LlamaCasty/Services/AlertClipDecoder.cs

using System.Diagnostics;
using ytLive.Services.Audio;
using ytLive.Services.Encoder;

namespace ytLive.Services;

/// <summary>
/// TASK 47: one alert clip's playback — video frames + on-stream audio from the
/// same file, both paced to real time. The alert box owns a decoder per event
/// (start at play, dispose at drain) — unlike MediaSource's app-wide refcounted
/// session, an alert REPLAYS the same clip each event, so a fresh decode per
/// play is required. Video is rawvideo BGRA scaled to the box rect (the
/// <see cref="MediaVideoSource"/> shape); audio is ffmpeg <c>f32le</c> at the
/// mixer's native 48 kHz stereo (so the mixer ring needs no resampler). The
/// fade is applied by the alert layer, not here.
/// </summary>
public interface IAlertClipDecoder : IDisposable
{
    /// <summary>Starts the decode loop (video + audio); raises the events off the
    /// calling thread. Paced to real time so playback matches wall clock.</summary>
    void Start();

    /// <summary>Stops and kills any running decode subprocesses.</summary>
    void Stop();

    /// <summary>One decoded video frame (box-sized, straight-alpha BGRA).</summary>
    event Action<VideoFrame>? FrameAvailable;

    /// <summary>One chunk of on-stream audio (48 kHz stereo float).</summary>
    event Action<AudioSample>? AudioReady;

    /// <summary>Raised once when both streams have fully played out.</summary>
    event Action? Completed;
}

/// <summary>Default <see cref="IAlertClipDecoder"/>: two ffmpeg subprocesses on
/// the same file. Video via <see cref="RawVideoFrameReader"/> on a <c>rawvideo
/// bgra</c> pipe (-an), audio via an <c>f32le -ac 2 -ar 48000</c> pipe (-vn).
/// The seams (locator, process factory, frame-rate probe, delay) let tests drive
/// a fake decoder without a real codec or Windows.</summary>
public sealed class AlertClipDecoder : IAlertClipDecoder
{
    public event Action<VideoFrame>? FrameAvailable;
    public event Action<AudioSample>? AudioReady;
    public event Action? Completed;

    private const int AudioSampleRate = 48000;
    private const int AudioChannels = 2;
    private const int AudioChunkFloats = 4800; // 50ms of stereo

    private readonly string _path;
    private readonly int _width;
    private readonly int _height;
    private readonly IFfmpegLocator _locator;
    private readonly Func<IDecodeProcess> _processFactory;
    private readonly IFrameRateProbe? _probe;
    private readonly Func<TimeSpan, CancellationToken, Task> _delay;
    private readonly CancellationTokenSource _cts = new();
    private Task? _run;

    public AlertClipDecoder(
        string path,
        int width,
        int height,
        IFfmpegLocator locator,
        Func<IDecodeProcess> processFactory,
        IFrameRateProbe? frameRateProbe = null,
        Func<TimeSpan, CancellationToken, Task>? delay = null)
    {
        _path = path;
        _width = width;
        _height = height;
        _locator = locator;
        _processFactory = processFactory;
        _probe = frameRateProbe;
        _delay = delay ?? ((t, ct) => Task.Delay(t, ct));
    }

    public void Start()
    {
        if (_run != null) return;
        _run = Task.Run(RunAsync, CancellationToken.None);
    }

    public void Stop() => _cts.Cancel();

    public void Dispose()
    {
        _cts.Cancel();
        _cts.Dispose();
    }

    private async Task RunAsync()
    {
        var ct = _cts.Token;
        try
        {
            double? frameRate = null;
            if (_probe != null)
                frameRate = await _probe.ProbeAsync(_path, ct).ConfigureAwait(false);
            var frameDuration = frameRate is > 0 ? TimeSpan.FromSeconds(1.0 / frameRate.Value) : TimeSpan.Zero;

            var ffmpegPath = await _locator.LocateAsync(ct).ConfigureAwait(false);

