1
0
mirror of https://github.com/nu774/fdkaac.git synced 2025-06-05 23:29:14 +02:00

6 Commits

Author SHA1 Message Date
5f0d784cd3 bump version 2013-10-25 00:03:13 +09:00
6a3b77de8e update README 2013-10-25 00:02:23 +09:00
1af8624b00 caf input support 2013-10-24 23:47:03 +09:00
29a8f73faf refactor pcm io routines 2013-10-24 10:38:27 +09:00
3b666b7546 cleanup metadata handling 2013-10-23 23:35:23 +09:00
8cb6378fca --tag-from-json: properly support number/total format in json track field 2013-10-23 12:19:14 +09:00
15 changed files with 1247 additions and 388 deletions

View File

@ -96,12 +96,14 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ </Command>
<ItemGroup>
<ClCompile Include="..\missings\getopt.c" />
<ClCompile Include="..\src\aacenc.c" />
<ClCompile Include="..\src\caf_reader.c" />
<ClCompile Include="..\src\compat_win32.c" />
<ClCompile Include="..\src\lpcm.c" />
<ClCompile Include="..\src\m4af.c" />
<ClCompile Include="..\src\main.c" />
<ClCompile Include="..\src\metadata.c" />
<ClCompile Include="..\src\parson.c" />
<ClCompile Include="..\src\pcm_readhelper.c" />
<ClCompile Include="..\src\pcm_sint16_converter.c" />
<ClCompile Include="..\src\progress.c" />
<ClCompile Include="..\src\wav_reader.c" />
@ -115,6 +117,7 @@ copy ..\fdk-aac\libSYS\include\machine_type.h include\fdk-aac\ </Command>
<ClInclude Include="..\src\m4af_endian.h" />
<ClInclude Include="..\src\metadata.h" />
<ClInclude Include="..\src\parson.h" />
<ClInclude Include="..\src\pcm_reader.h" />
<ClInclude Include="..\src\progress.h" />
<ClInclude Include="..\src\wav_reader.h" />
</ItemGroup>

View File

@ -18,6 +18,9 @@
<ClCompile Include="..\src\aacenc.c">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="..\src\caf_reader.c">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="..\src\compat_win32.c">
<Filter>Source Files</Filter>
</ClCompile>
@ -30,6 +33,12 @@
<ClCompile Include="..\src\main.c">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="..\src\pcm_readhelper.c">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="..\src\pcm_sint16_converter.c">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="..\src\progress.c">
<Filter>Source Files</Filter>
</ClCompile>
@ -44,6 +53,12 @@
<ClInclude Include="..\src\aacenc.h">
<Filter>Header Files</Filter>
</ClInclude>
<ClInclude Include="..\src\caf_reader.h">
<Filter>Header Files</Filter>
</ClInclude>
<ClInclude Include="..\src\catypes.h">
<Filter>Header Files</Filter>
</ClInclude>
<ClInclude Include="..\src\compat.h">
<Filter>Header Files</Filter>
</ClInclude>

View File

@ -5,11 +5,13 @@ bin_PROGRAMS = fdkaac
fdkaac_SOURCES = \
src/aacenc.c \
src/caf_reader.c \
src/lpcm.c \
src/m4af.c \
src/main.c \
src/metadata.c \
src/parson.c \
src/pcm_readhelper.c \
src/pcm_sint16_converter.c \
src/progress.c \
src/wav_reader.c

13
README
View File

@ -34,9 +34,16 @@ MSVC solution for Visual Studio 2010 is under MSVC directory.
Available input format
----------------------
WAV (or RF64), upto 32bit int / 64bit float format is supported.
However, since FDK AAC encoder is implemented based on fixed point integer,
encoder itself treats 16bit input only.
WAV, RF64, CAF, RAW, upto 32bit int / 64bit float format is supported.
Metadata in CAF info chunk can be read and copied to the resulting m4a.
This is especially useful and works well when you pipe from ffmpeg via CAF.
For example, you can copy tag from original "foo.flac" to "foo.m4a"
through the following pipeline:
$ ffmpeg -i foo.flac -f caf - | fdkaac -m3 - -o foo.m4a
Since FDK AAC encoder is implemented based on fixed point integer,
encoder itself handles 16bit input only.
Therefore, when feeding non-integer input, be careful so that input doesn't
exceed 0dBFS to avoid hard clips.
You might also want to apply dither/noise shape beforehand when your input

253
src/caf_reader.c Normal file
View File

@ -0,0 +1,253 @@
/*
* Copyright (C) 2013 nu774
* For conditions of distribution and use, see copyright notice in COPYING
*/
#if HAVE_CONFIG_H
# include "config.h"
#endif
#if HAVE_STDINT_H
# include <stdint.h>
#endif
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdarg.h>
#include "caf_reader.h"
#include "m4af.h"
typedef struct caf_reader_t {
pcm_reader_vtbl_t *vtbl;
pcm_sample_description_t sample_format;
int64_t length;
int64_t position;
int64_t data_offset;
pcm_io_context_t io;
aacenc_tag_callback_t tag_callback;
void *tag_ctx;
uint8_t chanmap[8];
} caf_reader_t;
static const pcm_sample_description_t *caf_get_format(pcm_reader_t *reader)
{
return &((caf_reader_t *)reader)->sample_format;
}
static int64_t caf_get_length(pcm_reader_t *reader)
{
return ((caf_reader_t *)reader)->length;
}
static int64_t caf_get_position(pcm_reader_t *reader)
{
return ((caf_reader_t *)reader)->position;
}
static void caf_teardown(pcm_reader_t **reader)
{
free(*reader);
*reader = 0;
}
static
uint32_t caf_next_chunk(caf_reader_t *reader, int64_t *chunk_size)
{
uint32_t fcc;
if (pcm_scanb(&reader->io, "LQ", &fcc, chunk_size) == 2)
return fcc;
return 0;
}
static
int caf_desc(caf_reader_t *reader, int64_t chunk_size)
{
double mSampleRate;
uint32_t mFormatID, mFormatFlags, mBytesPerPacket, mFramesPerPacket,
mChannelsPerFrame, mBitsPerChannel;
pcm_sample_description_t *desc = &reader->sample_format;
ENSURE(chunk_size >= 32);
TRY_IO(pcm_scanb(&reader->io, "QLLLLLL", &mSampleRate, &mFormatID,
&mFormatFlags, &mBytesPerPacket, &mFramesPerPacket,
&mChannelsPerFrame, &mBitsPerChannel) != 7);
ENSURE(mFormatID == M4AF_FOURCC('l','p','c','m'));
ENSURE(mSampleRate && mBytesPerPacket &&
mChannelsPerFrame >= 1 && mChannelsPerFrame <= 8 &&
mBitsPerChannel && mFramesPerPacket == 1 &&
mBytesPerPacket % mChannelsPerFrame == 0 &&
mBytesPerPacket >= mChannelsPerFrame * ((mBitsPerChannel + 7) / 8));
desc->sample_rate = mSampleRate;
desc->bits_per_channel = mBitsPerChannel;
desc->bytes_per_frame = mBytesPerPacket;
desc->channels_per_frame = mChannelsPerFrame;
switch (mFormatFlags) {
case 0: desc->sample_type = PCM_TYPE_SINT_BE; break;
case 1: desc->sample_type = PCM_TYPE_FLOAT_BE; break;
case 2: desc->sample_type = PCM_TYPE_SINT; break;
case 3: desc->sample_type = PCM_TYPE_FLOAT; break;
default: goto FAIL;
}
TRY_IO(pcm_skip(&reader->io, chunk_size - 32));
return 0;
FAIL:
return -1;
}
static
int caf_info(caf_reader_t *reader, int64_t chunk_size)
{
char *buf, *key, *val, *end;
size_t len;
if (chunk_size < 4 || (buf = malloc(chunk_size)) == 0)
return -1;
pcm_read(&reader->io, buf, chunk_size);
key = buf + 4;
end = buf + chunk_size;
do {
if ((val = key + strlen(key) + 1) < end) {
len = strlen(val);
if (reader->tag_callback)
reader->tag_callback(reader->tag_ctx, key, val, len);
key = val + len + 1;
}
} while (key < end && val < end);
if (reader->tag_callback)
reader->tag_callback(reader->tag_ctx, 0, 0, 0);
free(buf);
return 0;
}
static
int caf_read_frames(pcm_reader_t *preader, void *buffer, unsigned nframes)
{
int rc;
unsigned i, j, nbytes;
caf_reader_t *reader = (caf_reader_t *)preader;
unsigned bpf = reader->sample_format.bytes_per_frame;
unsigned nchannels = reader->sample_format.channels_per_frame;
unsigned bpc = bpf / nchannels;
uint8_t tmp[64]; /* enough room for maximum bpf: 8ch float64 */
uint8_t *bp;
uint8_t *chanmap = reader->chanmap;
if (nframes > reader->length - reader->position)
nframes = reader->length - reader->position;
nbytes = nframes * bpf;
if (nbytes) {
if ((rc = pcm_read(&reader->io, buffer, nbytes)) < 0)
return -1;
nframes = rc / bpf;
for (bp = buffer, i = 0; i < nframes; ++i, bp += bpf) {
memcpy(tmp, bp, bpf);
for (j = 0; j < nchannels; ++j)
memcpy(bp + bpc * j, tmp + bpc * chanmap[j], bpc);
}
reader->position += nframes;
}
if (nframes == 0) {
/* fetch info after data chunk */
uint32_t fcc;
int64_t chunk_size;
while ((fcc = caf_next_chunk(reader, &chunk_size)) != 0) {
if (fcc == M4AF_FOURCC('i','n','f','o'))
TRY_IO(caf_info(reader, chunk_size));
else
TRY_IO(pcm_skip(&reader->io, chunk_size));
}
}
return nframes;
FAIL:
return 0;
}
static
int caf_parse(caf_reader_t *reader, int64_t *data_length)
{
uint32_t fcc;
int64_t chunk_size;
*data_length = 0;
/* CAFFileHeader */
TRY_IO(pcm_read32be(&reader->io, &fcc));
ENSURE(fcc == M4AF_FOURCC('c','a','f','f'));
TRY_IO(pcm_skip(&reader->io, 4)); /* mFileVersion, mFileFlags */
while ((fcc = caf_next_chunk(reader, &chunk_size)) != 0) {
if (fcc == M4AF_FOURCC('d','e','s','c'))
TRY_IO(caf_desc(reader, chunk_size));
else if (fcc == M4AF_FOURCC('i','n','f','o'))
TRY_IO(caf_info(reader, chunk_size));
else if (fcc == M4AF_FOURCC('c','h','a','n')) {
ENSURE(reader->sample_format.channels_per_frame);
if (apple_chan_chunk(&reader->io, chunk_size,
&reader->sample_format, reader->chanmap) < 0)
goto FAIL;
} else if (fcc == M4AF_FOURCC('d','a','t','a')) {
TRY_IO(pcm_skip(&reader->io, 4)); /* mEditCount */
*data_length = (chunk_size == ~0ULL) ? chunk_size : chunk_size - 4;
reader->data_offset += 12;
break;
} else
TRY_IO(pcm_skip(&reader->io, chunk_size));
reader->data_offset += (chunk_size + 8);
}
ENSURE(reader->sample_format.channels_per_frame);
ENSURE(fcc == M4AF_FOURCC('d','a','t','a'));
return 0;
FAIL:
return -1;
}
static pcm_reader_vtbl_t caf_vtable = {
caf_get_format,
caf_get_length,
caf_get_position,
caf_read_frames,
caf_teardown
};
pcm_reader_t *caf_open(pcm_io_context_t *io,
aacenc_tag_callback_t tag_callback, void *tag_ctx)
{
caf_reader_t *reader = 0;
int64_t data_length;
unsigned bpf;
if ((reader = calloc(1, sizeof(caf_reader_t))) == 0)
return 0;
memcpy(&reader->io, io, sizeof(pcm_io_context_t));
reader->tag_callback = tag_callback;
reader->tag_ctx = tag_ctx;
if (caf_parse(reader, &data_length) < 0) {
free(reader);
return 0;
}
bpf = reader->sample_format.bytes_per_frame;
/* CAF uses -1 to indicate "unknown size" */
if (data_length < 0 || data_length % bpf)
reader->length = INT64_MAX;
else
reader->length = data_length / bpf;
if (reader->length == INT64_MAX) {
if (pcm_seek(&reader->io, 0, SEEK_END) >= 0) {
int64_t size = pcm_tell(&reader->io);
if (size > 0)
reader->length = (size - reader->data_offset) / bpf;
pcm_seek(&reader->io, reader->data_offset, SEEK_SET);
}
}
reader->vtbl = &caf_vtable;
return (pcm_reader_t *)reader;
}

