roadamp/internal/viz/capture.go

//go:build windows

// Package viz captures the Windows audio loopback via WASAPI and emits
// FFT spectrum data for visualisation in the web frontend.
package viz

import (
	"context"
	"fmt"
	"log"
	"math"
	"math/cmplx"
	"syscall"
	"time"
	"unsafe"

	"golang.org/x/sys/windows"
)

// NumBars is the number of frequency bars emitted per frame.
const NumBars = 64

const (
	fftN = 2048 // FFT window size (power of 2)

	// WASAPI
	audclntShareModeShared      = 0
	audclntStreamFlagsLoopback  = 0x00020000
	audclntBufferFlagsSilent    = 0x2
	bufDuration                 = 1_000_000 // 100 ms in 100-ns units

	// Wave format tags
	waveFormatPCM        = 1
	waveFormatFloat      = 3
	waveFormatExtensibleTag = 0xFFFE
)

// ── GUIDs ─────────────────────────────────────────────────────────────────────

var (
	clsidMMDeviceEnumerator = windows.GUID{
		Data1: 0xBCDE0395, Data2: 0xE52F, Data3: 0x467C,
		Data4: [8]byte{0x8E, 0x3D, 0xC4, 0x57, 0x92, 0x91, 0x69, 0x2E},
	}
	iidIMMDeviceEnumerator = windows.GUID{
		Data1: 0xA95664D2, Data2: 0x9614, Data3: 0x4F35,
		Data4: [8]byte{0xA7, 0x46, 0xDE, 0x8D, 0xB6, 0x36, 0x17, 0xE6},
	}
	iidIAudioClient = windows.GUID{
		Data1: 0x1CB9AD4C, Data2: 0xDBFA, Data3: 0x4c32,
		Data4: [8]byte{0xB1, 0x78, 0xC2, 0xF5, 0x68, 0xA7, 0x03, 0xB2},
	}
	iidIAudioCaptureClient = windows.GUID{
		Data1: 0xC8ADBD64, Data2: 0xE71E, Data3: 0x48a0,
		Data4: [8]byte{0xA4, 0xDE, 0x18, 0x5C, 0x39, 0x5C, 0xD3, 0x17},
	}
	subFormatFloat = windows.GUID{
		Data1: 0x00000003, Data2: 0x0000, Data3: 0x0010,
		Data4: [8]byte{0x80, 0x00, 0x00, 0xAA, 0x00, 0x38, 0x9B, 0x71},
	}
)

// ── WAVEFORMAT structs ────────────────────────────────────────────────────────

type waveFormatEx struct {
	FormatTag      uint16
	Channels       uint16
	SamplesPerSec  uint32
	AvgBytesPerSec uint32
	BlockAlign     uint16
	BitsPerSample  uint16
	Size           uint16
}

// waveFormatExtensibleEx is a flat representation of WAVEFORMATEXTENSIBLE.
// We cannot embed waveFormatEx because Go pads the struct to 20 bytes
// (alignment of largest field = uint32), but the C layout is 18 bytes —
// so SubFormat would land at the wrong offset if we used struct embedding.
type waveFormatExtensibleEx struct {
	FormatTag      uint16
	Channels       uint16
	SamplesPerSec  uint32
	AvgBytesPerSec uint32
	BlockAlign     uint16
	BitsPerSample  uint16
	Size           uint16
	Samples        uint16       // wValidBitsPerSample / wSamplesPerBlock
	ChannelMask    uint32
	SubFormat      windows.GUID // 16 bytes → total 40 bytes, matches C layout
}

// ── DLL procs ─────────────────────────────────────────────────────────────────

var (
	ole32            = windows.NewLazySystemDLL("ole32.dll")
	coInitializeEx   = ole32.NewProc("CoInitializeEx")
	coUninitialize   = ole32.NewProc("CoUninitialize")
	coCreateInstance = ole32.NewProc("CoCreateInstance")
	coTaskMemFree    = ole32.NewProc("CoTaskMemFree")
)

// ── COM vtable helpers ────────────────────────────────────────────────────────

var ptrSize = unsafe.Sizeof(uintptr(0))

func procAt(comObj uintptr, methodIdx int) uintptr {
	vtbl := *(*uintptr)(unsafe.Pointer(comObj))
	return *(*uintptr)(unsafe.Pointer(vtbl + uintptr(methodIdx)*ptrSize))
}

func comRelease(p uintptr) {
	if p != 0 {
		syscall.Syscall(procAt(p, 2), 1, p, 0, 0)
	}
}

// ── Capturer ──────────────────────────────────────────────────────────────────

// Capturer streams FFT spectrum bars from the system audio loopback.
type Capturer struct {
	// C receives slices of NumBars float32 values in [0.0, 1.0] at ~30 fps.
	// Slow consumers cause frames to be dropped (non-blocking send).
	C chan []float32
}

// NewCapturer creates a Capturer ready to Start.
func NewCapturer() *Capturer {
	return &Capturer{C: make(chan []float32, 4)}
}

