galahad2lcd/internal/transcoder/transcoder.go

// Package transcoder converts a source video or GIF into a raw Annex-B
// H.264 stream using the system ffmpeg/ffprobe binaries. Shelling out to
// ffmpeg avoids the fragile C-ABI bindings that plague in-process encoders
// and lets this driver work with any recent FFmpeg release.
package transcoder

import (
	"context"
	"errors"
	"fmt"
	"os"
	"os/exec"
	"strconv"
	"strings"

	"galahad2lcd/internal/config"
)

// Screen dimensions of the Lian Li Galahad II LCD.
const (
	screenWidth  = 480
	screenHeight = 480
)

// encoderParameters mirrors the reference driver: constrained VBR, low
// latency and a stream that the display firmware accepts.
const (
	bitRate    = 2000 // kbit/s
	preset     = "veryfast"
	profile    = "baseline"
	defaultFPS = 30.0
)

// Transcoder invokes ffmpeg and ffprobe. Binary paths are configurable so
// tests can substitute fakes.
type Transcoder struct {
	ffmpeg  string
	ffprobe string
}

// New returns a Transcoder that resolves ffmpeg and ffprobe from PATH.
func New() (*Transcoder, error) {
	ffmpeg, err := exec.LookPath("ffmpeg")
	if err != nil {
		return nil, fmt.Errorf("ffmpeg not found: %w (install ffmpeg)", err)
	}
	ffprobe, err := exec.LookPath("ffprobe")
	if err != nil {
		return nil, fmt.Errorf("ffprobe not found: %w (install ffmpeg)", err)
	}
	return &Transcoder{ffmpeg: ffmpeg, ffprobe: ffprobe}, nil
}

// NewWithBinaries is intended for tests.
func NewWithBinaries(ffmpeg, ffprobe string) *Transcoder {
	return &Transcoder{ffmpeg: ffmpeg, ffprobe: ffprobe}
}

// Transcode renders the configured media into an Annex-B H.264 cache file
// and returns the effective playback framerate. It is cancellable via ctx.
func (t *Transcoder) Transcode(ctx context.Context, cfg config.Config) (float64, error) {
	if cfg.Display.Input == "" {
		return 0, fmt.Errorf("no input file configured")
	}

	// Remove any stale cache file first: it may be owned by another user
	// (e.g. created by a manual run) which would make the service fail to
	// open it for writing. ffmpeg -y alone cannot fix that.
	if err := os.Remove(cfg.Stream.CachePath); err != nil && !errors.Is(err, os.ErrNotExist) {
		return 0, fmt.Errorf("remove stale cache %q: %w", cfg.Stream.CachePath, err)
	}

	fps := cfg.Stream.FPS
	if fps <= 0 {
		detected, err := t.DetectFPS(ctx, cfg.Display.Input)
		if err != nil {
			return 0, err
		}
		fps = detected
	}
	fps = clampFPS(fps)
	gop := max(1, int(fps+0.5))

	vf := buildFilter(cfg.Display.Rotate, fps)
	args := []string{
		"-y", "-v", "error",
		"-i", cfg.Display.Input,
		"-vf", vf,
		"-c:v", "libx264",
		"-preset", preset,
		"-profile:v", profile,
		"-b:v", strconv.Itoa(bitRate) + "k",
		"-maxrate", strconv.Itoa(bitRate) + "k",
		"-bufsize", strconv.Itoa(bitRate) + "k",
		"-x264-params", fmt.Sprintf(
			"nal-hrd=cbr:annexb=1:open-gop=0:scenecut=0:keyint=%d:min-keyint=%d", gop, gop),
		"-an", "-f", "h264", cfg.Stream.CachePath,
	}

	if err := run(ctx, t.ffmpeg, args...); err != nil {
		return 0, fmt.Errorf("transcode %q: %w", cfg.Display.Input, err)
	}
	return fps, nil
}

