talk/internal/audio/oto.go

package audio

import (
	"bytes"
	"context"
	"fmt"
	"sync"
	"time"

	"github.com/ebitengine/oto/v3"
)

// OtoSink plays audio embedded via oto (github.com/ebitengine/oto/v3,
// pinned to v3.5.0); no external player needed. oto loads the audio drivers
// at runtime without CGO (purego), which keeps talk cross-compilable.
type OtoSink struct{}

// Name implements Sink.
func (OtoSink) Name() string { return NameOto }

// contextCache keeps one oto context per PCM format: oto supports only ONE
// context per process ("Creating multiple contexts is NOT supported"), and
// a context is bound to exactly one format. The CLI plays once per process;
// the daemon benefits from the cache.
var contextCache = struct {
	sync.Mutex
	byFormat map[pcmFormat]*otoContext
}{byFormat: map[pcmFormat]*otoContext{}}

type otoContext struct {
	ctx   *oto.Context
	ready chan struct{}
}

// otoFormat maps a PCM format to the oto format.
func (f pcmFormat) otoFormat() oto.Format {
	switch {
	case f.Float:
		return oto.FormatFloat32LE
	case f.BitsPerCoded == 8:
		return oto.FormatUnsignedInt8
	default:
		return oto.FormatSignedInt16LE
	}
}

// contextFor returns the (cached) oto context for a format.
func contextFor(f pcmFormat) (*otoContext, error) {
	contextCache.Lock()
	defer contextCache.Unlock()
	if c, ok := contextCache.byFormat[f]; ok {
		return c, nil
	}
	op := &oto.NewContextOptions{
		SampleRate:   f.SampleRate,
		ChannelCount: f.Channels,
		Format:       f.otoFormat(),
	}
	ctx, ready, err := oto.NewContext(op)
	if err != nil {
		return nil, fmt.Errorf("oto: creating context failed (no audio device?): %w", err)
	}
	c := &otoContext{ctx: ctx, ready: ready}
	contextCache.byFormat[f] = c
	return c, nil
}

// Play implements Sink: decodes the data (RIFF/WAVE PCM from osbuiltin or
// MP3 from the cloud providers, see mp3.go) and plays the samples, blocking.
// Audio device and format errors are reported clearly.
func (OtoSink) Play(ctx context.Context, data []byte) error {
	f, pcm, err := decodeAudio(data)
	if err != nil {
		return err
	}
	c, err := contextFor(f)
	if err != nil {
		return err
	}
	select {
	case <-c.ready:
		if err := c.ctx.Err(); err != nil {
			return fmt.Errorf("oto: initialization failed: %w", err)
		}
	case <-ctx.Done():
		return ctx.Err()
	}

	player := c.ctx.NewPlayer(bytes.NewReader(pcm))
	defer player.Close()
	player.Play()
	for player.IsPlaying() {
		if err := ctx.Err(); err != nil {
			player.Close()
			return err
		}
		if err := player.Err(); err != nil {
			return fmt.Errorf("oto: playback failed: %w", err)
		}
		time.Sleep(5 * time.Millisecond)
	}
	return player.Err()
}