15
src/caf_reader.h Normal file
View File

@ -0,0 +1,15 @@
/*
* Copyright (C) 2013 nu774
* For conditions of distribution and use, see copyright notice in COPYING
*/
#ifndef CAF_READER_H
#define CAF_READER_H
#include "lpcm.h"
#include "pcm_reader.h"
#include "metadata.h"
pcm_reader_t *caf_open(pcm_io_context_t *io,
aacenc_tag_callback_t tag_callback, void *tag_ctx);
#endif

241
src/catypes.h Normal file
View File

@ -0,0 +1,241 @@
#if !defined(__CoreAudioTypes_h__)
#define __CoreAudioTypes_h__
enum { kVariableLengthArray = 1 };
typedef uint32_t AudioChannelLabel;
typedef uint32_t AudioChannelLayoutTag;
struct AudioChannelDescription
{
AudioChannelLabel mChannelLabel;
uint32_t mChannelFlags;
float mCoordinates[3];
};
typedef struct AudioChannelDescription AudioChannelDescription;
struct AudioChannelLayout
{
AudioChannelLayoutTag mChannelLayoutTag;
uint32_t mChannelBitmap;
uint32_t mNumberChannelDescriptions;
AudioChannelDescription mChannelDescriptions[kVariableLengthArray];
};
typedef struct AudioChannelLayout AudioChannelLayout;
enum
{
kAudioChannelLabel_Unknown = 0xFFFFFFFF, // unknown or unspecified other use
kAudioChannelLabel_Unused = 0, // channel is present, but has no intended use or destination
kAudioChannelLabel_UseCoordinates = 100, // channel is described by the mCoordinates fields.
kAudioChannelLabel_Left = 1,
kAudioChannelLabel_Right = 2,
kAudioChannelLabel_Center = 3,
kAudioChannelLabel_LFEScreen = 4,
kAudioChannelLabel_LeftSurround = 5, // WAVE: "Back Left"
kAudioChannelLabel_RightSurround = 6, // WAVE: "Back Right"
kAudioChannelLabel_LeftCenter = 7,
kAudioChannelLabel_RightCenter = 8,
kAudioChannelLabel_CenterSurround = 9, // WAVE: "Back Center" or plain "Rear Surround"
kAudioChannelLabel_LeftSurroundDirect = 10, // WAVE: "Side Left"
kAudioChannelLabel_RightSurroundDirect = 11, // WAVE: "Side Right"
kAudioChannelLabel_TopCenterSurround = 12,
kAudioChannelLabel_VerticalHeightLeft = 13, // WAVE: "Top Front Left"
kAudioChannelLabel_VerticalHeightCenter = 14, // WAVE: "Top Front Center"
kAudioChannelLabel_VerticalHeightRight = 15, // WAVE: "Top Front Right"
kAudioChannelLabel_TopBackLeft = 16,
kAudioChannelLabel_TopBackCenter = 17,
kAudioChannelLabel_TopBackRight = 18,
kAudioChannelLabel_RearSurroundLeft = 33,
kAudioChannelLabel_RearSurroundRight = 34,
kAudioChannelLabel_LeftWide = 35,
kAudioChannelLabel_RightWide = 36,
kAudioChannelLabel_LFE2 = 37,
kAudioChannelLabel_LeftTotal = 38, // matrix encoded 4 channels
kAudioChannelLabel_RightTotal = 39, // matrix encoded 4 channels
kAudioChannelLabel_HearingImpaired = 40,
kAudioChannelLabel_Narration = 41,
kAudioChannelLabel_Mono = 42,
kAudioChannelLabel_DialogCentricMix = 43,
kAudioChannelLabel_CenterSurroundDirect = 44, // back center, non diffuse
kAudioChannelLabel_Haptic = 45,
// first order ambisonic channels
kAudioChannelLabel_Ambisonic_W = 200,
kAudioChannelLabel_Ambisonic_X = 201,
kAudioChannelLabel_Ambisonic_Y = 202,
kAudioChannelLabel_Ambisonic_Z = 203,
// Mid/Side Recording
kAudioChannelLabel_MS_Mid = 204,
kAudioChannelLabel_MS_Side = 205,
// X-Y Recording
kAudioChannelLabel_XY_X = 206,
kAudioChannelLabel_XY_Y = 207,
// other
kAudioChannelLabel_HeadphonesLeft = 301,
kAudioChannelLabel_HeadphonesRight = 302,
kAudioChannelLabel_ClickTrack = 304,
kAudioChannelLabel_ForeignLanguage = 305,
// generic discrete channel
kAudioChannelLabel_Discrete = 400,
// numbered discrete channel
kAudioChannelLabel_Discrete_0 = (1L<<16) | 0,
kAudioChannelLabel_Discrete_1 = (1L<<16) | 1,
kAudioChannelLabel_Discrete_2 = (1L<<16) | 2,
kAudioChannelLabel_Discrete_3 = (1L<<16) | 3,
kAudioChannelLabel_Discrete_4 = (1L<<16) | 4,
kAudioChannelLabel_Discrete_5 = (1L<<16) | 5,
kAudioChannelLabel_Discrete_6 = (1L<<16) | 6,
kAudioChannelLabel_Discrete_7 = (1L<<16) | 7,
kAudioChannelLabel_Discrete_8 = (1L<<16) | 8,
kAudioChannelLabel_Discrete_9 = (1L<<16) | 9,
kAudioChannelLabel_Discrete_10 = (1L<<16) | 10,
kAudioChannelLabel_Discrete_11 = (1L<<16) | 11,
kAudioChannelLabel_Discrete_12 = (1L<<16) | 12,
kAudioChannelLabel_Discrete_13 = (1L<<16) | 13,
kAudioChannelLabel_Discrete_14 = (1L<<16) | 14,
kAudioChannelLabel_Discrete_15 = (1L<<16) | 15,
kAudioChannelLabel_Discrete_65535 = (1L<<16) | 65535
};
#define AudioChannelLayoutTag_GetNumberOfChannels(layoutTag) \
((uint32_t)((layoutTag) & 0x0000FFFF))
enum
{
kAudioChannelLayoutTag_UseChannelDescriptions = (0L<<16) | 0, // use the array of AudioChannelDescriptions to define the mapping.
kAudioChannelLayoutTag_UseChannelBitmap = (1L<<16) | 0, // use the bitmap to define the mapping.
kAudioChannelLayoutTag_Mono = (100L<<16) | 1, // a standard mono stream
kAudioChannelLayoutTag_Stereo = (101L<<16) | 2, // a standard stereo stream (L R) - implied playback
kAudioChannelLayoutTag_StereoHeadphones = (102L<<16) | 2, // a standard stereo stream (L R) - implied headphone playbac
kAudioChannelLayoutTag_MatrixStereo = (103L<<16) | 2, // a matrix encoded stereo stream (Lt, Rt)
kAudioChannelLayoutTag_MidSide = (104L<<16) | 2, // mid/side recording
kAudioChannelLayoutTag_XY = (105L<<16) | 2, // coincident mic pair (often 2 figure 8's)
kAudioChannelLayoutTag_Binaural = (106L<<16) | 2, // binaural stereo (left, right)
kAudioChannelLayoutTag_Ambisonic_B_Format = (107L<<16) | 4, // W, X, Y, Z
kAudioChannelLayoutTag_Quadraphonic = (108L<<16) | 4, // front left, front right, back left, back right
kAudioChannelLayoutTag_Pentagonal = (109L<<16) | 5, // left, right, rear left, rear right, center
kAudioChannelLayoutTag_Hexagonal = (110L<<16) | 6, // left, right, rear left, rear right, center, rear
kAudioChannelLayoutTag_Octagonal = (111L<<16) | 8, // front left, front right, rear left, rear right,
// front center, rear center, side left, side right
kAudioChannelLayoutTag_Cube = (112L<<16) | 8, // left, right, rear left, rear right
// top left, top right, top rear left, top rear right
// MPEG defined layouts
kAudioChannelLayoutTag_MPEG_1_0 = kAudioChannelLayoutTag_Mono, // C
kAudioChannelLayoutTag_MPEG_2_0 = kAudioChannelLayoutTag_Stereo, // L R
kAudioChannelLayoutTag_MPEG_3_0_A = (113L<<16) | 3, // L R C
kAudioChannelLayoutTag_MPEG_3_0_B = (114L<<16) | 3, // C L R
kAudioChannelLayoutTag_MPEG_4_0_A = (115L<<16) | 4, // L R C Cs
