audio/mp3: Make readMainL3 to return *mpeg1MainData

This commit is contained in:
Hajime Hoshi 2017-06-18 18:06:41 +09:00
parent 940c3e65a2
commit d43df208d2
3 changed files with 78 additions and 72 deletions

View File

@ -21,17 +21,17 @@ import (
"io"
)
var mpeg1_scalefac_sizes = [16][2]int{
var mpeg1ScalefacSizes = [16][2]int{
{0, 0}, {0, 1}, {0, 2}, {0, 3}, {3, 0}, {1, 1}, {1, 2}, {1, 3},
{2, 1}, {2, 2}, {2, 3}, {3, 1}, {3, 2}, {3, 3}, {4, 2}, {4, 3},
}
func (f *frame) readMainL3() error {
nch := f.header.numberOfChannels()
func readMainL3(prev *mainDataBytes, header *mpeg1FrameHeader, sideInfo *mpeg1SideInfo) (*mpeg1MainData, *mainDataBytes, error) {
nch := header.numberOfChannels()
// Calculate header audio data size
framesize := f.header.frameSize()
framesize := header.frameSize()
if framesize > 2000 {
return fmt.Errorf("mp3: framesize = %d", framesize)
return nil, nil, fmt.Errorf("mp3: framesize = %d", framesize)
}
// Sideinfo is 17 bytes for one channel and 32 bytes for two
sideinfo_size := 32
@ -41,29 +41,29 @@ func (f *frame) readMainL3() error {
// Main data size is the rest of the frame,including ancillary data
main_data_size := framesize - sideinfo_size - 4 // sync+header
// CRC is 2 bytes
if f.header.protection_bit == 0 {
if header.protection_bit == 0 {
main_data_size -= 2
}
// Assemble main data buffer with data from this frame and the previous
// two frames. main_data_begin indicates how many bytes from previous
// frames that should be used. This buffer is later accessed by the
// getMainBits function in the same way as the side info is.
m, err := getMainData(f.prev, main_data_size, f.sideInfo.main_data_begin)
m, err := getMainData(prev, main_data_size, sideInfo.main_data_begin)
if err != nil {
// This could be due to not enough data in reservoir
return err
return nil, nil, err
}
f.mainDataBytes = m
md := &mpeg1MainData{}
for gr := 0; gr < 2; gr++ {
for ch := 0; ch < nch; ch++ {
part_2_start := m.getMainPos()
// Number of bits in the bitstream for the bands
slen1 := mpeg1_scalefac_sizes[f.sideInfo.scalefac_compress[gr][ch]][0]
slen2 := mpeg1_scalefac_sizes[f.sideInfo.scalefac_compress[gr][ch]][1]
if (f.sideInfo.win_switch_flag[gr][ch] != 0) && (f.sideInfo.block_type[gr][ch] == 2) {
if f.sideInfo.mixed_block_flag[gr][ch] != 0 {
slen1 := mpeg1ScalefacSizes[sideInfo.scalefac_compress[gr][ch]][0]
slen2 := mpeg1ScalefacSizes[sideInfo.scalefac_compress[gr][ch]][1]
if (sideInfo.win_switch_flag[gr][ch] != 0) && (sideInfo.block_type[gr][ch] == 2) {
if sideInfo.mixed_block_flag[gr][ch] != 0 {
for sfb := 0; sfb < 8; sfb++ {
f.mainData.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
md.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
}
for sfb := 3; sfb < 12; sfb++ {
//slen1 for band 3-5,slen2 for 6-11
@ -72,7 +72,7 @@ func (f *frame) readMainL3() error {
nbits = slen1
}
for win := 0; win < 3; win++ {
f.mainData.scalefac_s[gr][ch][sfb][win] = m.getMainBits(nbits)
md.scalefac_s[gr][ch][sfb][win] = m.getMainBits(nbits)
}
}
} else {
@ -83,64 +83,64 @@ func (f *frame) readMainL3() error {
nbits = slen1
}
for win := 0; win < 3; win++ {
f.mainData.scalefac_s[gr][ch][sfb][win] = m.getMainBits(nbits)
md.scalefac_s[gr][ch][sfb][win] = m.getMainBits(nbits)
}
}
}
} else { // block_type == 0 if winswitch == 0
// Scale factor bands 0-5
if (f.sideInfo.scfsi[ch][0] == 0) || (gr == 0) {
if (sideInfo.scfsi[ch][0] == 0) || (gr == 0) {
for sfb := 0; sfb < 6; sfb++ {
f.mainData.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
md.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
}
} else if (f.sideInfo.scfsi[ch][0] == 1) && (gr == 1) {
