ADOFAIRenderer/Util/Audio/Mixer/VoiceMixer.cs

using System;
using System.Collections.Generic;
using System.Threading;
using ADOFAIRenderer.Util.Audio.Clip;

namespace ADOFAIRenderer.Util.Audio.Mixer;

/// <summary>
/// Mixes every <see cref="Voice"/> into interleaved double samples at the output sample rate and channel count.
///
/// Time: output sample n is at n / OutputRate seconds relative to the render start (the same time base as the times passed
/// to <see cref="Play"/> and the Voice Set methods). A time t maps to sample round(t * OutputRate).
///
/// Threads:
/// - Main thread: <see cref="Play"/> and the Voice Set methods queue commands under a lock. Commands must be queued in
///   non-decreasing time order per voice.
/// - Render thread: <see cref="Mix"/> takes the queued commands and mixes a block of samples. A command takes effect at its sample;
///   a command whose sample is before the block being mixed takes effect at the start of that block.
///   Chunks are read through <see cref="AudioClipStore.Acquire"/>, which waits for the decoder thread.
///
/// Sample reading: clip positions are in frames of the source at the output rate (sources are resampled when loaded).
/// An integer position copies the stored sample; a fractional position interpolates linearly between the frame and the next one
/// (the next frame of a looping voice after the last frame is frame 0; after the end of a non-looping clip it is 0).
/// Channel mapping follows <see cref="ChunkView.GetSample"/>.
/// </summary>
internal sealed class VoiceMixer : IDisposable
{
    private enum CommandKind
    {
        Play,
        Change
    }

    private readonly struct Command
    {
        public readonly CommandKind Kind;
        public readonly Voice Voice;
        public readonly Voice.Change Change;

        public Command(CommandKind kind, Voice voice, Voice.Change change)
        {
            Kind = kind;
            Voice = voice;
            Change = change;
        }
    }

    private readonly AudioClipStore _store;
    private readonly CancellationToken _cancellationToken;

    private readonly object _commandLock = new();
    private List<Command> _commands = new();
    private List<Command> _takenCommands = new();

    // Render thread only.
    private readonly List<Voice> _voices = new();
    private bool _disposed;

    /// <param name="store">Store the voices' sources belong to; its OutputRate is the mixer's sample rate.</param>
    /// <param name="outputChannels">Interleaved channel count of the mixed output.</param>
    /// <param name="cancellationToken">Cancels waits for chunks in <see cref="Mix"/>.</param>
    public VoiceMixer(AudioClipStore store, int outputChannels, CancellationToken cancellationToken = default)
    {
        if (outputChannels < 1) throw new ArgumentOutOfRangeException(nameof(outputChannels));
        _store = store;
        OutputRate = store.OutputRate;
        OutputChannels = outputChannels;
        _cancellationToken = cancellationToken;
    }

    public int OutputRate { get; }

    public int OutputChannels { get; }

    /// <summary>Voices currently mixed (started or waiting to start). Render thread only.</summary>
    public int ActiveVoiceCount => _voices.Count;

    /// <summary>
    /// Output sample containing <paramref name="seconds"/> (seconds × rate rounded down), the sample at which
    /// AudioSource.PlayScheduled starts a sound scheduled between two output samples.
    /// </summary>
    public static long ToSample(double seconds, int rate) => (long)Math.Floor(seconds * rate);

    // ── Main thread ──

    /// <summary>
    /// Schedules <paramref name="source"/> to start at <paramref name="startTime"/>.
    /// A start time before <paramref name="time"/> (the current time) starts the voice at <paramref name="time"/>.
    /// </summary>
    /// <param name="source">Clip to play; null plays silence.</param>
    /// <param name="name">Name for logs.</param>
    /// <param name="startTime">Start of playback, seconds relative to the render start.</param>
    /// <param name="volume">Linear gain.</param>
    /// <param name="pitch">Playback rate (1 = original speed); negative values are treated as 0.</param>
    /// <param name="loop">Repeats the clip until the voice is stopped.</param>
    /// <param name="time">Current time, seconds relative to the render start.</param>
    public Voice Play(AudioClipSource? source, string name, double startTime, double volume, double pitch, bool loop, double time)
    {
        var voice = new Voice(this, source, name)
        {
            StartSample = Math.Max(ToSample(startTime, OutputRate), ToSample(time, OutputRate)),
            Volume = volume,
            Pitch = Math.Max(0.0, pitch),
            Loop = loop,
            Period = source != null ? source.Duration * OutputRate : 0.0
        };
        voice.AnchorSample = voice.StartSample;

        lock (_commandLock)
        {
            _commands.Add(new Command(CommandKind.Play, voice, default));
        }
        return voice;
    }

    internal void Enqueue(Voice voice, Voice.ChangeKind kind, double time, double value)
    {
        var change = new Voice.Change(kind, ToSample(time, OutputRate), value);
        lock (_commandLock)
        {
            _commands.Add(new Command(CommandKind.Change, voice, change));
        }
    }

    // ── Render thread ──

    /// <summary>
    /// Adds the mix of output samples [<paramref name="firstSample"/>, firstSample + frames) to <paramref name="destination"/>
    /// (interleaved, frames = destination.Length / OutputChannels). Blocks are expected in increasing sample order.
    /// </summary>
    /// <exception cref="InvalidOperationException">Chunk loading failed fatally (see <see cref="AudioClipStore.Acquire"/>).</exception>
    /// <exception cref="OperationCanceledException">The cancellation token was canceled while waiting for a chunk.</exception>
    public void Mix(long firstSample, Span<double> destination)
    {
        if (_disposed) throw new ObjectDisposedException(nameof(VoiceMixer));
        var frames = destination.Length / OutputChannels;
        var endSample = firstSample + frames;

