Add TrackDescriptor concept

This commit is contained in:
Vanilagy
2026-04-08 12:21:08 +02:00
parent b532714926
commit bee1518100
21 changed files with 1280 additions and 887 deletions
+120 -275
View File
@@ -11,11 +11,18 @@ import { determineVideoPacketType } from './codec-data';
import { customAudioDecoders, customVideoDecoders } from './custom-coder';
import { Input } from './input';
import { EncodedPacketSink, PacketRetrievalOptions } from './media-sink';
import { assert, MaybePromise, NonFunctionKeys, Rational, Rotation, roundToDivisor, simplifyRational } from './misc';
import { assert, Rational, Rotation, roundToDivisor, simplifyRational } from './misc';
import { TrackType } from './output';
import { EncodedPacket, PacketType } from './packet';
import { TrackDisposition } from './metadata';
import { DurationMetadataRequestOptions } from './demuxer';
import {
InputTrackDescriptor,
mergeTrackDescriptorQueries,
type InputVideoTrackDescriptor,
type InputAudioTrackDescriptor,
type TrackDescriptorQuery,
} from './input-track-descriptor';
/**
* Contains aggregate statistics about the encoded packets of a track.
@@ -32,14 +39,15 @@ export type PacketStats = {
};
export interface InputTrackBacking {
getType(): TrackType;
getId(): number;
getNumber(): number;
getCodec(): MediaCodec | null;
getInternalCodecId(): string | number | Uint8Array | null;
getName(): string | null;
getLanguageCode(): string;
getTimeResolution(): number;
isRelativeToUnixEpoch(): boolean;
getTimeResolution(): number | undefined;
isRelativeToUnixEpoch(): boolean | undefined;
getDisposition(): TrackDisposition;
getPairingMask(): bigint;
getBitrate(): number | null;
@@ -47,6 +55,8 @@ export interface InputTrackBacking {
getDurationFromMetadata(options: DurationMetadataRequestOptions): Promise<number | null>;
getLiveRefreshInterval(): Promise<number | null>;
getHasOnlyKeyPackets?(): boolean | null;
getDecoderConfig(): Promise<VideoDecoderConfig | AudioDecoderConfig | null>;
getCodecParameterString?(): string | null;
getFirstPacket(options: PacketRetrievalOptions): Promise<EncodedPacket | null>;
getPacket(timestamp: number, options: PacketRetrievalOptions): Promise<EncodedPacket | null>;
@@ -68,9 +78,6 @@ export abstract class InputTrack {
readonly input: Input;
/** @internal */
_backing: InputTrackBacking;
/** @internal */
_hydrationPromise: Promise<void> | null = null;
/** @internal */
constructor(input: Input, backing: InputTrackBacking) {
this.input = input;
@@ -151,7 +158,9 @@ export abstract class InputTrack {
* integer multiples of 1/x.
*/
get timeResolution() {
return this._backing.getTimeResolution();
const value = this._backing.getTimeResolution();
assert(value !== undefined);
return value;
}
/**
@@ -159,7 +168,9 @@ export abstract class InputTrack {
* each timestamp maps to a definitive point in time.
*/
get isRelativeToUnixEpoch() {
return this._backing.isRelativeToUnixEpoch();
const value = this._backing.isRelativeToUnixEpoch();
assert(value !== undefined);
return value;
}
/** The track's disposition, i.e. information about its intended usage. */
@@ -181,10 +192,6 @@ export abstract class InputTrack {
* with a negative timestamp should not be presented.
*/
async getFirstTimestamp() {
if (!this.isHydrated) {
await this.hydrate();
}
const firstPacket = await this._backing.getFirstPacket({ metadataOnly: true });
return firstPacket?.timestamp ?? 0;
}
@@ -197,10 +204,6 @@ export abstract class InputTrack {
* in the options.
*/
async computeDuration(options?: PacketRetrievalOptions) {
if (!this.isHydrated) {
await this.hydrate();
}
const lastPacket = await this._backing.getPacket(Infinity, { metadataOnly: true, ...options });
const result = (lastPacket?.timestamp ?? 0) + (lastPacket?.duration ?? 0);
@@ -218,10 +221,6 @@ export abstract class InputTrack {
* {@link DurationMetadataRequestOptions.skipLiveWait} to `true` in the options.