} else if (sideInfo.scfsi[ch][0] == 1) && (gr == 1) {
// Copy scalefactors from granule 0 to granule 1
for sfb := 0; sfb < 6; sfb++ {
f.mainData.scalefac_l[1][ch][sfb] = f.mainData.scalefac_l[0][ch][sfb]
md.scalefac_l[1][ch][sfb] = md.scalefac_l[0][ch][sfb]
}
}
// Scale factor bands 6-10
if (f.sideInfo.scfsi[ch][1] == 0) || (gr == 0) {
if (sideInfo.scfsi[ch][1] == 0) || (gr == 0) {
for sfb := 6; sfb < 11; sfb++ {
f.mainData.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
md.scalefac_l[gr][ch][sfb] = m.getMainBits(slen1)
}
} else if (f.sideInfo.scfsi[ch][1] == 1) && (gr == 1) {
} else if (sideInfo.scfsi[ch][1] == 1) && (gr == 1) {
// Copy scalefactors from granule 0 to granule 1
for sfb := 6; sfb < 11; sfb++ {
f.mainData.scalefac_l[1][ch][sfb] = f.mainData.scalefac_l[0][ch][sfb]
md.scalefac_l[1][ch][sfb] = md.scalefac_l[0][ch][sfb]
}
}
// Scale factor bands 11-15
if (f.sideInfo.scfsi[ch][2] == 0) || (gr == 0) {
if (sideInfo.scfsi[ch][2] == 0) || (gr == 0) {
for sfb := 11; sfb < 16; sfb++ {
f.mainData.scalefac_l[gr][ch][sfb] = m.getMainBits(slen2)
md.scalefac_l[gr][ch][sfb] = m.getMainBits(slen2)
}
} else if (f.sideInfo.scfsi[ch][2] == 1) && (gr == 1) {
} else if (sideInfo.scfsi[ch][2] == 1) && (gr == 1) {
// Copy scalefactors from granule 0 to granule 1
for sfb := 11; sfb < 16; sfb++ {
f.mainData.scalefac_l[1][ch][sfb] = f.mainData.scalefac_l[0][ch][sfb]
md.scalefac_l[1][ch][sfb] = md.scalefac_l[0][ch][sfb]
}
}
// Scale factor bands 16-20
if (f.sideInfo.scfsi[ch][3] == 0) || (gr == 0) {
if (sideInfo.scfsi[ch][3] == 0) || (gr == 0) {
for sfb := 16; sfb < 21; sfb++ {
f.mainData.scalefac_l[gr][ch][sfb] = m.getMainBits(slen2)
md.scalefac_l[gr][ch][sfb] = m.getMainBits(slen2)
}
} else if (f.sideInfo.scfsi[ch][3] == 1) && (gr == 1) {
} else if (sideInfo.scfsi[ch][3] == 1) && (gr == 1) {
// Copy scalefactors from granule 0 to granule 1
for sfb := 16; sfb < 21; sfb++ {
f.mainData.scalefac_l[1][ch][sfb] = f.mainData.scalefac_l[0][ch][sfb]
md.scalefac_l[1][ch][sfb] = md.scalefac_l[0][ch][sfb]
}
}
}
// Read Huffman coded data. Skip stuffing bits.
if err := f.readHuffman(part_2_start, gr, ch); err != nil {
return err
if err := m.readHuffman(header, sideInfo, md, part_2_start, gr, ch); err != nil {
return nil, nil, err
}
}
}
// The ancillary data is stored here,but we ignore it.
return nil
return md, m, nil
}
type mainDataBytes struct {
@ -151,12 +151,12 @@ type mainDataBytes struct {
pos int
}
func getMainData(prevFrame *frame, size int, offset int) (*mainDataBytes, error) {
func getMainData(prev *mainDataBytes, size int, offset int) (*mainDataBytes, error) {
if size > 1500 {
return nil, fmt.Errorf("mp3: size = %d", size)
}
// Check that there's data available from previous frames if needed
if prevFrame != nil && offset > len(prevFrame.mainDataBytes.vec) {
if prev != nil && offset > len(prev.vec) {
// No,there is not, so we skip decoding this frame, but we have to
// read the main_data bits from the bitstream in case they are needed
// for decoding the next frame.
@ -174,17 +174,16 @@ func getMainData(prevFrame *frame, size int, offset int) (*mainDataBytes, error)
}
return nil, err
}
m := &mainDataBytes{}
if prevFrame != nil {
m.vec = append(prevFrame.mainDataBytes.vec, buf...)
m := &mainDataBytes{
vec: append(prev.vec, buf...),
}
// TODO: Define a special error and enable to continue the next frame.
return m, fmt.Errorf("mp3: frame can't be decoded")
}
// Copy data from previous frames
vec := []int{}
if prevFrame != nil {
v := prevFrame.mainDataBytes.vec
if prev != nil {
v := prev.vec
vec = v[len(v)-offset:]
}
// Read the main_data from file