// Start begins the capture loop; blocks until ctx is cancelled.
// Errors are logged but never fatal — the channel simply stays empty.
func (c *Capturer) Start(ctx context.Context) {
	if err := c.run(ctx); err != nil {
		log.Printf("viz: %v", err)
	}
}

func (c *Capturer) run(ctx context.Context) error {
	coInitializeEx.Call(0, 0) // COINIT_MULTITHREADED
	defer coUninitialize.Call()

	// ── IMMDeviceEnumerator ──────────────────────────────────────────────────
	var enumerator uintptr
	if hr, _, _ := coCreateInstance.Call(
		uintptr(unsafe.Pointer(&clsidMMDeviceEnumerator)), 0, 0x17,
		uintptr(unsafe.Pointer(&iidIMMDeviceEnumerator)),
		uintptr(unsafe.Pointer(&enumerator)),
	); hr != 0 {
		return fmt.Errorf("CoCreateInstance(MMDeviceEnumerator): 0x%08X", hr)
	}
	defer comRelease(enumerator)

	// ── Default render device ────────────────────────────────────────────────
	// GetDefaultAudioEndpoint(eRender, eConsole, &device) — vtable index 4, 4 args
	var device uintptr
	if hr, _, _ := syscall.Syscall6(
		procAt(enumerator, 4), 4,
		enumerator, 0, 0, uintptr(unsafe.Pointer(&device)), 0, 0,
	); hr != 0 {
		return fmt.Errorf("GetDefaultAudioEndpoint: 0x%08X", hr)
	}
	defer comRelease(device)

	// ── IAudioClient ────────────────────────────────────────────────────────
	// IMMDevice::Activate(riid, clsCtx, pParams, &ppv) — vtable index 3, 5 args
	var ac uintptr
	if hr, _, _ := syscall.Syscall6(
		procAt(device, 3), 5,
		device, uintptr(unsafe.Pointer(&iidIAudioClient)), 0x17, 0,
		uintptr(unsafe.Pointer(&ac)), 0,
	); hr != 0 {
		return fmt.Errorf("Activate(IAudioClient): 0x%08X", hr)
	}
	defer comRelease(ac)

	// ── Mix format ──────────────────────────────────────────────────────────
	var fmtPtr uintptr
	if hr, _, _ := syscall.Syscall(
		procAt(ac, 8), 2, // GetMixFormat
		ac, uintptr(unsafe.Pointer(&fmtPtr)), 0,
	); hr != 0 {
		return fmt.Errorf("GetMixFormat: 0x%08X", hr)
	}
	defer coTaskMemFree.Call(fmtPtr)

	wfx := (*waveFormatEx)(unsafe.Pointer(fmtPtr))
	sampleRate := int(wfx.SamplesPerSec)
	channels := int(wfx.Channels)
	isFloat := wfx.FormatTag == waveFormatFloat
	if wfx.FormatTag == waveFormatExtensibleTag && wfx.Size >= 22 {
		ext := (*waveFormatExtensibleEx)(unsafe.Pointer(fmtPtr))
		isFloat = ext.SubFormat == subFormatFloat
	}
	log.Printf("viz: loopback format %d Hz, %d ch, %d bit, float=%v",
		sampleRate, channels, wfx.BitsPerSample, isFloat)

	if !isFloat || wfx.BitsPerSample != 32 {
		return fmt.Errorf("viz: unsupported format (need float32); got tag=%04X bits=%d",
			wfx.FormatTag, wfx.BitsPerSample)
	}

	// ── Initialize loopback ──────────────────────────────────────────────────
	if hr, _, _ := syscall.Syscall9(
		procAt(ac, 3), 7, // IAudioClient::Initialize
		ac,
		audclntShareModeShared,
		audclntStreamFlagsLoopback,
		uintptr(bufDuration), 0, // hnsBufferDuration, hnsPeriodicity
		fmtPtr, 0, // pFormat, AudioSessionGuid
		0, 0,
	); hr != 0 {
		return fmt.Errorf("IAudioClient::Initialize: 0x%08X", hr)
	}

	// ── IAudioCaptureClient ──────────────────────────────────────────────────
	var acc uintptr
	if hr, _, _ := syscall.Syscall(
		procAt(ac, 14), 3, // GetService
		ac,
		uintptr(unsafe.Pointer(&iidIAudioCaptureClient)),
		uintptr(unsafe.Pointer(&acc)),
	); hr != 0 {
		return fmt.Errorf("GetService(IAudioCaptureClient): 0x%08X", hr)
	}
	defer comRelease(acc)

	// ── Start ────────────────────────────────────────────────────────────────
	if hr, _, _ := syscall.Syscall(procAt(ac, 10), 1, ac, 0, 0); hr != 0 {
		return fmt.Errorf("IAudioClient::Start: 0x%08X", hr)
	}
	defer syscall.Syscall(procAt(ac, 11), 1, ac, 0, 0) // Stop