        TakeCommands();

        for (var i = _voices.Count - 1; i >= 0; i--)
        {
            var voice = _voices[i];
            var sample = firstSample;
            while (!voice.Finished)
            {
                var segmentEnd = endSample;
                if (voice.Pending.Count > 0)
                {
                    var next = voice.Pending.Peek();
                    if (next.Sample <= sample)
                    {
                        voice.Apply(voice.Pending.Dequeue(), sample, OutputRate);
                        continue;
                    }
                    segmentEnd = Math.Min(segmentEnd, next.Sample);
                }

                if (segmentEnd > sample)
                    MixSegment(voice, sample, segmentEnd, destination.Slice((int)(sample - firstSample) * OutputChannels));
                sample = segmentEnd;
                if (sample >= endSample) break;
            }

            if (!voice.Finished && endSample >= voice.EndSample)
                voice.Finished = true;
            if (voice.Finished)
            {
                ReleaseVoice(voice);
                _voices.RemoveAt(i);
            }
        }
    }

    private void TakeCommands()
    {
        lock (_commandLock)
        {
            (_commands, _takenCommands) = (_takenCommands, _commands);
        }

        foreach (var command in _takenCommands)
        {
            var voice = command.Voice;
            if (command.Kind == CommandKind.Play)
            {
                if (voice.Source is { IsFailed: false } source)
                    voice.Cursor = source.AddCursor(0, voice.Loop);
                _voices.Add(voice);
            }
            else if (!voice.Finished)
            {
                voice.Pending.Enqueue(command.Change);
            }
        }
        _takenCommands.Clear();
    }

    /// <summary>Mixes <paramref name="voice"/> over samples [from, to) with its current state into <paramref name="destination"/> (starting at sample from).</summary>
    private void MixSegment(Voice voice, long from, long to, Span<double> destination)
    {
        var start = Math.Max(from, voice.StartSample);
        var end = Math.Min(to, voice.EndSample);
        if (start >= end) return;

        var source = voice.Source;
        if (source == null || source.IsFailed)
        {
            voice.Finished = true;
            return;
        }

        var totalFrames = source.TotalFrames;
        if (voice.Volume == 0.0)
        {
            // Nothing to add; a non-looping voice still ends at the end of the clip.
            if (!voice.Loop && voice.PositionAt(end - 1) >= totalFrames) voice.Finished = true;
            return;
        }

        var channels = OutputChannels;
        var volume = voice.Volume;
        var offset = (int)(start - from) * channels;

        UpdateCursor(voice, (long)voice.WrappedPositionAt(start));

        for (var n = start; n < end; n++, offset += channels)
        {
            var position = voice.WrappedPositionAt(n);
            if (position >= totalFrames)
            {
                if (!voice.Loop)
                {
                    voice.Finished = true;
                    return;
                }
                position = totalFrames - 1;
            }

            var index = (long)position;
            var fraction = position - index;

            if (!EnsureView(voice, index)) return;
            var view = voice.View;

            if (fraction == 0.0)
            {
                for (var c = 0; c < channels; c++)
                    destination[offset + c] += view.GetSample(index, c, channels) * volume;
                continue;
            }

            var nextIndex = index + 1;
            var nextInView = view.Contains(nextIndex);
            for (var c = 0; c < channels; c++)
            {
                double low = view.GetSample(index, c, channels);
                double high = nextInView ? view.GetSample(nextIndex, c, channels) : ReadOutside(voice, nextIndex, c);
                destination[offset + c] += (low + fraction * (high - low)) * volume;
            }
        }
    }

    /// <summary>Sample of frame <paramref name="frame"/> outside the voice's current view (next chunk, loop start, or past the end).</summary>
    private double ReadOutside(Voice voice, long frame, int channel)
    {
        var source = voice.Source!;
        if (frame >= source.TotalFrames)
        {
            if (!voice.Loop) return 0.0;
            frame = 0;
        }

        var view = _store.Acquire(source, frame, _cancellationToken);
        if (!view.IsValid) return 0.0;
        try
        {
            return view.GetSample(frame, channel, OutputChannels);
        }
        finally
        {
            source.Release(view);
        }
    }

    /// <summary>Makes voice.View the chunk containing <paramref name="frame"/>. False when the source cannot provide it (voice finished).</summary>
    private bool EnsureView(Voice voice, long frame)
    {
        if (voice.View.IsValid && voice.View.Contains(frame)) return true;

        var source = voice.Source!;
        if (voice.View.IsValid)
        {
            source.Release(voice.View);
            voice.View = default;
        }

        UpdateCursor(voice, frame);
        voice.View = _store.Acquire(source, frame, _cancellationToken);
        if (voice.View.IsValid) return true;

        voice.Finished = true;
        return false;
    }

    private static void UpdateCursor(Voice voice, long frame)
    {
        var cursor = voice.Cursor;
        if (cursor == null) return;
        if (cursor.Loop != voice.Loop) cursor.Loop = voice.Loop;
        if (cursor.Position != frame) cursor.Position = frame;
    }

    private static void ReleaseVoice(Voice voice)
    {
        var source = voice.Source;
        if (source == null) return;
        if (voice.View.IsValid)
        {
            source.Release(voice.View);
            voice.View = default;
        }
        if (voice.Cursor != null)
        {
            source.RemoveCursor(voice.Cursor);
            voice.Cursor = null;
        }
        voice.Pending.Clear();
    }

    /// <summary>Releases every voice's chunk and cursor. Render thread (or after the render thread stopped).</summary>
    public void Dispose()
    {
        if (_disposed) return;
        _disposed = true;
        foreach (var voice in _voices)
            ReleaseVoice(voice);
        _voices.Clear();
        lock (_commandLock)
        {
            _commands.Clear();
        }
    }
}