*/
async getDurationFromMetadata(options: DurationMetadataRequestOptions = {}) {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.getDurationFromMetadata(options);
}
@@ -277,136 +276,129 @@ export abstract class InputTrack {
}
async getLiveRefreshInterval() {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.getLiveRefreshInterval();
}
get isHydrated() {
return this._backing.isHydrated?.() ?? true;
}
hydrate(): Promise<void> {
if (this.isHydrated || !this._backing.hydrate) {
return Promise.resolve();
canBePairedWith(other: InputTrack | InputTrackDescriptor | null) {
if (!(other instanceof InputTrack || other instanceof InputTrackDescriptor || other === null)) {
throw new TypeError('other must be an InputTrack, InputTrackDescriptor, or null.');
}
return this._hydrationPromise ??= this._backing.hydrate();
}
canBePairedWith(otherTrack: InputTrack | null) {
if (!otherTrack) {
if (!other) {
return true;
}
if (this.input !== otherTrack.input || this === otherTrack) {
if (this.input !== other.input || this === other) {
return false;
}
return (this._backing.getPairingMask() & otherTrack._backing.getPairingMask()) !== 0n;
return (this._backing.getPairingMask() & other._backing.getPairingMask()) !== 0n;
}
async getPairableTracks(query?: TrackQuery<InputTrack>) {
const tracks = await this.input.getTracks();
return queryTracks(tracks.filter(x => this.canBePairedWith(x)), query);
async getPairableTracks(query?: TrackDescriptorQuery<InputTrackDescriptor>) {
const descriptors = await this.input.getTrackDescriptors(mergeTrackDescriptorQueries({
filter: d => d.canBePairedWith(this),
}, query));
return Promise.all(descriptors.map(d => d.getTrack()));
}
async pluckPairableTrack(query?: TrackQuery<InputTrack>) {
return (await this.getPairableTracks(query))[0];
async pluckPairableTrack(query?: TrackDescriptorQuery<InputTrackDescriptor>) {
return (await this.getPairableTracks(query))[0] ?? null;
}
async getPairableVideoTracks(query?: TrackQuery<InputVideoTrack>) {
const tracks = await this.getPairableTracks({
filter: track => track.isVideoTrack(),
});
return queryTracks(tracks as InputVideoTrack[], query);
async getPairableVideoTracks(query?: TrackDescriptorQuery<InputVideoTrackDescriptor>) {
const descriptors = await this.input.getVideoTrackDescriptors(mergeTrackDescriptorQueries({
filter: d => d.canBePairedWith(this),
}, query));
return Promise.all(descriptors.map(d => d.getTrack()));
}
async pluckPairableVideoTrack(query?: TrackQuery<InputVideoTrack>) {
async pluckPairableVideoTrack(query?: TrackDescriptorQuery<InputVideoTrackDescriptor>) {
return (await this.getPairableVideoTracks(query))[0] ?? null;
}
async getPairableAudioTracks(query?: TrackQuery<InputAudioTrack>) {
const tracks = await this.getPairableTracks({
filter: track => track.isAudioTrack(),
});
return queryTracks(tracks as InputAudioTrack[], query);
async getPairableAudioTracks(query?: TrackDescriptorQuery<InputAudioTrackDescriptor>) {
const descriptors = await this.input.getAudioTrackDescriptors(mergeTrackDescriptorQueries({
filter: d => d.canBePairedWith(this),
}, query));
return Promise.all(descriptors.map(d => d.getTrack()));
}
async pluckPairableAudioTrack(query?: TrackQuery<InputAudioTrack>) {
async pluckPairableAudioTrack(query?: TrackDescriptorQuery<InputAudioTrackDescriptor>) {
return (await this.getPairableAudioTracks(query))[0] ?? null;
}
async getPrimaryPairableVideoTrack(query?: TrackQuery<InputVideoTrack>) {
return this.input.getPrimaryVideoTrack(mergeTrackQueries({
filter: track => track.canBePairedWith(this),
async getPrimaryPairableVideoTrack(query?: TrackDescriptorQuery<InputVideoTrackDescriptor>) {
return this.input.getPrimaryVideoTrack(mergeTrackDescriptorQueries({
filter: d => d.canBePairedWith(this),
}, query));
}
async getPrimaryPairableAudioTrack(query?: TrackQuery<InputAudioTrack>) {
return this.input.getPrimaryAudioTrack(mergeTrackQueries({
filter: track => track.canBePairedWith(this),