kAudioChannelLayoutTag_MPEG_4_0_B = (116L<<16) | 4, // C L R Cs
kAudioChannelLayoutTag_MPEG_5_0_A = (117L<<16) | 5, // L R C Ls Rs
kAudioChannelLayoutTag_MPEG_5_0_B = (118L<<16) | 5, // L R Ls Rs C
kAudioChannelLayoutTag_MPEG_5_0_C = (119L<<16) | 5, // L C R Ls Rs
kAudioChannelLayoutTag_MPEG_5_0_D = (120L<<16) | 5, // C L R Ls Rs
kAudioChannelLayoutTag_MPEG_5_1_A = (121L<<16) | 6, // L R C LFE Ls Rs
kAudioChannelLayoutTag_MPEG_5_1_B = (122L<<16) | 6, // L R Ls Rs C LFE
kAudioChannelLayoutTag_MPEG_5_1_C = (123L<<16) | 6, // L C R Ls Rs LFE
kAudioChannelLayoutTag_MPEG_5_1_D = (124L<<16) | 6, // C L R Ls Rs LFE
kAudioChannelLayoutTag_MPEG_6_1_A = (125L<<16) | 7, // L R C LFE Ls Rs Cs
kAudioChannelLayoutTag_MPEG_7_1_A = (126L<<16) | 8, // L R C LFE Ls Rs Lc Rc
kAudioChannelLayoutTag_MPEG_7_1_B = (127L<<16) | 8, // C Lc Rc L R Ls Rs LFE (doc: IS-13818-7 MPEG2-AAC Table 3.1)
kAudioChannelLayoutTag_MPEG_7_1_C = (128L<<16) | 8, // L R C LFE Ls Rs Rls Rrs
kAudioChannelLayoutTag_Emagic_Default_7_1 = (129L<<16) | 8, // L R Ls Rs C LFE Lc Rc
kAudioChannelLayoutTag_SMPTE_DTV = (130L<<16) | 8, // L R C LFE Ls Rs Lt Rt
// (kAudioChannelLayoutTag_ITU_5_1 plus a matrix encoded stereo mix)
// ITU defined layouts
kAudioChannelLayoutTag_ITU_1_0 = kAudioChannelLayoutTag_Mono, // C
kAudioChannelLayoutTag_ITU_2_0 = kAudioChannelLayoutTag_Stereo, // L R
kAudioChannelLayoutTag_ITU_2_1 = (131L<<16) | 3, // L R Cs
kAudioChannelLayoutTag_ITU_2_2 = (132L<<16) | 4, // L R Ls Rs
kAudioChannelLayoutTag_ITU_3_0 = kAudioChannelLayoutTag_MPEG_3_0_A, // L R C
kAudioChannelLayoutTag_ITU_3_1 = kAudioChannelLayoutTag_MPEG_4_0_A, // L R C Cs
kAudioChannelLayoutTag_ITU_3_2 = kAudioChannelLayoutTag_MPEG_5_0_A, // L R C Ls Rs
kAudioChannelLayoutTag_ITU_3_2_1 = kAudioChannelLayoutTag_MPEG_5_1_A, // L R C LFE Ls Rs
kAudioChannelLayoutTag_ITU_3_4_1 = kAudioChannelLayoutTag_MPEG_7_1_C, // L R C LFE Ls Rs Rls Rrs
// DVD defined layouts
kAudioChannelLayoutTag_DVD_0 = kAudioChannelLayoutTag_Mono, // C (mono)
kAudioChannelLayoutTag_DVD_1 = kAudioChannelLayoutTag_Stereo, // L R
kAudioChannelLayoutTag_DVD_2 = kAudioChannelLayoutTag_ITU_2_1, // L R Cs
kAudioChannelLayoutTag_DVD_3 = kAudioChannelLayoutTag_ITU_2_2, // L R Ls Rs
kAudioChannelLayoutTag_DVD_4 = (133L<<16) | 3, // L R LFE
kAudioChannelLayoutTag_DVD_5 = (134L<<16) | 4, // L R LFE Cs
kAudioChannelLayoutTag_DVD_6 = (135L<<16) | 5, // L R LFE Ls Rs
kAudioChannelLayoutTag_DVD_7 = kAudioChannelLayoutTag_MPEG_3_0_A, // L R C
kAudioChannelLayoutTag_DVD_8 = kAudioChannelLayoutTag_MPEG_4_0_A, // L R C Cs
kAudioChannelLayoutTag_DVD_9 = kAudioChannelLayoutTag_MPEG_5_0_A, // L R C Ls Rs
kAudioChannelLayoutTag_DVD_10 = (136L<<16) | 4, // L R C LFE
kAudioChannelLayoutTag_DVD_11 = (137L<<16) | 5, // L R C LFE Cs
kAudioChannelLayoutTag_DVD_12 = kAudioChannelLayoutTag_MPEG_5_1_A, // L R C LFE Ls Rs
// 13 through 17 are duplicates of 8 through 12.
kAudioChannelLayoutTag_DVD_13 = kAudioChannelLayoutTag_DVD_8, // L R C Cs
kAudioChannelLayoutTag_DVD_14 = kAudioChannelLayoutTag_DVD_9, // L R C Ls Rs
kAudioChannelLayoutTag_DVD_15 = kAudioChannelLayoutTag_DVD_10, // L R C LFE
kAudioChannelLayoutTag_DVD_16 = kAudioChannelLayoutTag_DVD_11, // L R C LFE Cs
kAudioChannelLayoutTag_DVD_17 = kAudioChannelLayoutTag_DVD_12, // L R C LFE Ls Rs
kAudioChannelLayoutTag_DVD_18 = (138L<<16) | 5, // L R Ls Rs LFE
kAudioChannelLayoutTag_DVD_19 = kAudioChannelLayoutTag_MPEG_5_0_B, // L R Ls Rs C
kAudioChannelLayoutTag_DVD_20 = kAudioChannelLayoutTag_MPEG_5_1_B, // L R Ls Rs C LFE
// These layouts are recommended for AudioUnit usage
// These are the symmetrical layouts
kAudioChannelLayoutTag_AudioUnit_4 = kAudioChannelLayoutTag_Quadraphonic,
kAudioChannelLayoutTag_AudioUnit_5 = kAudioChannelLayoutTag_Pentagonal,
kAudioChannelLayoutTag_AudioUnit_6 = kAudioChannelLayoutTag_Hexagonal,
kAudioChannelLayoutTag_AudioUnit_8 = kAudioChannelLayoutTag_Octagonal,
// These are the surround-based layouts
kAudioChannelLayoutTag_AudioUnit_5_0 = kAudioChannelLayoutTag_MPEG_5_0_B, // L R Ls Rs C
kAudioChannelLayoutTag_AudioUnit_6_0 = (139L<<16) | 6, // L R Ls Rs C Cs
kAudioChannelLayoutTag_AudioUnit_7_0 = (140L<<16) | 7, // L R Ls Rs C Rls Rrs
kAudioChannelLayoutTag_AudioUnit_7_0_Front = (148L<<16) | 7, // L R Ls Rs C Lc Rc
kAudioChannelLayoutTag_AudioUnit_5_1 = kAudioChannelLayoutTag_MPEG_5_1_A, // L R C LFE Ls Rs
kAudioChannelLayoutTag_AudioUnit_6_1 = kAudioChannelLayoutTag_MPEG_6_1_A, // L R C LFE Ls Rs Cs
kAudioChannelLayoutTag_AudioUnit_7_1 = kAudioChannelLayoutTag_MPEG_7_1_C, // L R C LFE Ls Rs Rls Rrs
kAudioChannelLayoutTag_AudioUnit_7_1_Front = kAudioChannelLayoutTag_MPEG_7_1_A, // L R C LFE Ls Rs Lc Rc
kAudioChannelLayoutTag_AAC_3_0 = kAudioChannelLayoutTag_MPEG_3_0_B, // C L R
kAudioChannelLayoutTag_AAC_Quadraphonic = kAudioChannelLayoutTag_Quadraphonic, // L R Ls Rs
kAudioChannelLayoutTag_AAC_4_0 = kAudioChannelLayoutTag_MPEG_4_0_B, // C L R Cs
kAudioChannelLayoutTag_AAC_5_0 = kAudioChannelLayoutTag_MPEG_5_0_D, // C L R Ls Rs
kAudioChannelLayoutTag_AAC_5_1 = kAudioChannelLayoutTag_MPEG_5_1_D, // C L R Ls Rs Lfe
kAudioChannelLayoutTag_AAC_6_0 = (141L<<16) | 6, // C L R Ls Rs Cs
kAudioChannelLayoutTag_AAC_6_1 = (142L<<16) | 7, // C L R Ls Rs Cs Lfe
kAudioChannelLayoutTag_AAC_7_0 = (143L<<16) | 7, // C L R Ls Rs Rls Rrs
kAudioChannelLayoutTag_AAC_7_1 = kAudioChannelLayoutTag_MPEG_7_1_B, // C Lc Rc L R Ls Rs Lfe
kAudioChannelLayoutTag_AAC_Octagonal = (144L<<16) | 8, // C L R Ls Rs Rls Rrs Cs
kAudioChannelLayoutTag_TMH_10_2_std = (145L<<16) | 16, // L R C Vhc Lsd Rsd Ls Rs Vhl Vhr Lw Rw Csd Cs LFE1 LFE2
kAudioChannelLayoutTag_TMH_10_2_full = (146L<<16) | 21, // TMH_10_2_std plus: Lc Rc HI VI Haptic
kAudioChannelLayoutTag_AC3_1_0_1 = (149L<<16) | 2, // C LFE
kAudioChannelLayoutTag_AC3_3_0 = (150L<<16) | 3, // L C R
kAudioChannelLayoutTag_AC3_3_1 = (151L<<16) | 4, // L C R Cs
kAudioChannelLayoutTag_AC3_3_0_1 = (152L<<16) | 4, // L C R LFE
kAudioChannelLayoutTag_AC3_2_1_1 = (153L<<16) | 4, // L R Cs LFE
kAudioChannelLayoutTag_AC3_3_1_1 = (154L<<16) | 5, // L C R Cs LFE
kAudioChannelLayoutTag_DiscreteInOrder = (147L<<16) | 0, // needs to be ORed with the actual number of channels
kAudioChannelLayoutTag_Unknown = 0xFFFF0000 // needs to be ORed with the actual number of channels
};
#endif