View File

@ -73,9 +73,16 @@ func (f *frame) readNextFrame() (*frame, error) {
// If there's not enough main data in the bit reservoir,
// signal to calling function so that decoding isn't done!
// Get main data(scalefactors and Huffman coded frequency data)
if err := nf.readMainL3(); err != nil {
var prevM *mainDataBytes
if f != nil {
prevM = f.mainDataBytes
}
md, mdb, err := readMainL3(prevM, h, s)
if err != nil {
return nil, err
}
nf.mainData = md
nf.mainDataBytes = mdb
return nf, nil
}
@ -176,90 +183,90 @@ func readHeader() (*mpeg1FrameHeader, error) {
return h, nil
}
func (f *frame) readHuffman(part_2_start, gr, ch int) error {
func (m *mainDataBytes) readHuffman(header *mpeg1FrameHeader, sideInfo *mpeg1SideInfo, mainData *mpeg1MainData, part_2_start, gr, ch int) error {
// Check that there is any data to decode. If not,zero the array.
if f.sideInfo.part2_3_length[gr][ch] == 0 {
if sideInfo.part2_3_length[gr][ch] == 0 {
for is_pos := 0; is_pos < 576; is_pos++ {
f.mainData.is[gr][ch][is_pos] = 0.0
mainData.is[gr][ch][is_pos] = 0.0
}
return nil
}
// Calculate bit_pos_end which is the index of the last bit for this part.
bit_pos_end := part_2_start + f.sideInfo.part2_3_length[gr][ch] - 1
bit_pos_end := part_2_start + sideInfo.part2_3_length[gr][ch] - 1
// Determine region boundaries
region_1_start := 0
region_2_start := 0
if (f.sideInfo.win_switch_flag[gr][ch] == 1) && (f.sideInfo.block_type[gr][ch] == 2) {
if (sideInfo.win_switch_flag[gr][ch] == 1) && (sideInfo.block_type[gr][ch] == 2) {
region_1_start = 36 // sfb[9/3]*3=36
region_2_start = 576 // No Region2 for short block case.
} else {
sfreq := f.header.sampling_frequency
sfreq := header.sampling_frequency
region_1_start =
sfBandIndicesSet[sfreq].l[f.sideInfo.region0_count[gr][ch]+1]
sfBandIndicesSet[sfreq].l[sideInfo.region0_count[gr][ch]+1]
region_2_start =
sfBandIndicesSet[sfreq].l[f.sideInfo.region0_count[gr][ch]+
f.sideInfo.region1_count[gr][ch]+2]
sfBandIndicesSet[sfreq].l[sideInfo.region0_count[gr][ch]+
sideInfo.region1_count[gr][ch]+2]
}
// Read big_values using tables according to region_x_start
for is_pos := 0; is_pos < f.sideInfo.big_values[gr][ch]*2; is_pos++ {
for is_pos := 0; is_pos < sideInfo.big_values[gr][ch]*2; is_pos++ {
table_num := 0
if is_pos < region_1_start {
table_num = f.sideInfo.table_select[gr][ch][0]
table_num = sideInfo.table_select[gr][ch][0]
} else if is_pos < region_2_start {