async getPrimaryPairableAudioTrack(query?: TrackDescriptorQuery<InputAudioTrackDescriptor>) {
return this.input.getPrimaryAudioTrack(mergeTrackDescriptorQueries({
filter: d => d.canBePairedWith(this),
}, query));
}
hasPairableTrack(predicate?: (track: InputTrack) => boolean) {
assert(this.input._tracksCache);
return this.input._tracksCache.some(x => this.canBePairedWith(x) && (!predicate || predicate(x)));
}
hasPairableTrack(predicate?: (descriptor: InputTrackDescriptor) => boolean) {
predicate &&= toValidatedPredicate(predicate);
hasPairableVideoTrack(predictate?: (track: InputVideoTrack) => boolean): boolean {
return this.hasPairableTrack(x =>
x.isVideoTrack() && (!predictate || predictate(x)),
const descriptors = [...this.input._backingToDescriptor.values()];
return descriptors.some(x =>
this.canBePairedWith(x) && (!predicate || predicate(x)),
);
}
hasPairableAudioTrack(predictate?: (track: InputAudioTrack) => boolean): boolean {
hasPairableVideoTrack(predicate?: (descriptor: InputVideoTrackDescriptor) => boolean) {
predicate &&= toValidatedPredicate(predicate);
return this.hasPairableTrack(x =>
x.isAudioTrack() && (!predictate || predictate(x)),
x.isVideoTrackDescriptor() && (!predicate || predicate(x)),
);
}
get<K extends NonFunctionKeys<this>>(key: K): this[K] | null {
try {
return this[key];
} catch (error) {
if (error instanceof TrackNotHydratedError) {
return null;
}
hasPairableAudioTrack(predicate?: (descriptor: InputAudioTrackDescriptor) => boolean) {
predicate &&= toValidatedPredicate(predicate);
throw error;
}
}
async resolve<K extends NonFunctionKeys<this>>(key: K): Promise<this[K]> {
try {
return this[key];
} catch (error) {
if (error instanceof TrackNotHydratedError) {
await this.hydrate();
return this[key];
}
throw error;
}
return this.hasPairableTrack(x =>
x.isAudioTrackDescriptor() && (!predicate || predicate(x)),
);
}
}
const toValidatedPredicate = <T extends InputTrackDescriptor>(predicate?: (descriptor: T) => boolean) => {
if (predicate !== undefined && typeof predicate !== 'function') {
throw new TypeError('predicate, when provided, must be a function.');
}
return predicate
? (desc: T) => {
const result = predicate(desc);
if (typeof result !== 'boolean') {
throw new TypeError('predicate must return a boolean value.');
}
return result;
}
: undefined;
};
export interface InputVideoTrackBacking extends InputTrackBacking {
getType(): 'video';
getCodec(): VideoCodec | null;
getCodedWidth(): number;
getCodedHeight(): number;
getSquarePixelWidth(): number;
getSquarePixelHeight(): number;
getDisplayWidth?(): number | null;
getDisplayHeight?(): number | null;
getRotation(): Rotation;
getCodedWidth(): number | undefined;
getCodedHeight(): number | undefined;
getSquarePixelWidth(): number | undefined;
getSquarePixelHeight(): number | undefined;
getMetadataDisplayWidth?(): number | null;
getMetadataDisplayHeight?(): number | null;
getRotation(): Rotation | undefined;
getColorSpace(): Promise<VideoColorSpaceInit>;
canBeTransparent(): Promise<boolean>;
getDecoderConfig(): Promise<VideoDecoderConfig | null>;
getCodecParameterString?(): Promise<string | null>;
}
/**
@@ -441,27 +433,37 @@ export class InputVideoTrack extends InputTrack {
/** The width in pixels of the track's coded samples, before any transformations or rotations. */
get codedWidth() {
return this._backing.getCodedWidth();
const value = this._backing.getCodedWidth();
assert(value !== undefined);
return value;
}
/** The height in pixels of the track's coded samples, before any transformations or rotations. */
get codedHeight() {
return this._backing.getCodedHeight();
const value = this._backing.getCodedHeight();
assert(value !== undefined);
return value;
}
/** The angle in degrees by which the track's frames should be rotated (clockwise). */
get rotation() {
return this._backing.getRotation();
const value = this._backing.getRotation();
assert(value !== undefined);
return value;
}