View File

@ -34,6 +34,7 @@
#endif
#include "compat.h"
#include "wav_reader.h"
#include "caf_reader.h"
#include "aacenc.h"
#include "m4af.h"
#include "progress.h"
@ -216,7 +217,9 @@ typedef struct aacenc_param_ex_t {
unsigned raw_rate;
const char *raw_format;
aacenc_tag_param_t tags;
aacenc_tag_store_t tags;
aacenc_tag_store_t source_tags;
aacenc_translate_generic_text_tag_ctx_t source_tag_ctx;
char *json_filename;
} aacenc_param_ex_t;
@ -398,8 +401,8 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
case M4AF_TAG_TRACK:
case M4AF_TAG_DISK:
case M4AF_TAG_TEMPO:
aacenc_param_add_itmf_entry(&params->tags, ch, 0, optarg,
strlen(optarg), 0);
aacenc_add_tag_to_store(&params->tags, ch, 0, optarg,
strlen(optarg), 0);
break;
case OPT_SHORT_TAG:
case OPT_SHORT_TAG_FILE:
@ -432,9 +435,9 @@ int parse_options(int argc, char **argv, aacenc_param_ex_t *params)
for (; *optarg; ++optarg)
fcc = ((fcc << 8) | (*optarg & 0xff));
}
aacenc_param_add_itmf_entry(&params->tags, fcc, optarg,
val, strlen(val),
ch == OPT_SHORT_TAG_FILE);
aacenc_add_tag_to_store(&params->tags, fcc, optarg,
val, strlen(val),
ch == OPT_SHORT_TAG_FILE);
}
break;
case OPT_TAG_FROM_JSON:
@ -570,13 +573,18 @@ int finalize_m4a(m4af_ctx_t *m4af, const aacenc_param_ex_t *params,
HANDLE_AACENCODER encoder)
{
unsigned i;
aacenc_tag_entry_t *tag = params->tags.tag_table;
aacenc_tag_entry_t *tag;
tag = params->source_tags.tag_table;
for (i = 0; i < params->source_tags.tag_count; ++i, ++tag)
aacenc_write_tag_entry(m4af, tag);
if (params->json_filename)
aacenc_put_tags_from_json(m4af, params->json_filename);
aacenc_write_tags_from_json(m4af, params->json_filename);
tag = params->tags.tag_table;
for (i = 0; i < params->tags.tag_count; ++i, ++tag)
aacenc_put_tag_entry(m4af, tag);
aacenc_write_tag_entry(m4af, tag);
put_tool_tag(m4af, params, encoder);
@ -643,12 +651,15 @@ int parse_raw_spec(const char *spec, pcm_sample_description_t *desc)
return 0;
}
static pcm_io_vtbl_t pcm_io_vtbl = {
read_callback, seek_callback, tell_callback
};
static pcm_io_vtbl_t pcm_io_vtbl_noseek = { read_callback, 0, 0 };
static
pcm_reader_t *open_input(aacenc_param_ex_t *params)
{
wav_io_context_t wav_io = {
read_callback, seek_callback, tell_callback
};
pcm_io_context_t io = { 0 };
pcm_reader_t *reader = 0;
struct stat stb = { 0 };
@ -657,11 +668,13 @@ pcm_reader_t *open_input(aacenc_param_ex_t *params)
strerror(errno));
goto END;
}
if (fstat(fileno(params->input_fp), &stb) == 0
&& (stb.st_mode & S_IFMT) != S_IFREG) {
wav_io.seek = 0;
wav_io.tell = 0;
}
io.cookie = params->input_fp;
if (fstat(fileno(io.cookie), &stb) == 0
&& (stb.st_mode & S_IFMT) == S_IFREG)
io.vtbl = &pcm_io_vtbl;
else
io.vtbl = &pcm_io_vtbl_noseek;
if (params->is_raw) {
int bytes_per_channel;
pcm_sample_description_t desc = { 0 };
@ -673,14 +686,32 @@ pcm_reader_t *open_input(aacenc_param_ex_t *params)
desc.channels_per_frame = params->raw_channels;
bytes_per_channel = (desc.bits_per_channel + 7) / 8;
desc.bytes_per_frame = params->raw_channels * bytes_per_channel;
if ((reader = raw_open(&wav_io, params->input_fp, &desc)) == 0) {
if ((reader = raw_open(&io, &desc)) == 0) {
fprintf(stderr, "ERROR: failed to open raw input\n");
goto END;
}
} else {
if ((reader = wav_open(&wav_io, params->input_fp,
params->ignore_length)) == 0) {
fprintf(stderr, "ERROR: broken / unsupported input file\n");
int c;
ungetc(c = getc(params->input_fp), params->input_fp);
switch (c) {
case 'R':
if ((reader = wav_open(&io, params->ignore_length)) == 0) {
fprintf(stderr, "ERROR: broken / unsupported input file\n");
goto END;
}
break;
case 'c':
params->source_tag_ctx.add = aacenc_add_tag_entry_to_store;
params->source_tag_ctx.add_ctx = &params->source_tags;
if ((reader = caf_open(&io,
aacenc_translate_generic_text_tag,
&params->source_tag_ctx)) == 0) {
fprintf(stderr, "ERROR: broken / unsupported input file\n");
goto END;
}
break;
default:
goto END;
}
}
@ -774,7 +805,10 @@ END:
if (params.output_fp) fclose(params.output_fp);
if (encoder) aacEncClose(&encoder);
if (output_filename) free(output_filename);
if (params.tags.tag_table) free(params.tags.tag_table);
if (params.tags.tag_table)
aacenc_free_tag_store(&params.tags);
if (params.source_tags.tag_table)
aacenc_free_tag_store(&params.source_tags);
return result;
}

