🚀 perf: stream frame generation to disk

This commit is contained in:
2026-07-16 21:56:41 +07:00
parent d7e831f373
commit 172615f68f
2 changed files with 342 additions and 40 deletions
+219 -39
View File
@@ -1,15 +1,20 @@
package main package main
import ( import (
"bufio"
"bytes" "bytes"
"encoding/binary"
"encoding/json" "encoding/json"
"errors" "errors"
"fmt" "fmt"
"io" "io"
"math" "math"
"os"
"os/exec" "os/exec"
"path/filepath"
"strconv" "strconv"
"strings" "strings"
"time"
) )
var ( var (
@@ -17,29 +22,177 @@ var (
jpegEOI = []byte{0xff, 0xd9} jpegEOI = []byte{0xff, 0xd9}
) )
const (
maxJPEGFrameBytes = 64 << 20
maxFFmpegLogBytes = 64 << 10
)
type jpegFrameSplitter struct { type jpegFrameSplitter struct {
buffer []byte buffer []byte
scan int
inJPEG bool
} }
func (s *jpegFrameSplitter) push(chunk []byte) [][]byte { func (s *jpegFrameSplitter) push(chunk []byte, emit func([]byte) error) (int, error) {
s.buffer = append(s.buffer, chunk...) s.buffer = append(s.buffer, chunk...)
var frames [][]byte emitted := 0
for { for {
start := bytes.Index(s.buffer, jpegSOI) if !s.inJPEG {
start := bytes.Index(s.buffer[s.scan:], jpegSOI)
if start == -1 { if start == -1 {
break // Retain only a possible marker prefix spanning two reads.
if len(s.buffer) > 0 && s.buffer[len(s.buffer)-1] == jpegSOI[0] {
s.buffer = s.buffer[len(s.buffer)-1:]
} else {
s.buffer = s.buffer[:0]
} }
end := bytes.Index(s.buffer[start+len(jpegSOI):], jpegEOI) s.scan = 0
return emitted, nil
}
start += s.scan
s.buffer = s.buffer[start:]
s.scan = len(jpegSOI)
s.inJPEG = true
}
end := bytes.Index(s.buffer[s.scan:], jpegEOI)
if end == -1 { if end == -1 {
break if len(s.buffer) > maxJPEGFrameBytes {
return emitted, fmt.Errorf("JPEG frame exceeds %d MiB limit", maxJPEGFrameBytes>>20)
} }
frameEnd := start + len(jpegSOI) + end + len(jpegEOI) s.scan = max(len(jpegSOI), len(s.buffer)-1)
frame := make([]byte, frameEnd-start) return emitted, nil
copy(frame, s.buffer[start:frameEnd]) }
frames = append(frames, frame) frameEnd := s.scan + end + len(jpegEOI)
if frameEnd > maxJPEGFrameBytes {
return emitted, fmt.Errorf("JPEG frame exceeds %d MiB limit", maxJPEGFrameBytes>>20)
}
if err := emit(s.buffer[:frameEnd]); err != nil {
return emitted, err
}
emitted++
s.buffer = s.buffer[frameEnd:] s.buffer = s.buffer[frameEnd:]
s.scan = 0
s.inJPEG = false
}
}
func (s *jpegFrameSplitter) finish() error {
if s.inJPEG {
return fmt.Errorf("ffmpeg produced a truncated JPEG frame")
}
return nil
}
type boundedLog struct {
buffer bytes.Buffer
limit int
}
func (w *boundedLog) Write(p []byte) (int, error) {
n := len(p)
if remaining := w.limit - w.buffer.Len(); remaining > 0 {
_, _ = w.buffer.Write(p[:min(len(p), remaining)])
}
return n, nil
}
func (w *boundedLog) String() string {
return strings.TrimSpace(w.buffer.String())
}
type streamingTSF struct {
file *os.File
writer *bufio.Writer
path string
count uint32
}
func newStreamingTSF(output string, fps float64) (*streamingTSF, error) {
if math.IsNaN(fps) || math.IsInf(fps, 0) || fps <= 0 || fps > maxTSFFPS {