            var videoTask = RunVideoAsync(ffmpegPath, frameDuration, ct);
            var audioTask = RunAudioAsync(ffmpegPath, ct);
            await Task.WhenAll(videoTask, audioTask).ConfigureAwait(false);
        }
        catch (OperationCanceledException)
        {
            return;
        }
        catch
        {
            // A failed decode is not fatal: the alert layer falls back to the six
            // animations (or the frame it already holds) exactly as a missing file.
        }

        if (!ct.IsCancellationRequested)
            Completed?.Invoke();
    }

    private async Task RunVideoAsync(string ffmpegPath, TimeSpan frameDuration, CancellationToken ct)
    {
        var args =
            $"-hide_banner -loglevel error -i \"{_path}\" " +
            $"-f rawvideo -pix_fmt bgra -vf scale={_width}:{_height} -an -";
        using var process = _processFactory();
        process.Start(new ProcessStartInfo
        {
            FileName = ffmpegPath,
            Arguments = args,
            UseShellExecute = false,
            RedirectStandardOutput = true,
            RedirectStandardError = true,
            CreateNoWindow = true,
        });

        var reader = new RawVideoFrameReader(_width, _height);
        var buffer = new byte[65536];
        var stream = process.StandardOutput;
        while (!ct.IsCancellationRequested && !process.HasExited)
        {
            var read = await stream.ReadAsync(buffer, 0, buffer.Length, ct).ConfigureAwait(false);
            if (read <= 0) break;
            var chunk = new byte[read];
            Buffer.BlockCopy(buffer, 0, chunk, 0, read);
            foreach (var frame in reader.Feed(chunk))
            {
                FrameAvailable?.Invoke(frame);
                if (frameDuration > TimeSpan.Zero)
                    await _delay(frameDuration, ct).ConfigureAwait(false);
            }
        }
    }

    private async Task RunAudioAsync(string ffmpegPath, CancellationToken ct)
    {
        var args =
            $"-hide_banner -loglevel error -i \"{_path}\" " +
            $"-f f32le -ac {AudioChannels} -ar {AudioSampleRate} -vn -";
        using var process = _processFactory();
        process.Start(new ProcessStartInfo
        {
            FileName = ffmpegPath,
            Arguments = args,
            UseShellExecute = false,
            RedirectStandardOutput = true,
            RedirectStandardError = true,
            CreateNoWindow = true,
        });

        var stream = process.StandardOutput;
        var raw = new byte[65536];
        var carry = new byte[0];
        var floatScratch = new float[AudioChunkFloats];
        var floatIndex = 0;
        // The pipe is byte-delimited only: a read may split a float, so leftover
        // bytes are carried into the next read instead of dropped (a dropped
        // half-float would desync the entire stream).
        while (!ct.IsCancellationRequested && !process.HasExited)
        {
            var read = await stream.ReadAsync(raw, 0, raw.Length, ct).ConfigureAwait(false);
            if (read <= 0) break;
            var total = carry.Length + read;
            var bytes = total <= raw.Length ? raw : new byte[total];
            if (carry.Length > 0)
            {
                Buffer.BlockCopy(carry, 0, bytes, 0, carry.Length);
                Buffer.BlockCopy(raw, 0, bytes, carry.Length, read);
            }
            var floatsTo = total - (total % 4);
            for (var i = 0; i < floatsTo && floatIndex < floatScratch.Length; i += 4)
                floatScratch[floatIndex++] = BitConverter.ToSingle(bytes, i);
            carry = total > floatsTo ? bytes[floatsTo..total] : new byte[0];

            while (floatIndex == floatScratch.Length)
            {
                AudioReady?.Invoke(new AudioSample(floatScratch, AudioSampleRate, AudioChannels));
                // Pace the NEXT chunk so the whole clip's audio lands at real-time
                // cadence (the fade in the alert layer tracks the same wall clock).
                await _delay(TimeSpan.FromSeconds(AudioChunkFloats / (double)(AudioSampleRate * AudioChannels)), ct).ConfigureAwait(false);
                floatScratch = new float[AudioChunkFloats];
                floatIndex = 0;
            }
        }
    }
}