View File

@ -50,12 +50,13 @@ static tag_key_mapping_t tag_mapping_table[] = {
{ "copyright", M4AF_TAG_COPYRIGHT },
{ "date", M4AF_TAG_DATE },
{ "disc", M4AF_TAG_DISK },
{ "disctotal", TAG_TOTAL_DISCS },
{ "discnumber", M4AF_TAG_DISK },
{ "disctotal", TAG_TOTAL_DISCS },
{ "genre", M4AF_TAG_GENRE },
{ "grouping", M4AF_TAG_GROUPING },
{ "itunescompilation", M4AF_TAG_COMPILATION },
{ "lyrics", M4AF_TAG_LYRICS },
{ "performer", M4AF_TAG_ARTIST },
{ "title", M4AF_TAG_TITLE },
{ "titlesort", M4AF_FOURCC('s','o','n','m') },
{ "titlesortorder", M4AF_FOURCC('s','o','n','m') },
@ -95,6 +96,7 @@ uint32_t get_tag_fcc_from_name(const char *name)
return ent ? ent->fcc : 0;
}
static
char *aacenc_load_tag_from_file(const char *path, uint32_t *data_size)
{
FILE *fp = 0;
@ -121,48 +123,152 @@ END:
return data;
}
void aacenc_param_add_itmf_entry(aacenc_tag_param_t *params, uint32_t tag,
const char *key, const char *value,
uint32_t size, int is_file_name)
static
int aacenc_is_string_tag(uint32_t tag)
{
aacenc_tag_entry_t *entry;
switch (tag) {
case M4AF_TAG_TITLE:
case M4AF_TAG_ARTIST:
case M4AF_TAG_ALBUM:
case M4AF_TAG_GENRE:
case M4AF_TAG_DATE:
case M4AF_TAG_COMPOSER:
case M4AF_TAG_GROUPING:
case M4AF_TAG_COMMENT:
case M4AF_TAG_LYRICS:
case M4AF_TAG_TOOL:
case M4AF_TAG_ALBUM_ARTIST:
case M4AF_TAG_DESCRIPTION:
case M4AF_TAG_LONG_DESCRIPTION:
case M4AF_TAG_COPYRIGHT:
case M4AF_FOURCC('a','p','I','D'):
case M4AF_FOURCC('c','a','t','g'):
case M4AF_FOURCC('k','e','y','w'):
case M4AF_FOURCC('p','u','r','d'):
case M4AF_FOURCC('p','u','r','l'):
case M4AF_FOURCC('s','o','a','a'):
case M4AF_FOURCC('s','o','a','l'):
case M4AF_FOURCC('s','o','a','r'):
case M4AF_FOURCC('s','o','c','o'):
case M4AF_FOURCC('s','o','n','m'):
case M4AF_FOURCC('s','o','s','n'):
case M4AF_FOURCC('t','v','e','n'):
case M4AF_FOURCC('t','v','n','n'):
case M4AF_FOURCC('t','v','s','h'):
case M4AF_FOURCC('x','i','d',' '):
case M4AF_FOURCC('\xa9','e','n','c'):
case M4AF_FOURCC('\xa9','s','t','3'):
case M4AF_FOURCC('-','-','-','-'):
return 1;
}
return 0;
}
void aacenc_add_tag_to_store(aacenc_tag_store_t *store, uint32_t tag,
const char *key, const char *value,
uint32_t size, int is_file_name)
{
aacenc_tag_entry_t entry = { 0 };
char *dp = 0;
if (!is_file_name && !size)
return;
if (params->tag_count == params->tag_table_capacity) {
unsigned newsize = params->tag_table_capacity;
newsize = newsize ? newsize * 2 : 1;
params->tag_table =
realloc(params->tag_table, newsize * sizeof(aacenc_tag_entry_t));
params->tag_table_capacity = newsize;
}
entry = params->tag_table + params->tag_count;
entry->tag = tag;
entry.tag = tag;
if (tag == M4AF_FOURCC('-','-','-','-'))
entry->name = key;
entry->data = value;
entry->data_size = size;
entry->is_file_name = is_file_name;
params->tag_count++;
entry.name = (char *)key;
if (is_file_name) {
entry.data = dp = aacenc_load_tag_from_file(value, &size);
entry.data_size = size;
} else if (aacenc_is_string_tag(tag)) {
entry.data = dp = aacenc_to_utf8(value);
entry.data_size = strlen(entry.data);
} else {
entry.data = (char *)value;
entry.data_size = size;
}
aacenc_add_tag_entry_to_store(store, &entry);
free(dp);
}
void aacenc_add_tag_entry_to_store(void *ctx, const aacenc_tag_entry_t *tag)
{
aacenc_tag_store_t *store = ctx;
aacenc_tag_entry_t *entry;
if (store->tag_count == store->tag_table_capacity) {
unsigned newsize = store->tag_table_capacity;
newsize = newsize ? newsize * 2 : 1;
store->tag_table =
realloc(store->tag_table, newsize * sizeof(aacenc_tag_entry_t));
store->tag_table_capacity = newsize;
}
entry = store->tag_table + store->tag_count;
entry->tag = tag->tag;
entry->data_size = tag->data_size;
entry->name = tag->name ? strdup(tag->name) : 0;
entry->data = malloc(tag->data_size + 1);
memcpy(entry->data, tag->data, tag->data_size);
entry->data[tag->data_size] = 0;
store->tag_count++;
}
static
void tag_put_number_pair(m4af_ctx_t *m4af, uint32_t fcc,
const char *snumber, const char *stotal)
void tag_put_number_pair(aacenc_translate_generic_text_tag_ctx_t *ctx,
uint32_t fcc, unsigned number, unsigned total)
{
unsigned number = 0, total = 0;
char buf[128];
aacenc_tag_entry_t entry = { 0 };
if (snumber) sscanf(snumber, "%u", &number);
if (stotal) sscanf(stotal, "%u", &total);
if (number) {
if (total) sprintf(buf, "%u/%u", number, total);
else sprintf(buf, "%u", number);
entry.tag = fcc;
entry.data = buf;
entry.data_size = strlen(buf);
aacenc_put_tag_entry(m4af, &entry);
ctx->add(ctx->add_ctx, &entry);
}
}
void aacenc_translate_generic_text_tag(void *pctx, const char *key,
const char *val, uint32_t size)
{
aacenc_translate_generic_text_tag_ctx_t *ctx = pctx;
aacenc_tag_entry_t entry = { 0 };
uint32_t fcc;
/*
* Since track/disc number pair (number and total) can be stored within
* either single tag or separately, we cannot instantly translate
* them in one-to-one manner.
* Instead, we keep and store them until all tags are processed,
* then finally submit.
*/
if (!key) {
/* use null key as flushing signal */
tag_put_number_pair(ctx, M4AF_TAG_TRACK, ctx->track, ctx->track_total);
tag_put_number_pair(ctx, M4AF_TAG_DISK, ctx->disc, ctx->disc_total);
return;
}
if (!val || !size)
return;
if ((fcc = get_tag_fcc_from_name(key)) == 0)
return;
switch (fcc) {
case TAG_TOTAL_DISCS:
sscanf(val, "%d", &ctx->disc_total); break;
case TAG_TOTAL_TRACKS:
sscanf(val, "%d", &ctx->track_total); break;
case M4AF_TAG_DISK:
sscanf(val, "%d/%d", &ctx->disc, &ctx->disc_total); break;
case M4AF_TAG_TRACK:
sscanf(val, "%d/%d", &ctx->track, &ctx->track_total); break;
default:
{
entry.tag = fcc;
entry.data = (char *)val;
entry.data_size = (size == ~0U) ? strlen(val) : size;
ctx->add(ctx->add_ctx, &entry);
}
}
}
@ -188,7 +294,7 @@ const char *aacenc_json_object_get_string(JSON_Object *obj, const char *key,
return val;
}
void aacenc_put_tags_from_json(m4af_ctx_t *m4af, const char *json_filename)
void aacenc_write_tags_from_json(m4af_ctx_t *m4af, const char *json_filename)
{
char *data = 0;
JSON_Value *json = 0;
@ -197,12 +303,11 @@ void aacenc_put_tags_from_json(m4af_ctx_t *m4af, const char *json_filename)
