go-ts/pkg/audio/playback.go

package audio

import (
	"encoding/binary"
	"fmt"
	"sync"
	"time"
	"unsafe"

	"github.com/go-ole/go-ole"
	"github.com/moutend/go-wca/pkg/wca"
)

// Player handles WASAPI audio playback
type Player struct {
	client       *wca.IAudioClient
	renderClient *wca.IAudioRenderClient
	waveFormat   *wca.WAVEFORMATEX
	bufferSize   uint32
	volume       float32
	muted        bool
	mu           sync.Mutex
	running      bool
	stopChan     chan struct{}

	// Audio buffer - accumulates incoming audio
	audioBuffer []int16
	bufferMu    sync.Mutex

	// Frame queue (960 samples = 20ms at 48kHz)
	frameQueue chan []int16
}

const frameSamples = 960 // 20ms at 48kHz

// NewPlayer creates a new WASAPI audio player
func NewPlayer() (*Player, error) {
	// Initialize COM using go-ole
	if err := ole.CoInitializeEx(0, ole.COINIT_APARTMENTTHREADED); err != nil {
		// Ignore if already initialized
	}

	// Get default audio endpoint
	var deviceEnumerator *wca.IMMDeviceEnumerator
	if err := wca.CoCreateInstance(
		wca.CLSID_MMDeviceEnumerator,
		0,
		wca.CLSCTX_ALL,
		wca.IID_IMMDeviceEnumerator,
		&deviceEnumerator,
	); err != nil {
		return nil, fmt.Errorf("failed to create device enumerator: %w", err)
	}
	defer deviceEnumerator.Release()

	var device *wca.IMMDevice
	if err := deviceEnumerator.GetDefaultAudioEndpoint(wca.ERender, wca.EConsole, &device); err != nil {
		return nil, fmt.Errorf("failed to get default render device: %w", err)
	}
	defer device.Release()

	// Activate audio client
	var audioClient *wca.IAudioClient
	if err := device.Activate(wca.IID_IAudioClient, wca.CLSCTX_ALL, nil, &audioClient); err != nil {
		return nil, fmt.Errorf("failed to activate audio client: %w", err)
	}

	// Set up format for 48kHz mono 16-bit (TeamSpeak format)
	waveFormat := &wca.WAVEFORMATEX{
		WFormatTag:      wca.WAVE_FORMAT_PCM,
		NChannels:       1,
		NSamplesPerSec:  48000,
		WBitsPerSample:  16,
		NBlockAlign:     2,
		NAvgBytesPerSec: 96000,
		CbSize:          0,
	}

	// Initialize in shared mode - 100ms buffer
	duration := wca.REFERENCE_TIME(100 * 10000) // 100ms in 100-nanosecond units
	if err := audioClient.Initialize(
		wca.AUDCLNT_SHAREMODE_SHARED,
		wca.AUDCLNT_STREAMFLAGS_AUTOCONVERTPCM|wca.AUDCLNT_STREAMFLAGS_SRC_DEFAULT_QUALITY,
		duration,
		0,
		waveFormat,
		nil,
	); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to initialize audio client: %w", err)
	}

	// Get buffer size
	var bufferSize uint32
	if err := audioClient.GetBufferSize(&bufferSize); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to get buffer size: %w", err)
	}

	// Get render client
	var renderClient *wca.IAudioRenderClient
	if err := audioClient.GetService(wca.IID_IAudioRenderClient, &renderClient); err != nil {
		audioClient.Release()
		return nil, fmt.Errorf("failed to get render client: %w", err)
	}

	return &Player{
		client:       audioClient,
		renderClient: renderClient,
		waveFormat:   waveFormat,
		bufferSize:   bufferSize,
		volume:       1.0,
		muted:        false,
		stopChan:     make(chan struct{}),
		audioBuffer:  make([]int16, 0, frameSamples*50), // ~1 second buffer
		frameQueue:   make(chan []int16, 100),           // ~2 seconds of frames
	}, nil
}