// DetectFPS probes the source video stream and returns its average frame
// rate. It falls back to a safe default when the rate cannot be determined.
func (t *Transcoder) DetectFPS(ctx context.Context, input string) (float64, error) {
	out, err := runCapture(ctx, t.ffprobe,
		"-v", "error",
		"-select_streams", "v:0",
		"-show_entries", "stream=avg_frame_rate",
		"-of", "csv=p=0", input)
	if err != nil {
		return 0, fmt.Errorf("probe %q: %w", input, err)
	}

	fps, err := parseFPSRational(strings.TrimSpace(out))
	if err != nil {
		// Streams without a meaningful avg_frame_rate (some GIFs) report 0/0.
		if r, rerr := t.rateFromHeader(ctx, input); rerr == nil && r > 0 {
			return r, nil
		}
		return defaultFPS, nil
	}
	return fps, nil
}

// rateFromHeader reads the raw codec frame rate as a secondary probe.
func (t *Transcoder) rateFromHeader(ctx context.Context, input string) (float64, error) {
	out, err := runCapture(ctx, t.ffprobe,
		"-v", "error",
		"-select_streams", "v:0",
		"-show_entries", "stream=r_frame_rate",
		"-of", "csv=p=0", input)
	if err != nil {
		return 0, err
	}
	return parseFPSRational(strings.TrimSpace(out))
}

// buildFilter assembles the video filter graph: scale to the display, apply
// the requested rotation, normalise to the target frame rate and force a
// format libx264 accepts.
func buildFilter(rotate int, fps float64) string {
	var parts []string
	parts = append(parts, fmt.Sprintf("scale=%d:%d:flags=lanczos", screenWidth, screenHeight))

	switch rotate {
	case 90:
		parts = append(parts, "transpose=1") // 90° clockwise
	case 180:
		parts = append(parts, "transpose=1,transpose=1")
	case 270:
		parts = append(parts, "transpose=2") // 90° counter-clockwise
	}

	parts = append(parts, "format=yuv420p")
	return strings.Join(parts, ",")
}

// parseFPSRational parses an ffprobe frame rate such as "30000/1001",
// "25" or "0/0".
func parseFPSRational(s string) (float64, error) {
	s = strings.TrimSpace(s)
	if s == "" {
		return 0, fmt.Errorf("empty frame rate")
	}
	num, den := int64(1), int64(1)
	if i := strings.IndexByte(s, '/'); i >= 0 {
		var err error
		if num, err = strconv.ParseInt(s[:i], 10, 64); err != nil {
			return 0, fmt.Errorf("parse numerator %q: %w", s[:i], err)
		}
		if den, err = strconv.ParseInt(s[i+1:], 10, 64); err != nil {
			return 0, fmt.Errorf("parse denominator %q: %w", s[i+1:], err)
		}
	} else {
		n, err := strconv.ParseFloat(s, 64)
		if err != nil {
			return 0, fmt.Errorf("parse rate %q: %w", s, err)
		}
		return n, nil
	}
	if num <= 0 || den <= 0 {
		return 0, fmt.Errorf("non-positive frame rate %q", s)
	}
	return float64(num) / float64(den), nil
}

func clampFPS(fps float64) float64 {
	switch {
	case fps <= 0:
		return defaultFPS
	case fps > 120:
		return 120
	default:
		return fps
	}
}

func run(ctx context.Context, bin string, args ...string) error {
	cmd := exec.CommandContext(ctx, bin, args...)
	if out, err := cmd.CombinedOutput(); err != nil {
		msg := strings.TrimSpace(string(out))
		if msg != "" {
			return fmt.Errorf("%s: %w", msg, err)
		}
		return err
	}
	return nil
}

func runCapture(ctx context.Context, bin string, args ...string) (string, error) {
	cmd := exec.CommandContext(ctx, bin, args...)
	out, err := cmd.CombinedOutput()
	if err != nil {
		return "", err
	}
	return string(out), nil
}

func max(a, b int) int {
	if a > b {
		return a
	}
	return b
}