uint32_t data_size;
char *json_dot_path;
char *filename = 0;
char *disc = 0;
char *track = 0;
char *total_discs = 0;
char *total_tracks = 0;
aacenc_tag_entry_t entry = { 0 };
aacenc_translate_generic_text_tag_ctx_t ctx = { 0 };
ctx.add = aacenc_write_tag_entry;
ctx.add_ctx = m4af;
filename = strdup(json_filename);
if ((json_dot_path = strchr(filename, '?')) != 0)
*json_dot_path++ = '\0';
@ -226,59 +331,36 @@ void aacenc_put_tags_from_json(m4af_ctx_t *m4af, const char *json_filename)
char buf[256];
const char *key = json_object_get_name(root, i);
const char *val = aacenc_json_object_get_string(root, key, buf);
uint32_t fcc = get_tag_fcc_from_name(key);
if (!val || !fcc)
continue;
switch (fcc) {
case TAG_TOTAL_DISCS:
total_discs = realloc(total_discs, strlen(val) + 1);
strcpy(total_discs, val);
break;
case TAG_TOTAL_TRACKS:
total_tracks = realloc(total_tracks, strlen(val) + 1);
strcpy(total_tracks, val);
break;
case M4AF_TAG_DISK:
disc = realloc(disc, strlen(val) + 1);
strcpy(disc, val);
break;
case M4AF_TAG_TRACK:
track = realloc(track, strlen(val) + 1);
strcpy(track, val);
break;
default:
{
entry.tag = fcc;
entry.data = val;
entry.data_size = strlen(val);
aacenc_put_tag_entry(m4af, &entry);
}
}
if (val) aacenc_translate_generic_text_tag(&ctx, key, val, ~0U);
}
tag_put_number_pair(m4af, M4AF_TAG_TRACK, track, total_tracks);
tag_put_number_pair(m4af, M4AF_TAG_DISK, disc, total_discs);
aacenc_translate_generic_text_tag(&ctx, 0, 0, 0);
DONE:
if (track) free(track);
if (disc) free(disc);
if (total_tracks) free(total_tracks);
if (total_discs) free(total_discs);
if (data) free(data);
if (filename) free(filename);
if (json) json_value_free(json);
}
void aacenc_put_tag_entry(m4af_ctx_t *m4af, const aacenc_tag_entry_t *tag)
void aacenc_free_tag_store(aacenc_tag_store_t *store)
{
if (store->tag_table) {
unsigned i;
for (i = 0; i < store->tag_count; ++i) {
aacenc_tag_entry_t *ent = &store->tag_table[i];
free(ent->name);
free(ent->data);
}
free(store->tag_table);
store->tag_table = 0;
store->tag_count = 0;
}
}
void aacenc_write_tag_entry(void *ctx, const aacenc_tag_entry_t *tag)
{
m4af_ctx_t *m4af = ctx;
unsigned m, n = 0;
const char *data = tag->data;
uint32_t data_size = tag->data_size;
char *file_contents = 0;
if (tag->is_file_name) {
data = file_contents = aacenc_load_tag_from_file(tag->data, &data_size);
if (!data) return;
}
switch (tag->tag) {
case M4AF_TAG_TRACK:
if (sscanf(data, "%u/%u", &m, &n) >= 1)
@ -334,59 +416,22 @@ void aacenc_put_tag_entry(m4af_ctx_t *m4af, const aacenc_tag_entry_t *tag)
data_type = M4AF_PNG;
if (data_type)
m4af_add_itmf_short_tag(m4af, tag->tag, data_type,
data, data_size);
data, tag->data_size);
break;
}
case M4AF_FOURCC('-','-','-','-'):
{
char *u8 = aacenc_to_utf8(data);
m4af_add_itmf_long_tag(m4af, tag->name, u8);
free(u8);
break;
}
case M4AF_TAG_TITLE:
case M4AF_TAG_ARTIST:
case M4AF_TAG_ALBUM:
case M4AF_TAG_GENRE:
case M4AF_TAG_DATE:
case M4AF_TAG_COMPOSER:
case M4AF_TAG_GROUPING:
case M4AF_TAG_COMMENT:
case M4AF_TAG_LYRICS:
case M4AF_TAG_TOOL:
case M4AF_TAG_ALBUM_ARTIST:
case M4AF_TAG_DESCRIPTION:
case M4AF_TAG_LONG_DESCRIPTION:
case M4AF_TAG_COPYRIGHT:
case M4AF_FOURCC('a','p','I','D'):
case M4AF_FOURCC('c','a','t','g'):
case M4AF_FOURCC('k','e','y','w'):
case M4AF_FOURCC('p','u','r','d'):
case M4AF_FOURCC('p','u','r','l'):
case M4AF_FOURCC('s','o','a','a'):
case M4AF_FOURCC('s','o','a','l'):
case M4AF_FOURCC('s','o','a','r'):
case M4AF_FOURCC('s','o','c','o'):
case M4AF_FOURCC('s','o','n','m'):
case M4AF_FOURCC('s','o','s','n'):
case M4AF_FOURCC('t','v','e','n'):
case M4AF_FOURCC('t','v','n','n'):
case M4AF_FOURCC('t','v','s','h'):
case M4AF_FOURCC('x','i','d',' '):
case M4AF_FOURCC('\xa9','e','n','c'):
case M4AF_FOURCC('\xa9','s','t','3'):
{
char *u8 = aacenc_to_utf8(data);
m4af_add_itmf_string_tag(m4af, tag->tag, u8);
free(u8);
m4af_add_itmf_long_tag(m4af, tag->name, data);
break;
}
default:
fprintf(stderr, "WARNING: unknown/unsupported tag: %c%c%c%c\n",
tag->tag >> 24, (tag->tag >> 16) & 0xff,
(tag->tag >> 8) & 0xff, tag->tag & 0xff);
if (aacenc_is_string_tag(tag->tag))
m4af_add_itmf_string_tag(m4af, tag->tag, data);
else
fprintf(stderr, "WARNING: unknown/unsupported tag: %c%c%c%c\n",
tag->tag >> 24, (tag->tag >> 16) & 0xff,
(tag->tag >> 8) & 0xff, tag->tag & 0xff);
}
if (file_contents) free(file_contents);
}

View File

@ -1,28 +1,44 @@
#ifndef METADATA_H
#define METADATA_H
#include "m4af.h"
typedef struct aacenc_tag_entry_t {
uint32_t tag;
const char *name;
const char *data;
uint32_t data_size;
int is_file_name;
uint32_t tag;
char *name;
char *data;
uint32_t data_size;
} aacenc_tag_entry_t;
typedef struct aacenc_tag_param_t {
typedef struct aacenc_tag_store_t {
aacenc_tag_entry_t *tag_table;
unsigned tag_count;
unsigned tag_table_capacity;
} aacenc_tag_param_t;
} aacenc_tag_store_t;
char *aacenc_load_tag_from_file(const char *path, uint32_t *data_size);
typedef struct aacenc_translate_generic_text_tag_ctx_t {
unsigned track, track_total, disc, disc_total;
void (*add)(void *, const aacenc_tag_entry_t *);
void *add_ctx;
} aacenc_translate_generic_text_tag_ctx_t;
void aacenc_param_add_itmf_entry(aacenc_tag_param_t *params, uint32_t tag,
const char *key, const char *value,
uint32_t size, int is_file_name);
typedef void (*aacenc_tag_callback_t)(void *ctx, const char *key,
const char *value, uint32_t size);
void aacenc_put_tags_from_json(m4af_ctx_t *m4af, const char *json_filename);
void aacenc_translate_generic_text_tag(void *ctx, const char *key,
const char *val, uint32_t size);
void aacenc_put_tag_entry(m4af_ctx_t *m4af, const aacenc_tag_entry_t *tag);
void aacenc_add_tag_to_store(aacenc_tag_store_t *store, uint32_t tag,
const char *key, const char *value,
uint32_t size, int is_file_name);
void aacenc_add_tag_entry_to_store(void *store, const aacenc_tag_entry_t *tag);
void aacenc_free_tag_store(aacenc_tag_store_t *store);
void aacenc_write_tags_from_json(m4af_ctx_t *m4af, const char *json_filename);
void aacenc_write_tag_entry(void *m4af, const aacenc_tag_entry_t *tag);
#endif

