feat: decode Vorbis audio tracks

Author: nerg4l

extract_audio.go

@@ -2,13 +2,17 @@ package matroska
 
 import (
 	"fmt"
+	"github.com/coding-socks/matroska/internal/vorbis"
 	"io"
+	"math/rand/v2"
 )
 
 func extractTrackAudio(w io.Writer, s *Scanner, t TrackEntry) error {
 	switch t.CodecID {
 	case AudioCodecMP2, AudioCodecMP3:
 		return extractTrackMPEG(w, s, t)
+	case AudioCodecVORBIS:
+		return extractTrackVORBIS(w, s, t)
 	}
 	return fmt.Errorf("matroska: unknown audio codec %s", t.CodecID)
 }
@@ -25,23 +29,120 @@ func extractTrackMPEG(w io.Writer, s *Scanner, t TrackEntry) error {
 			if err != nil {
 				return fmt.Errorf("matroska: could not create block struct: %w", err)
 			}
-			if block.TrackNumber() == t.TrackNumber {
-				if _, err := io.Copy(w, block.Data()); err != nil {
-					return err
-				}
+			if block.TrackNumber() != t.TrackNumber {
+				continue
+			}
+			if _, err := io.Copy(w, block.Data()); err != nil {
+				return err
 			}
 		}
 		for i := range c.BlockGroup {
 			block, err := ReadBlock(c.BlockGroup[i].Block, c.Timestamp)
 			if err != nil {
 				return fmt.Errorf("matroska: could not create block struct: %w", err)
 			}
-			if block.TrackNumber() == t.TrackNumber {
-				if _, err := io.Copy(w, block.Data()); err != nil {
+			if block.TrackNumber() != t.TrackNumber {
+				continue
+			}
+			if _, err := io.Copy(w, block.Data()); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
+// extractTrackVORBIS is based on the Vorbis I specification created by the Xiph.Org Foundation.
+// See: https://xiph.org/vorbis/doc/Vorbis_I_spec.pdf
+func extractTrackVORBIS(w io.Writer, s *Scanner, track TrackEntry) error {
+	serialNum := rand.Int32()
+
+	vw := vorbis.NewWriter(w, serialNum)
+
+	codecPrivate := *track.CodecPrivate
+	frames := Frames(LacingFlagXiph, codecPrivate)
+	if len(frames) != 3 {
+		return fmt.Errorf("matroska: Vorbis audio track requires 3 header pages, got %d", len(frames))
+	}
+
+	ih := frames[0]
+	if err := vw.WriteIdentHeader(ih); err != nil {
+		return err
+	}
+	iheader, err := vorbis.ParseIdentificationHeader([30]byte(ih))
+	if err != nil {
+		return err
+	}
+	blockSizes := []uint16{
+		iheader.Blocksize0,
+		iheader.Blocksize1,
+	}
+	cheader, err := vorbis.ParseCommentHeader(frames[1])
+	if err != nil {
+		return err
+	}
+	_ = cheader
+
+	if err := vw.WriteHeaders(frames[1], frames[2]); err != nil {
+		return err
+	}
+
+	var (
+		prevBlockSize uint64 = 0
+		granpos       uint64 = 0
+
+		prevFrame []byte //
+	)
+	frames = frames[:0] // clear value
+
+	for s.Next() {
+		c := s.Cluster()
+		if (len(c.SimpleBlock) + len(c.BlockGroup)) == 0 {
+			continue
+		}
+
+		for i := range c.SimpleBlock {
+			block, err := ReadSimpleBlock(c.SimpleBlock[i], c.Timestamp)
+			if err != nil {
+				return fmt.Errorf("matroska: could not create block struct: %w", err)
+			}
+			if block.TrackNumber() != track.TrackNumber {
+				continue
+			}
+			frames = append(frames, block.Frames()...)
+		}
+		for i := range c.BlockGroup {
+			block, err := ReadBlock(c.BlockGroup[i].Block, c.Timestamp)
+			if err != nil {
+				return fmt.Errorf("matroska: could not create block struct: %w", err)
+			}
+			if block.TrackNumber() != track.TrackNumber {
+				continue
+			}
+			frames = append(frames, block.Frames()...)
+		}
+
+		for _, frame := range frames {
+			blockSize := uint64(blockSizes[(frame[0]>>1)&1])
+
+			if prevFrame != nil {
+				if err := vw.Segment(prevFrame, granpos, false); err != nil {
 					return err
 				}
+
+				// We need at least two segment to calculate this
+				granpos += (blockSize + prevBlockSize) / 4
 			}
+
+			prevBlockSize = blockSize
+			prevFrame = frame
 		}
+
+		frames = frames[:0]
 	}
-	return nil
+	if prevFrame == nil {
+		return nil
+	}
+	// only this element can be the last.
+	return vw.Segment(prevFrame, granpos, true)
 }