	// ── Capture loop ─────────────────────────────────────────────────────────
	buf := make([]float64, 0, fftN*2)
	smooth := make([]float32, NumBars)
	tick := time.NewTicker(10 * time.Millisecond)
	defer tick.Stop()

	for {
		select {
		case <-ctx.Done():
			return nil
		case <-tick.C:
			buf = drainLoopback(acc, channels, buf)
			for len(buf) >= fftN {
				bars := spectrum(buf[:fftN], sampleRate, smooth)
				copy(smooth, bars)
				select {
				case c.C <- bars:
				default:
				}
				buf = buf[fftN:]
			}
		}
	}
}

// drainLoopback reads all pending audio frames into buf and returns it.
func drainLoopback(acc uintptr, channels int, buf []float64) []float64 {
	for {
		// GetNextPacketSize
		var packetFrames uint32
		if hr, _, _ := syscall.Syscall(
			procAt(acc, 5), 2,
			acc, uintptr(unsafe.Pointer(&packetFrames)), 0,
		); hr != 0 || packetFrames == 0 {
			break
		}

		// GetBuffer(ppData, &numFrames, &flags, NULL, NULL) — 6 args
		var dataPtr uintptr
		var numFrames uint32
		var flags uint32
		if hr, _, _ := syscall.Syscall6(
			procAt(acc, 3), 6,
			acc,
			uintptr(unsafe.Pointer(&dataPtr)),
			uintptr(unsafe.Pointer(&numFrames)),
			uintptr(unsafe.Pointer(&flags)),
			0, 0,
		); hr != 0 {
			break
		}

		if flags&audclntBufferFlagsSilent == 0 && dataPtr != 0 && numFrames > 0 {
			samples := unsafe.Slice((*float32)(unsafe.Pointer(dataPtr)), int(numFrames)*channels)
			for i := 0; i < int(numFrames); i++ {
				var mono float64
				for ch := 0; ch < channels; ch++ {
					mono += float64(samples[i*channels+ch])
				}
				buf = append(buf, mono/float64(channels))
			}
		}

		// ReleaseBuffer
		syscall.Syscall(procAt(acc, 4), 2, acc, uintptr(numFrames), 0)
	}
	return buf
}

// ── Spectrum analysis ─────────────────────────────────────────────────────────

// spectrum applies a Hanning window, runs the FFT, maps to NumBars
// log-spaced frequency bins, and applies fast-attack/slow-decay smoothing.
func spectrum(samples []float64, sampleRate int, prev []float32) []float32 {
	n := len(samples)

	// Hanning window
	cx := make([]complex128, n)
	for i, s := range samples {
		w := 0.5 * (1 - math.Cos(2*math.Pi*float64(i)/float64(n-1)))
		cx[i] = complex(s*w, 0)
	}

	ditFFT(cx)

	// Magnitude of positive frequencies, normalised
	bins := make([]float64, n/2)
	scale := 2.0 / float64(n)
	for i := range bins {
		bins[i] = cmplx.Abs(cx[i]) * scale
	}

	// Log-spaced output bars: 40 Hz → 20 kHz
	const fMin, fMax = 40.0, 20_000.0
	freqRes := float64(sampleRate) / float64(n)
	bars := make([]float32, NumBars)

	for b := 0; b < NumBars; b++ {
		t := float64(b) / float64(NumBars-1)
		f := fMin * math.Pow(fMax/fMin, t)

		var fNext float64
		if b < NumBars-1 {
			t2 := float64(b+1) / float64(NumBars-1)
			fNext = fMin * math.Pow(fMax/fMin, t2)
		} else {
			fNext = fMax
		}

		lo := clamp(int(f/freqRes), 0, len(bins)-1)
		hi := clamp(int(fNext/freqRes), lo+1, len(bins))

		var sum float64
		for i := lo; i < hi; i++ {
			sum += bins[i]
		}
		avg := sum / float64(hi-lo)

		// dB → [0, 1]
		dB := 20 * math.Log10(avg+1e-9)
		norm := float32((dB + 80) / 80)
		if norm < 0 {
			norm = 0
		}
		if norm > 1 {
			norm = 1
		}

		// Fast attack, slow decay
		if norm > prev[b] {
			bars[b] = norm
		} else {
			bars[b] = prev[b] * 0.88
		}
	}
	return bars
}

func clamp(v, lo, hi int) int {
	if v < lo {
		return lo
	}
	if v > hi {
		return hi
	}
	return v
}

// ── Cooley-Tukey FFT ─────────────────────────────────────────────────────────

// ditFFT is an in-place, decimation-in-time FFT. len(x) must be a power of 2.
func ditFFT(x []complex128) {
	n := len(x)
	// Bit-reversal permutation
	j := 0
	for i := 1; i < n; i++ {
		bit := n >> 1
		for j&bit != 0 {
			j ^= bit
			bit >>= 1
		}
		j ^= bit
		if i < j {
			x[i], x[j] = x[j], x[i]
		}
	}
	// Butterfly stages
	for length := 2; length <= n; length <<= 1 {
		half := length >> 1
		wStep := cmplx.Exp(complex(0, -math.Pi/float64(half)))
		for i := 0; i < n; i += length {
			w := complex(1, 0)
			for k := 0; k < half; k++ {
				u := x[i+k]
				v := x[i+k+half] * w
				x[i+k] = u + v
				x[i+k+half] = u - v
				w *= wStep
			}
		}
	}
}