClosestCue(seekTime);
+
+            if (!cuePoint) {
+              return seekWithoutCues(seekTime);
+            }
+
+            return createRangedEbmlStream({
+              ...options,
+              url,
+              byteStart: seekSystem.offsetFromSeekPosition(
+                cueSystem.getCueTrackPositions(cuePoint)
+                  .CueClusterPosition as number
+              ),
+            }).pipe(
+              switchMap((req) => req.ebml$),
+              filter(isTagIdPos(EbmlTagIdEnum.Cluster, EbmlTagPosition.End)),
+              map(clusterSystem.addClusterWithTag.bind(clusterSystem))
+            );
+          };
+
+          const seek = (
+            seekTime: number
+          ): Observable<SegmentComponent<ClusterType>> => {
+            if (seekTime === 0) {
+              const subscription = merge(withCues$, withoutCues$).subscribe();
+
+              // if seekTime equals to 0 at start, reuse the initialize stream
+              return seekWithoutCues(seekTime).pipe(
+                finalize(() => {
+                  subscription.unsubscribe();
+                })
+              );
+            }
+            return merge(
+              withCues$.pipe(switchMap((s) => seekWithCues(s.cue, seekTime))),
+              withoutCues$.pipe(switchMap((_) => seekWithoutCues(seekTime)))
+            );
+          };
+
+          return {
+            startTag,
+            head$,
+            segment,
+            meta$,
+            withMeta$,
+            withCues$,
+            withoutCues$,
+            withTags$,
+            withoutTags$,
+            seekWithCues,
+            seekWithoutCues,
+            seek,
+          };
+        })
+      );
+
+      return {
+        segments$,
+        head$,
+        totalSize,
+        ebml$,
+        controller,
+        response,
+      };
+    }),
+    shareReplay(1)
+  );
+
+  return {
+    controller$,
+    request$: metaRequest$,
+  };
+}

packages/matroska/src/util.ts

@@ -0,0 +1,62 @@
+import type { Type } from 'arktype';
+import { EbmlElementType, EbmlTagIdEnum, type EbmlTagType } from 'konoebml';
+import { IdMultiSet } from './schema';
+
+export type InferType<T extends Type<any>> = T['infer'];
+
+export type PredicateIdExtract<T, K> = Extract<T, { id: K }>;
+
+export type PredicatePositionExtract<
+  T extends { position: string },
+  P,
+> = P extends T['position'] ? T : never;
+
+export function isTagIdPos<
+  I extends EbmlTagIdEnum,
+  P extends PredicateIdExtract<EbmlTagType, I>['position'] | '*' = '*',
+>(id: I, pos?: P) {
+  return (tag: EbmlTagType): tag is PredicateIdExtract<EbmlTagType, I> =>
+    tag.id === id && (pos === '*' || pos === tag.position);
+}
+
+export function isTagPos<
+  T extends { position: string },
+  P extends T['position'],
+>(pos: P | '*' = '*') {
+  return (tag: T): tag is PredicatePositionExtract<T, P> =>
+    pos === '*' || pos === tag.position;
+}
+
+export function convertEbmlTagToComponent(tag: EbmlTagType) {
+  if (tag.type === EbmlElementType.Master) {
+    const obj: Record<string, any> = {};
+    const children = tag.children;
+    for (const c of children) {
+      const name = EbmlTagIdEnum[c.id];
+      const converted = convertEbmlTagToComponent(c);
+      if (IdMultiSet.has(c.id)) {
+        if (obj[name]) {
+          obj[name].push(converted);
+        } else {
+          obj[name] = [converted];
+        }
+      } else {
+        obj[name] = converted;
+      }
+    }
+    return obj;
+  }
+  if (tag.id === EbmlTagIdEnum.SimpleBlock || tag.id === EbmlTagIdEnum.Block) {
+    return tag;
+  }
+  return tag.data;
+}
+
+export function waitTick() {
+  return new Promise<void>((resolve) => {
+    const timeout = setTimeout(() => {
+      resolve();
+      timeout && clearTimeout(timeout);
+    }, 0);
+  });
+}

packages/matroska/tsconfig.json

@@ -0,0 +1,20 @@
+{
+    "extends": "../../tsconfig.base.json",
+    "compilerOptions": {
+        "composite": true,
+        "outDir": "./dist",
+        "paths": {
+            "@konoplayer/core/*": [
+                "../core/src/*"
+            ]
+        }
+    },
+    "include": [
+        "src"
+    ],
+    "references": [
+        {
+            "path": "../core"
+        }
+    ]
+}

packages/symphonia/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
-name = "konoplayer-demuxing"
+name = "konoplayer-symphonia"
 version = "0.1.0"
 edition = "2024"
 
 [dependencies]
-symphonia-format-mkv = "0.5.4"
+symphonia = "0.5.4"