/** The width of the track's frames in square pixels, adjusted for pixel aspect ratio but before rotation. */
get squarePixelWidth() {
return this._backing.getSquarePixelWidth();
const value = this._backing.getSquarePixelWidth();
assert(value !== undefined);
return value;
}
/** The height of the track's frames in square pixels, adjusted for pixel aspect ratio but before rotation. */
get squarePixelHeight() {
return this._backing.getSquarePixelHeight();
const value = this._backing.getSquarePixelHeight();
assert(value !== undefined);
return value;
}
/**
@@ -470,39 +472,23 @@ export class InputVideoTrack extends InputTrack {
*/
get pixelAspectRatio() {
return this._pixelAspectRatioCache ??= simplifyRational({
num: this._backing.getSquarePixelWidth() * this._backing.getCodedHeight(),
den: this._backing.getSquarePixelHeight() * this._backing.getCodedWidth(),
num: this.squarePixelWidth * this.codedHeight,
den: this.squarePixelHeight * this.codedWidth,
});
}
/** The display width of the track's frames in pixels, after aspect ratio adjustment and rotation. */
get displayWidth() {
const customValue = this._backing.getDisplayWidth?.() ?? null;
if (customValue !== null) {
return customValue;
}
const rotation = this._backing.getRotation();
return rotation % 180 === 0 ? this.squarePixelWidth : this.squarePixelHeight;
return this.rotation % 180 === 0 ? this.squarePixelWidth : this.squarePixelHeight;
}
/** The display height of the track's frames in pixels, after aspect ratio adjustment and rotation. */
get displayHeight() {
const customValue = this._backing.getDisplayHeight?.() ?? null;
if (customValue !== null) {
return customValue;
}
const rotation = this._backing.getRotation();
return rotation % 180 === 0 ? this.squarePixelHeight : this.squarePixelWidth;
return this.rotation % 180 === 0 ? this.squarePixelHeight : this.squarePixelWidth;
}
/** Returns the color space of the track's samples. */
async getColorSpace() {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.getColorSpace();
}
@@ -517,10 +503,6 @@ export class InputVideoTrack extends InputTrack {
/** Checks if this track may contain transparent samples with alpha data. */
async canBeTransparent() {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.canBeTransparent();
}
@@ -530,32 +512,16 @@ export class InputVideoTrack extends InputTrack {
* null if the track's codec is unknown.
*/
async getDecoderConfig() {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.getDecoderConfig();
}
async getCodecParameterString() {
if (this._backing.getCodecParameterString) {
return this._backing.getCodecParameterString();
}
if (!this.isHydrated) {
await this.hydrate();
}
const decoderConfig = await this._backing.getDecoderConfig();
return decoderConfig?.codec ?? null;
}
async canDecode() {
try {
if (!this.isHydrated) {
await this.hydrate();
}
const decoderConfig = await this._backing.getDecoderConfig();
if (!decoderConfig) {
return false;
@@ -600,11 +566,11 @@ export class InputVideoTrack extends InputTrack {
}
export interface InputAudioTrackBacking extends InputTrackBacking {
getType(): 'audio';
getCodec(): AudioCodec | null;
getNumberOfChannels(): number;
getSampleRate(): number;
getNumberOfChannels(): number | undefined;
getSampleRate(): number | undefined;
getDecoderConfig(): Promise<AudioDecoderConfig | null>;
getCodecParameterString?(): Promise<string | null>;
}
/**
@@ -637,12 +603,16 @@ export class InputAudioTrack extends InputTrack {
/** The number of audio channels in the track. */
get numberOfChannels() {
return this._backing.getNumberOfChannels();
const value = this._backing.getNumberOfChannels();
assert(value !== undefined);
return value;
}
/** The track's audio sample rate in hertz. */
get sampleRate() {
return this._backing.getSampleRate();
const value = this._backing.getSampleRate();
assert(value !== undefined);
return value;
}
/**
@@ -651,32 +621,16 @@ export class InputAudioTrack extends InputTrack {
* null if the track's codec is unknown.