// Start begins audio playback
func (p *Player) Start() error {
	p.mu.Lock()
	if p.running {
		p.mu.Unlock()
		return nil
	}
	p.running = true
	p.mu.Unlock()

	if err := p.client.Start(); err != nil {
		return fmt.Errorf("failed to start audio client: %w", err)
	}

	// Playback loop writes frames from queue to WASAPI
	go p.playbackLoop()

	return nil
}

// Stop stops audio playback
func (p *Player) Stop() {
	p.mu.Lock()
	if !p.running {
		p.mu.Unlock()
		return
	}
	p.running = false
	p.mu.Unlock()

	close(p.stopChan)
	p.client.Stop()
}

// Close releases all resources
func (p *Player) Close() {
	p.Stop()
	if p.renderClient != nil {
		p.renderClient.Release()
	}
	if p.client != nil {
		p.client.Release()
	}
	ole.CoUninitialize()
}

// PlayPCM queues PCM audio for playback
// Accumulates samples and queues complete 960-sample frames
func (p *Player) PlayPCM(samples []int16) {
	if p.muted {
		return
	}

	// Apply volume
	adjusted := samples
	if p.volume != 1.0 {
		adjusted = make([]int16, len(samples))
		for i, s := range samples {
			adjusted[i] = int16(float32(s) * p.volume)
		}
	}

	p.bufferMu.Lock()
	p.audioBuffer = append(p.audioBuffer, adjusted...)

	// Queue complete 960-sample frames
	for len(p.audioBuffer) >= frameSamples {
		frame := make([]int16, frameSamples)
		copy(frame, p.audioBuffer[:frameSamples])
		p.audioBuffer = p.audioBuffer[frameSamples:]

		select {
		case p.frameQueue <- frame:
		default:
			// Queue full, drop oldest frame
			select {
			case <-p.frameQueue:
			default:
			}
			p.frameQueue <- frame
		}
	}
	p.bufferMu.Unlock()
}

// SetVolume sets playback volume (0.0 to 1.0)
func (p *Player) SetVolume(vol float32) {
	if vol < 0 {
		vol = 0
	}
	if vol > 1.0 {
		vol = 1.0
	}
	p.mu.Lock()
	p.volume = vol
	p.mu.Unlock()
}

// GetVolume returns current volume (0.0 to 1.0)
func (p *Player) GetVolume() float32 {
	p.mu.Lock()
	defer p.mu.Unlock()
	return p.volume
}

// SetMuted sets mute state
func (p *Player) SetMuted(muted bool) {
	p.mu.Lock()
	p.muted = muted
	p.mu.Unlock()
}

// IsMuted returns mute state
func (p *Player) IsMuted() bool {
	p.mu.Lock()
	defer p.mu.Unlock()
	return p.muted
}

func (p *Player) playbackLoop() {
	// Use 20ms ticker matching TeamSpeak frame rate
	ticker := time.NewTicker(20 * time.Millisecond)
	defer ticker.Stop()

	for {
		select {
		case <-p.stopChan:
			return
		case <-ticker.C:
			p.writeFrame()
		}
	}
}

func (p *Player) writeFrame() {
	// Get current padding (samples already in buffer)
	var padding uint32
	if err := p.client.GetCurrentPadding(&padding); err != nil {
		return
	}

	available := p.bufferSize - padding
	if available < frameSamples {
		return // Not enough space for a full frame
	}

	// Try to get a frame from the queue
	select {
	case frame := <-p.frameQueue:
		var buffer *byte
		if err := p.renderClient.GetBuffer(frameSamples, &buffer); err != nil {
			return
		}

		// Write frame to WASAPI buffer
		bufSlice := unsafe.Slice(buffer, frameSamples*2)
		for i := 0; i < frameSamples; i++ {
			binary.LittleEndian.PutUint16(bufSlice[i*2:], uint16(frame[i]))
		}

		p.renderClient.ReleaseBuffer(frameSamples, 0)

	default:
		// No audio available - optionally write silence
		// (skip for now to avoid crackling)
	}
}