return nil, fmt.Errorf("cannot write .tsf: fps must be finite and between 0 and %d", maxTSFFPS)
}
dir := filepath.Dir(output)
f, err := os.CreateTemp(dir, "."+filepath.Base(output)+"-*.tmp")
if err != nil {
return nil, err
}
s := &streamingTSF{file: f, writer: bufio.NewWriterSize(f, 1<<20), path: f.Name()}
if err := f.Chmod(0o644); err != nil {
s.abort()
return nil, err
}
if _, err := s.writer.WriteString(tsfMagic); err != nil {
s.abort()
return nil, err
}
var hdr [14]byte
binary.LittleEndian.PutUint16(hdr[0:], tsfVersion)
binary.LittleEndian.PutUint64(hdr[2:], math.Float64bits(fps))
if _, err := s.writer.Write(hdr[:]); err != nil {
s.abort()
return nil, err
}
return s, nil
}
func (s *streamingTSF) addFrame(frame []byte) error {
if s.count >= maxTSFFrameCount {
return fmt.Errorf("too many video frames")
}
if uint64(len(frame)) > math.MaxUint32 {
return fmt.Errorf("JPEG frame is too large")
}
var size [4]byte
binary.LittleEndian.PutUint32(size[:], uint32(len(frame)))
if _, err := s.writer.Write(size[:]); err != nil {
return err
}
if _, err := s.writer.Write(frame); err != nil {
return err
}
s.count++
return nil
}
func (s *streamingTSF) commit(output string) error {
if s.count == 0 {
return fmt.Errorf("no frames were decoded from the video")
}
if err := s.writer.Flush(); err != nil {
return err
}
if _, err := s.file.Seek(14, io.SeekStart); err != nil {
return err
}
var count [4]byte
binary.LittleEndian.PutUint32(count[:], s.count)
if _, err := s.file.Write(count[:]); err != nil {
return err
}
if err := s.file.Sync(); err != nil {
return err
}
if err := s.file.Close(); err != nil {
return err
}
s.file = nil
if err := os.Rename(s.path, output); err != nil {
return err
}
s.path = ""
return nil
}
func (s *streamingTSF) abort() {
if s.file != nil {
_ = s.file.Close()
s.file = nil
}
if s.path != "" {
_ = os.Remove(s.path)
s.path = ""
} }
return frames
} }
type ffprobeOutput struct { type ffprobeOutput struct {
@@ -72,59 +225,82 @@ func parseFrameRate(rate string) float64 {
return num return num
} }
func extractFrames(path, vf, label string, maxDimension, totalFrames int) ([][]byte, error) { func extractFrames(path, vf, label string, totalFrames int, emit func([]byte) error) (int, error) {
cmd := exec.Command( cmd := exec.Command(
"ffmpeg", "-i", path, "ffmpeg", "-hide_banner", "-loglevel", "error", "-nostats",
"-i", path,
"-c:v", "mjpeg", "-c:v", "mjpeg",
"-q:v", "3", "-q:v", "3",
"-vf", vf, "-vf", vf,
"-f", "image2pipe", "-f", "image2pipe",
"pipe:1", "pipe:1",
) )
var stderr bytes.Buffer stderr := &boundedLog{limit: maxFFmpegLogBytes}
cmd.Stderr = &stderr cmd.Stderr = stderr
stdout, err := cmd.StdoutPipe() stdout, err := cmd.StdoutPipe()
if err != nil { if err != nil {
return nil, fmt.Errorf("ffmpeg failed: %w", err) return 0, fmt.Errorf("ffmpeg failed: %w", err)
} }
if err := cmd.Start(); err != nil { if err := cmd.Start(); err != nil {
return nil, fmt.Errorf("ffmpeg failed: %w", err) return 0, fmt.Errorf("ffmpeg failed: %w", err)
} }
var frames [][]byte