*/
async getDecoderConfig() {
if (!this.isHydrated) {
await this.hydrate();
}
return this._backing.getDecoderConfig();
}
async getCodecParameterString() {
if (this._backing.getCodecParameterString) {
return this._backing.getCodecParameterString();
}
if (!this.isHydrated) {
await this.hydrate();
}
const decoderConfig = await this._backing.getDecoderConfig();
return decoderConfig?.codec ?? null;
}
async canDecode() {
try {
if (!this.isHydrated) {
await this.hydrate();
}
const decoderConfig = await this._backing.getDecoderConfig();
if (!decoderConfig) {
return false;
@@ -717,112 +671,3 @@ export class InputAudioTrack extends InputTrack {
return 'key'; // No audio codec with delta packets
}
}
export class TrackNotHydratedError extends Error {
/** Creates a new {@link InputDisposedError}. */
constructor(
message = 'InputTrack is not hydrated; please call hydrate() first, or use the resolve() or get() methods.',
) {
super(message);
this.name = 'TrackNotHydratedError';
}
}
export type TrackQuery<T extends InputTrack> = {
filter?: (track: T) => MaybePromise<boolean>;
sortBy?: (track: T) => MaybePromise<number | number[]>;
};
export const mergeTrackQueries = <T extends InputTrack>(
queryA: TrackQuery<T> | undefined,
queryB: TrackQuery<T> | undefined,
): TrackQuery<T> => {
return {
filter: queryA?.filter || queryB?.filter
? (track) => {
const resultA = queryA?.filter?.(track) ?? true;
const handleResultA = (resultA: boolean) => {
if (resultA === false) {
return false;
}
return queryB?.filter?.(track) ?? true;
};
if (resultA instanceof Promise) {
return resultA.then(handleResultA);
} else {
return handleResultA(resultA);
}
}
: undefined,
sortBy: queryA?.sortBy || queryB?.sortBy
? (track) => {
const resultA = queryA?.sortBy?.(track) ?? [];
const resultB = queryB?.sortBy?.(track) ?? [];
type Result = Awaited<typeof resultA>;
const join = (resultA: Result, resultB: Result) => {
return [
...(Array.isArray(resultA) ? resultA : [resultA]),
...(Array.isArray(resultB) ? resultB : [resultB]),
];
};
if (resultA instanceof Promise || resultB instanceof Promise) {
return Promise.all([resultA, resultB]).then(([resultA, resultB]) => {
return join(resultA, resultB);
});
} else {
return join(resultA, resultB);
}
}
: undefined,
};
};
export const queryTracks = async <T extends InputTrack>(tracks: T[], query?: TrackQuery<T>): Promise<T[]> => {
let matchedTracks = tracks;
if (query?.filter) {
const filterMatches = tracks.map(track => query.filter!(track));
const hasAsyncFilter = filterMatches.some(x => x instanceof Promise);
if (hasAsyncFilter) {
// eslint-disable-next-line @typescript-eslint/await-thenable
const resolvedFilterMatches = await Promise.all(filterMatches);
matchedTracks = tracks.filter((_, i) => resolvedFilterMatches[i]);
} else {
matchedTracks = tracks.filter((_, i) => filterMatches[i] as boolean);
}
}
if (!query?.sortBy) {
return matchedTracks;
}
const sortValues = matchedTracks.map(track => query.sortBy!(track));
const hasAsyncSort = sortValues.some(x => x instanceof Promise);
const resolvedSortValues = hasAsyncSort
// eslint-disable-next-line @typescript-eslint/await-thenable
? await Promise.all(sortValues)
: sortValues as (number | number[])[];
return matchedTracks
.map((track, i) => ({ track, sortValue: resolvedSortValues[i] }))
.sort((a, b) => {
const aValues = Array.isArray(a.sortValue) ? a.sortValue : [a.sortValue];
const bValues = Array.isArray(b.sortValue) ? b.sortValue : [b.sortValue];
const maxLength = Math.max(aValues.length, bValues.length);
for (let i = 0; i < maxLength; i++) {
const aValue = aValues[i] ?? 0;
const bValue = bValues[i] ?? 0;
if (aValue === bValue) {
continue;
}
return aValue - bValue;
}
return 0;
})
.map(x => x.track);
};