extract_video.go

@@ -27,18 +27,20 @@ func extractTrackMSCOMP(w io.Writer, s *Scanner, t TrackEntry) error {
 			if err != nil {
 				return fmt.Errorf("matroska: could not create block struct: %w", err)
 			}
-			if block.TrackNumber() == t.TrackNumber {
-				blocks = append(blocks, block)
+			if block.TrackNumber() != t.TrackNumber {
+				continue
 			}
+			blocks = append(blocks, block)
 		}
 		for i := range c.BlockGroup {
 			block, err := ReadBlock(c.BlockGroup[i].Block, c.Timestamp)
 			if err != nil {
 				return fmt.Errorf("matroska: could not create block struct: %w", err)
 			}
-			if block.TrackNumber() == t.TrackNumber {
-				blocks = append(blocks, block)
+			if block.TrackNumber() != t.TrackNumber {
+				continue
 			}
+			blocks = append(blocks, block)
 		}
 	}
 	_ = scale

go.mod

@@ -7,7 +7,7 @@ toolchain go1.24.1
 require (
 	github.com/charmbracelet/huh v0.6.0
 	github.com/charmbracelet/huh/spinner v0.0.0-20250410174039-76d1f8226680
-	github.com/coding-socks/ebml v0.0.0-20250409185428-3a0bcb87ec98
+	github.com/coding-socks/ebml v0.0.0-20250413191753-908e87a1adf5
 	github.com/spf13/pflag v1.0.6
 	golang.org/x/tools v0.32.0
 )
@@ -41,3 +41,5 @@ require (
 	golang.org/x/sys v0.32.0 // indirect
 	golang.org/x/text v0.24.0 // indirect
 )
+
+tool github.com/coding-socks/ebml

go.sum

@@ -32,6 +32,8 @@ github.com/charmbracelet/x/term v0.2.1 h1:AQeHeLZ1OqSXhrAWpYUtZyX1T3zVxfpZuEQMIQ
 github.com/charmbracelet/x/term v0.2.1/go.mod h1:oQ4enTYFV7QN4m0i9mzHrViD7TQKvNEEkHUMCmsxdUg=
 github.com/coding-socks/ebml v0.0.0-20250409185428-3a0bcb87ec98 h1:OMkiM9+3+Q0Q+6r6s5Y3xJXNrFq6uAdfODbTF3iUxmo=
 github.com/coding-socks/ebml v0.0.0-20250409185428-3a0bcb87ec98/go.mod h1:6RUr58mWiQHHZiL2UFb8JygHNS5E35wGJfEmoSsDyI0=
+github.com/coding-socks/ebml v0.0.0-20250413191753-908e87a1adf5 h1:fgjgpV5PH7z03eKverW8/e7F2c/HlSVigr57N7HslWw=
+github.com/coding-socks/ebml v0.0.0-20250413191753-908e87a1adf5/go.mod h1:6RUr58mWiQHHZiL2UFb8JygHNS5E35wGJfEmoSsDyI0=
 github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY=
 github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto=
 github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f h1:Y/CXytFA4m6baUTXGLOoWe4PQhGxaX0KpnayAqC48p4=

internal/ogg/crc32.go

@@ -0,0 +1,66 @@
+package ogg
+
+import (
+	"hash"
+	"hash/crc32"
+)
+
+var crcTable = makeOggCRC32Table()
+
+func makeOggCRC32Table() *crc32.Table {
+	crc32.NewIEEE()
+	t := new(crc32.Table)
+	for i := 0; i < 256; i++ {
+		r := uint32(i) << 24
+		for j := 0; j < 8; j++ {
+			if r&0x80000000 != 0 {
+				r = (r << 1) ^ 0x04c11db7
+			} else {
+				r <<= 1
+			}
+		}
+		t[i] = r
+	}
+	return t
+}
+
+func update(crc uint32, tab *crc32.Table, data []byte) uint32 {
+	for _, v := range data {
+		crc = (crc << 8) ^ tab[byte(crc>>24)^v]
+	}
+	return crc
+}
+
+func CRC32Checksum(data []byte) uint32 {
+	return update(0, crcTable, data)
+}
+
+// digest represents the partial evaluation of a checksum.
+type digest struct {
+	crc uint32
+	tab *crc32.Table
+}
+
+func NewCRC32() hash.Hash32 {
+	return &digest{0, crcTable}
+}
+
+func (d *digest) Size() int { return crc32.Size }
+
+func (d *digest) BlockSize() int { return 1 }
+
+func (d *digest) Reset() { d.crc = 0 }
+
+func (d *digest) Write(p []byte) (n int, err error) {
+	// We only create digest objects through New() which takes care of
+	// initialization in this case.
+	d.crc = update(d.crc, d.tab, p)
+	return len(p), nil
+}
+
+func (d *digest) Sum32() uint32 { return d.crc }
+
+func (d *digest) Sum(in []byte) []byte {
+	s := d.Sum32()
+	return append(in, byte(s>>24), byte(s>>16), byte(s>>8), byte(s))
+}

internal/ogg/crc32_test.go

@@ -0,0 +1,16 @@
+package ogg
+
+import (
+	"io"
+	"testing"
+)
+
+func TestSum32(t *testing.T) {
+	c := NewCRC32()
+	in := "123456789"
+	io.WriteString(c, in)
+	s := c.Sum32()
+	if out := uint32(0x89a1897f); s != out {
+		t.Fatalf("jones crc64(%s) = 0x%x want 0x%x", in, s, out)
+	}
+}