splitter := &jpegFrameSplitter{} splitter := &jpegFrameSplitter{}
reportProgress := func(count int) { frameCount := 0
if totalFrames > 0 { lastReport := time.Now()
pct := min(100, int(math.Round(float64(count)/float64(totalFrames)*100))) reportProgress := func(force bool) {
fmt.Printf("\rGenerating %s frames: %d/%d (%d%%)", label, count, totalFrames, pct) if !force && time.Since(lastReport) < 250*time.Millisecond {
} else { return
fmt.Printf("\rGenerating %s frames: %d", label, count)
} }
if totalFrames > 0 {
pct := min(100, int(math.Round(float64(frameCount)/float64(totalFrames)*100)))
fmt.Printf("\rGenerating %s frames: %d/%d (%d%%)", label, frameCount, totalFrames, pct)
} else {
fmt.Printf("\rGenerating %s frames: %d", label, frameCount)
}
lastReport = time.Now()
}
failStream := func(err error) (int, error) {
_ = cmd.Process.Kill()
_ = cmd.Wait()
return frameCount, err
} }
buf := make([]byte, 256*1024) buf := make([]byte, 256*1024)
for { for {
n, err := stdout.Read(buf) n, readErr := stdout.Read(buf)
if n > 0 { if n > 0 {
frames = append(frames, splitter.push(buf[:n])...) emitted, splitErr := splitter.push(buf[:n], emit)
reportProgress(len(frames)) frameCount += emitted
if splitErr != nil {
return failStream(fmt.Errorf("ffmpeg stream error: %w", splitErr))
} }
if err == io.EOF { if emitted > 0 {
reportProgress(false)
}
}
if readErr == io.EOF {
break break
} }
if err != nil { if readErr != nil {
_ = cmd.Wait() return failStream(fmt.Errorf("ffmpeg stream error: %w", readErr))
return nil, fmt.Errorf("ffmpeg stream error: %s", err.Error())
} }
} }
if err := cmd.Wait(); err != nil { if err := cmd.Wait(); err != nil {
return nil, fmt.Errorf("ffmpeg failed: %s", strings.TrimSpace(stderr.String())) if msg := stderr.String(); msg != "" {
return frameCount, fmt.Errorf("ffmpeg failed: %s", msg)
} }
if len(frames) == 0 { return frameCount, fmt.Errorf("ffmpeg failed: %w", err)
return nil, fmt.Errorf("no frames were decoded from the video")
} }
if err := splitter.finish(); err != nil {
return frameCount, err
}
if frameCount == 0 {
return 0, fmt.Errorf("no frames were decoded from the video")
}
reportProgress(true)
fmt.Println() fmt.Println()
return frames, nil return frameCount, nil
} }
func processVideo(path, output string, maxDimension int) error { func processVideo(path, output string, maxDimension int) error {
@@ -175,15 +351,19 @@ func processVideo(path, output string, maxDimension int) error {
maxDimension, maxDimension, maxDimension, maxDimension,
) )
colorFrames, err := extractFrames(path, scaleFilter, "color", maxDimension, totalFrames) outputFile, err := newStreamingTSF(output, fps)
if err != nil { if err != nil {
return err return err
} }
defer outputFile.abort()
videoData := FramesContainer{FPS: fps, ColorFrames: colorFrames} frameCount, err := extractFrames(path, scaleFilter, "color", totalFrames, outputFile.addFrame)
if err := writeTSF(output, &videoData); err != nil { if err != nil {
return err return err
} }
logInfo(fmt.Sprintf("Saved %d frames to %s", len(videoData.ColorFrames), output)) if err := outputFile.commit(output); err != nil {