View File

@ -17,6 +17,21 @@ struct pcm_reader_t {
pcm_reader_vtbl_t *vtbl;
};
typedef int (*pcm_read_callback)(void *cookie, void *data, uint32_t count);
typedef int (*pcm_seek_callback)(void *cookie, int64_t off, int whence);
typedef int64_t (*pcm_tell_callback)(void *cookie);
typedef struct pcm_io_vtbl_t {
pcm_read_callback read;
pcm_seek_callback seek;
pcm_tell_callback tell;
} pcm_io_vtbl_t;
typedef struct pcm_io_context_t {
pcm_io_vtbl_t *vtbl;
void *cookie;
} pcm_io_context_t;
static inline
const pcm_sample_description_t *pcm_get_format(pcm_reader_t *r)
{
@ -47,6 +62,51 @@ void pcm_teardown(pcm_reader_t **r)
(*r)->vtbl->teardown(r);
}
static inline
uint32_t bitcount(uint32_t bits)
{
bits = (bits & 0x55555555) + (bits >> 1 & 0x55555555);
bits = (bits & 0x33333333) + (bits >> 2 & 0x33333333);
bits = (bits & 0x0f0f0f0f) + (bits >> 4 & 0x0f0f0f0f);
bits = (bits & 0x00ff00ff) + (bits >> 8 & 0x00ff00ff);
return (bits & 0x0000ffff) + (bits >>16 & 0x0000ffff);
}
#define TRY_IO(expr) \
do { \
if ((expr)) goto FAIL; \
} while (0)
#define ENSURE(expr) \
do { \
if (!(expr)) goto FAIL;\
} while (0)
int pcm_read(pcm_io_context_t *io, void *buffer, uint32_t size);
int pcm_skip(pcm_io_context_t *io, int64_t count);
static int pcm_seek(pcm_io_context_t *io, int64_t off, int whence)
{
return io->vtbl->seek ? io->vtbl->seek(io->cookie, off, whence) : -1;
}
static inline int64_t pcm_tell(pcm_io_context_t *io)
{
return io->vtbl->tell ? io->vtbl->tell(io->cookie) : -1;
}
int pcm_read16le(pcm_io_context_t *io, uint16_t *value);
int pcm_read16be(pcm_io_context_t *io, uint16_t *value);
int pcm_read32le(pcm_io_context_t *io, uint32_t *value);
int pcm_read32be(pcm_io_context_t *io, uint32_t *value);
int pcm_read64le(pcm_io_context_t *io, uint64_t *value);
int pcm_read64be(pcm_io_context_t *io, uint64_t *value);
int pcm_scanl(pcm_io_context_t *io, const char *fmt, ...);
int pcm_scanb(pcm_io_context_t *io, const char *fmt, ...);
int apple_chan_chunk(pcm_io_context_t *io, uint32_t chunk_size,
pcm_sample_description_t *fmt, uint8_t *mapping);
pcm_reader_t *pcm_open_sint16_converter(pcm_reader_t *reader);
#endif