table_num = f.sideInfo.table_select[gr][ch][1]
table_num = sideInfo.table_select[gr][ch][1]
} else {
table_num = f.sideInfo.table_select[gr][ch][2]
table_num = sideInfo.table_select[gr][ch][2]
}
// Get next Huffman coded words
x, y, _, _, err := huffmanDecode(f.mainDataBytes, table_num)
x, y, _, _, err := huffmanDecode(m, table_num)
if err != nil {
return err
}
// In the big_values area there are two freq lines per Huffman word
f.mainData.is[gr][ch][is_pos] = float32(x)
mainData.is[gr][ch][is_pos] = float32(x)
is_pos++
f.mainData.is[gr][ch][is_pos] = float32(y)
mainData.is[gr][ch][is_pos] = float32(y)
}
// Read small values until is_pos = 576 or we run out of huffman data
table_num := f.sideInfo.count1table_select[gr][ch] + 32
is_pos := f.sideInfo.big_values[gr][ch] * 2
for (is_pos <= 572) && (f.mainDataBytes.getMainPos() <= bit_pos_end) {
table_num := sideInfo.count1table_select[gr][ch] + 32
is_pos := sideInfo.big_values[gr][ch] * 2
for (is_pos <= 572) && (m.getMainPos() <= bit_pos_end) {
// Get next Huffman coded words
x, y, v, w, err := huffmanDecode(f.mainDataBytes, table_num)
x, y, v, w, err := huffmanDecode(m, table_num)
if err != nil {
return err
}
f.mainData.is[gr][ch][is_pos] = float32(v)
mainData.is[gr][ch][is_pos] = float32(v)
is_pos++
if is_pos >= 576 {
break
}
f.mainData.is[gr][ch][is_pos] = float32(w)
mainData.is[gr][ch][is_pos] = float32(w)
is_pos++
if is_pos >= 576 {
break
}
f.mainData.is[gr][ch][is_pos] = float32(x)
mainData.is[gr][ch][is_pos] = float32(x)
is_pos++
if is_pos >= 576 {
break
}
f.mainData.is[gr][ch][is_pos] = float32(y)
mainData.is[gr][ch][is_pos] = float32(y)
is_pos++
}
// Check that we didn't read past the end of this section
if f.mainDataBytes.getMainPos() > (bit_pos_end + 1) {
if m.getMainPos() > (bit_pos_end + 1) {
// Remove last words read
is_pos -= 4
}
// Setup count1 which is the index of the first sample in the rzero reg.
f.sideInfo.count1[gr][ch] = is_pos
sideInfo.count1[gr][ch] = is_pos
// Zero out the last part if necessary
for is_pos < 576 {
f.mainData.is[gr][ch][is_pos] = 0.0
mainData.is[gr][ch][is_pos] = 0.0
is_pos++
}
// Set the bitpos to point to the next part to read
f.mainDataBytes.setMainPos(bit_pos_end + 1)
m.setMainPos(bit_pos_end + 1)
return nil
}

View File

@ -88,7 +88,7 @@ type frame struct {
header *mpeg1FrameHeader
sideInfo *mpeg1SideInfo
mainData mpeg1MainData
mainData *mpeg1MainData
mainDataBytes *mainDataBytes
store [2][32][18]float32