return err
}
logInfo(fmt.Sprintf("Saved %d frames to %s", frameCount, output))
return nil return nil
} }
+122
View File
@@ -0,0 +1,122 @@
package main
import (
"bytes"
"os"
"path/filepath"
"testing"
)
func TestJPEGFrameSplitterAcrossChunks(t *testing.T) {
splitter := &jpegFrameSplitter{}
var frames [][]byte
emit := func(frame []byte) error {
frames = append(frames, bytes.Clone(frame))
return nil
}
chunks := [][]byte{
{0x01, 0x02, 0xff},
{0xd8, 0x10, 0xff},
{0xd9, 0xff, 0xd8, 0x20},
{0x30, 0xff},
{0xd9, 0x03},
}
for _, chunk := range chunks {
if _, err := splitter.push(chunk, emit); err != nil {
t.Fatalf("push: %v", err)
}
}
if err := splitter.finish(); err != nil {
t.Fatalf("finish: %v", err)
}
want := [][]byte{
{0xff, 0xd8, 0x10, 0xff, 0xd9},
{0xff, 0xd8, 0x20, 0x30, 0xff, 0xd9},
}
if len(frames) != len(want) {
t.Fatalf("got %d frames, want %d", len(frames), len(want))
}
for i := range want {
if !bytes.Equal(frames[i], want[i]) {
t.Errorf("frame %d = %x, want %x", i, frames[i], want[i])
}
}
}
func TestJPEGFrameSplitterRejectsTruncatedFrame(t *testing.T) {
splitter := &jpegFrameSplitter{}
if _, err := splitter.push([]byte{0xff, 0xd8, 0x01}, func([]byte) error { return nil }); err != nil {
t.Fatalf("push: %v", err)
}
if err := splitter.finish(); err == nil {
t.Fatal("finish accepted a truncated JPEG")
}
}
func TestBoundedLog(t *testing.T) {
log := &boundedLog{limit: 4}
if n, err := log.Write([]byte("abcdefgh")); err != nil || n != 8 {
t.Fatalf("Write = %d, %v", n, err)
}
if got := log.String(); got != "abcd" {
t.Fatalf("String = %q, want %q", got, "abcd")
}
}
func TestStreamingTSFCommitAndAbort(t *testing.T) {
dir := t.TempDir()
output := filepath.Join(dir, "frames.tsf")
stream, err := newStreamingTSF(output, 24)
if err != nil {
t.Fatalf("newStreamingTSF: %v", err)
}
for _, frame := range [][]byte{{1, 2, 3}, {4, 5}} {
if err := stream.addFrame(frame); err != nil {
t.Fatalf("addFrame: %v", err)
}
}
if err := stream.commit(output); err != nil {
t.Fatalf("commit: %v", err)
}
stream.abort()
got, err := loadTSF(output)
if err != nil {
t.Fatalf("loadTSF: %v", err)
}
defer func() { _ = got.Close() }()
if got.FPS != 24 || len(got.ColorFrames) != 2 || !bytes.Equal(got.ColorFrames[1], []byte{4, 5}) {
t.Fatalf("unexpected streamed TSF: %+v", got)
}
original := []byte("existing destination")
if err := os.WriteFile(output, original, 0o644); err != nil {
t.Fatalf("WriteFile: %v", err)
}
failed, err := newStreamingTSF(output, 24)
if err != nil {
t.Fatalf("newStreamingTSF: %v", err)
}
if err := failed.addFrame([]byte{9}); err != nil {
t.Fatalf("addFrame: %v", err)
}
failed.abort()
contents, err := os.ReadFile(output)
if err != nil {
t.Fatalf("ReadFile: %v", err)
}
if !bytes.Equal(contents, original) {
t.Fatalf("destination changed after abort: %q", contents)
}
matches, err := filepath.Glob(filepath.Join(dir, ".frames.tsf-*.tmp"))
if err != nil {
t.Fatalf("Glob: %v", err)
}
if len(matches) != 0 {
t.Fatalf("temporary files remain after abort: %v", matches)
}
}