356
src/pcm_readhelper.c Normal file
View File

@ -0,0 +1,356 @@
/*
* Copyright (C) 2013 nu774
* For conditions of distribution and use, see copyright notice in COPYING
*/
#if HAVE_CONFIG_H
# include "config.h"
#endif
#if HAVE_STDINT_H
# include <stdint.h>
#endif
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdarg.h>
#include "pcm_reader.h"
#include "m4af_endian.h"
#include "catypes.h"
int pcm_read(pcm_io_context_t *io, void *buffer, uint32_t size)
{
int rc;
uint32_t count = 0;
do {
rc = io->vtbl->read(io->cookie, buffer, size - count);
if (rc > 0)
count += rc;
} while (rc > 0 && count < size);
return count > 0 ? count : rc;
}
int pcm_skip(pcm_io_context_t *io, int64_t count)
{
char buff[8192];
int rc;
pcm_io_vtbl_t *vp = io->vtbl;
if (count == 0 || pcm_seek(io, count, SEEK_CUR) >= 0)
return 0;
do {
if ((rc = vp->read(io->cookie, buff, count > 8192 ? 8192 : count)) > 0)
count -= rc;
} while (rc > 0 && count > 0);
return count == 0 ? 0 : -1;
}
int pcm_read16le(pcm_io_context_t *io, uint16_t *value)
{
if (pcm_read(io, value, 2) == 2) {
*value = m4af_ltoh16(*value);
return 0;
}
return -1;
}
int pcm_read16be(pcm_io_context_t *io, uint16_t *value)
{
if (pcm_read(io, value, 2) == 2) {
*value = m4af_btoh16(*value);
return 0;
}
return -1;
}
int pcm_read32le(pcm_io_context_t *io, uint32_t *value)
{
if (pcm_read(io, value, 4) == 4) {
*value = m4af_ltoh32(*value);
return 0;
}
return -1;
}
int pcm_read32be(pcm_io_context_t *io, uint32_t *value)
{
if (pcm_read(io, value, 4) == 4) {
*value = m4af_btoh32(*value);
return 0;
}
return -1;
}
int pcm_read64le(pcm_io_context_t *io, uint64_t *value)
{
if (pcm_read(io, value, 8) == 8) {
*value = m4af_ltoh64(*value);
return 0;
}
return -1;
}
int pcm_read64be(pcm_io_context_t *io, uint64_t *value)
{
if (pcm_read(io, value, 8) == 8) {
*value = m4af_btoh64(*value);
return 0;
}
return -1;
}
int pcm_scanl(pcm_io_context_t *io, const char *fmt, ...)
{
int c, count = 0;
va_list ap;
va_start(ap, fmt);
while ((c = *fmt++)) {
switch (c) {
case 'S':
TRY_IO(pcm_read16le(io, va_arg(ap, uint16_t*)));
++count;
break;
case 'L':
TRY_IO(pcm_read32le(io, va_arg(ap, uint32_t*)));
++count;
break;
case 'Q':
TRY_IO(pcm_read64le(io, va_arg(ap, uint64_t*)));
++count;
break;
}
}
FAIL:
va_end(ap);
return count;
}
int pcm_scanb(pcm_io_context_t *io, const char *fmt, ...)
{
int c, count = 0;
va_list ap;
va_start(ap, fmt);
while ((c = *fmt++)) {
switch (c) {
case 'S':
TRY_IO(pcm_read16be(io, va_arg(ap, uint16_t*)));
++count;
break;
case 'L':
TRY_IO(pcm_read32be(io, va_arg(ap, uint32_t*)));
++count;
break;
case 'Q':
TRY_IO(pcm_read64be(io, va_arg(ap, uint64_t*)));
++count;
break;
}
}
FAIL:
va_end(ap);
return count;
}
static
int channel_compare(const void *a, const void *b)
{
return (*(const uint8_t **)a)[0] - (*(const uint8_t **)b)[0];
}
void apple_translate_channel_labels(uint8_t *channels, unsigned n)
{
unsigned i;
char *has_side = strpbrk((char*)channels, "\x0A\x0B");
for (i = 0; i < n; ++i) {
switch (channels[i]) {
case kAudioChannelLabel_LeftSurround:
case kAudioChannelLabel_RightSurround:
if (!has_side) channels[i] += 5; // map to SL/SR
break;
case kAudioChannelLabel_RearSurroundLeft:
case kAudioChannelLabel_RearSurroundRight:
if (!has_side) channels[i] -= 28; // map to BL/BR
break;
case kAudioChannelLabel_Mono:
channels[i] = kAudioChannelLabel_Center;
break;
}
}
}
int apple_chan_chunk(pcm_io_context_t *io, uint32_t chunk_size,
pcm_sample_description_t *fmt, uint8_t *mapping)
{
/*
* Although FDK encoder supports upto 5.1ch, we handle upto
* 8 channels here.
*/
uint32_t i, mChannelLayoutTag, mChannelBitmap, mNumberChannelDescriptions;
uint32_t mask = 0;
const uint32_t nchannels = fmt->channels_per_frame;
uint8_t channels[9] = { 0 };
uint8_t *index[8] = { 0 };
const char *layout = 0;
ENSURE(chunk_size >= 12);
TRY_IO(pcm_scanb(io, "LLL", &mChannelLayoutTag, &mChannelBitmap,
&mNumberChannelDescriptions) != 3);
switch (mChannelLayoutTag) {
case kAudioChannelLayoutTag_UseChannelBitmap:
ENSURE(bitcount(mask) == nchannels);
TRY_IO(pcm_skip(io, chunk_size - 12));
fmt->channel_mask = mChannelBitmap;
for (i = 0; i < nchannels; ++i)
mapping[i] = i;
return 0;
case kAudioChannelLayoutTag_UseChannelDescriptions:
ENSURE(mNumberChannelDescriptions == nchannels);
ENSURE(chunk_size >= 12 + nchannels * 20);
for (i = 0; i < mNumberChannelDescriptions; ++i) {
uint32_t mChannelLabel;
TRY_IO(pcm_read32be(io, &mChannelLabel));
ENSURE(mChannelLabel && mChannelLabel <= 0xff);
channels[i] = mChannelLabel;
TRY_IO(pcm_skip(io, 16));
}
TRY_IO(pcm_skip(io, chunk_size - 12 - nchannels * 20));
apple_translate_channel_labels(channels, nchannels);
for (i = 0; i < nchannels; ++i)
if (channels[i] > kAudioChannelLabel_TopBackLeft)
goto FAIL;
break;
default:
ENSURE((mChannelLayoutTag & 0xffff) == nchannels);
TRY_IO(pcm_skip(io, chunk_size - 12));
switch (mChannelLayoutTag) {
/* 1ch */
case kAudioChannelLayoutTag_Mono:
layout = "\x03"; break;
/* 1.1ch */
case kAudioChannelLayoutTag_AC3_1_0_1:
layout = "\x03\x04"; break;
/* 2ch */
case kAudioChannelLayoutTag_Stereo:
case kAudioChannelLayoutTag_MatrixStereo:
case kAudioChannelLayoutTag_Binaural:
layout = "\x01\x02"; break;
/* 2.1ch */
case kAudioChannelLayoutTag_DVD_4:
layout = "\x01\x02\x04"; break;
/* 3ch */
case kAudioChannelLayoutTag_MPEG_3_0_A:
layout = "\x01\x02\x03"; break;
case kAudioChannelLayoutTag_AC3_3_0:
layout = "\x01\x03\x02"; break;
case kAudioChannelLayoutTag_MPEG_3_0_B:
layout = "\x03\x01\x02"; break;
case kAudioChannelLayoutTag_ITU_2_1:
layout = "\x01\x02\x09"; break;
/* 3.1ch */
case kAudioChannelLayoutTag_DVD_10:
layout = "\x01\x02\x03\x04"; break;
case kAudioChannelLayoutTag_AC3_3_0_1:
layout = "\x01\x03\x02\x04"; break;
case kAudioChannelLayoutTag_DVD_5:
layout = "\x01\x02\x04\x09"; break;
case kAudioChannelLayoutTag_AC3_2_1_1:
layout = "\x01\x02\x09\x04"; break;
/* 4ch */
case kAudioChannelLayoutTag_Quadraphonic:
case kAudioChannelLayoutTag_ITU_2_2:
layout = "\x01\x02\x0A\x0B"; break;
case kAudioChannelLayoutTag_MPEG_4_0_A:
layout = "\x01\x02\x03\x09"; break;
case kAudioChannelLayoutTag_MPEG_4_0_B:
layout = "\x03\x01\x02\x09"; break;
case kAudioChannelLayoutTag_AC3_3_1:
layout = "\x01\x03\x02\x09"; break;
/* 4.1ch */
case kAudioChannelLayoutTag_DVD_6:
layout = "\x01\x02\x04\x0A\x0B"; break;
case kAudioChannelLayoutTag_DVD_18:
layout = "\x01\x02\x0A\x0B\x04"; break;
case kAudioChannelLayoutTag_DVD_11:
layout = "\x01\x02\x03\x04\x09"; break;
case kAudioChannelLayoutTag_AC3_3_1_1:
layout = "\x01\x03\x02\x09\x04"; break;
/* 5ch */
case kAudioChannelLayoutTag_MPEG_5_0_A:
layout = "\x01\x02\x03\x0A\x0B"; break;
case kAudioChannelLayoutTag_Pentagonal:
case kAudioChannelLayoutTag_MPEG_5_0_B:
layout = "\x01\x02\x0A\x0B\x03"; break;
case kAudioChannelLayoutTag_MPEG_5_0_C:
layout = "\x01\x03\x02\x0A\x0B"; break;
case kAudioChannelLayoutTag_MPEG_5_0_D:
layout = "\x03\x01\x02\x0A\x0B"; break;
/* 5.1ch */
case kAudioChannelLayoutTag_MPEG_5_1_A:
layout = "\x01\x02\x03\x04\x0A\x0B"; break;
case kAudioChannelLayoutTag_MPEG_5_1_B:
layout = "\x01\x02\x0A\x0B\x03\x04"; break;
case kAudioChannelLayoutTag_MPEG_5_1_C:
layout = "\x01\x03\x02\x0A\x0B\x04"; break;
case kAudioChannelLayoutTag_MPEG_5_1_D:
layout = "\x03\x01\x02\x0A\x0B\x04"; break;
/* 6ch */
case kAudioChannelLayoutTag_Hexagonal:
case kAudioChannelLayoutTag_AudioUnit_6_0:
layout = "\x01\x02\x0A\x0B\x03\x09"; break;
case kAudioChannelLayoutTag_AAC_6_0:
layout = "\x03\x01\x02\x0A\x0B\x09"; break;
/* 6.1ch */
case kAudioChannelLayoutTag_MPEG_6_1_A:
layout = "\x01\x02\x03\x04\x0A\x0B\x09"; break;
case kAudioChannelLayoutTag_AAC_6_1:
layout = "\x03\x01\x02\x0A\x0B\x09\x04"; break;
/* 7ch */
case kAudioChannelLayoutTag_AudioUnit_7_0:
layout = "\x01\x02\x0A\x0B\x03\x05\x06"; break;
case kAudioChannelLayoutTag_AudioUnit_7_0_Front:
layout = "\x01\x02\x0A\x0B\x03\x07\x08"; break;
case kAudioChannelLayoutTag_AAC_7_0:
layout = "\x03\x01\x02\x0A\x0B\x05\x06"; break;
/* 7.1ch */
case kAudioChannelLayoutTag_MPEG_7_1_A:
layout = "\x01\x02\x03\x04\x0A\x0B\x07\x08"; break;
case kAudioChannelLayoutTag_MPEG_7_1_B:
layout = "\x03\x07\x08\x01\x02\x05\x06\x04"; break;
case kAudioChannelLayoutTag_MPEG_7_1_C:
layout = "\x01\x02\x03\x04\x0A\x0B\x05\x06"; break;
case kAudioChannelLayoutTag_Emagic_Default_7_1:
layout = "\x01\x02\x0A\x0B\x03\x04\x07\x08"; break;
/* 8ch */
case kAudioChannelLayoutTag_Octagonal:
layout = "\x01\x02\x05\x06\x03\x09\x0A\x0B"; break;
case kAudioChannelLayoutTag_AAC_Octagonal:
layout = "\x03\x01\x02\x0A\x0B\x05\x06\x09"; break;
default:
goto FAIL;
}
strcpy((char*)channels, layout);
}
for (i = 0; i < nchannels; ++i)
mask |= 1 << (channels[i] - 1);
fmt->channel_mask = mask;
ENSURE(bitcount(mask) == nchannels);
for (i = 0; i < nchannels; ++i)
index[i] = channels + i;
qsort(index, nchannels, sizeof(char*), channel_compare);
for (i = 0; i < nchannels; ++i)
mapping[i] = index[i] - channels;
return 0;
FAIL:
return -1;
}

View File

@ -15,36 +15,18 @@
#include <string.h>
#include <stdarg.h>
#include "wav_reader.h"
#include "m4af_endian.h"
#define RIFF_FOURCC(a,b,c,d) ((a)|((b)<<8)|((c)<<16)|((d)<<24))
#define TRY_IO(expr) \
do { \
if (expr) \
goto FAIL; \
} while (0)
#define ASSERT_FORMAT(ctx, expr) \
do { \
if (!expr) { \
if (!ctx->last_error) \
ctx->last_error = WAV_INVALID_FORMAT; \
goto FAIL;\
} \
} while (0)
struct wav_reader_t {
typedef struct wav_reader_t {
pcm_reader_vtbl_t *vtbl;
pcm_sample_description_t sample_format;
int64_t length;
int64_t position;
int32_t data_offset;
int ignore_length;
int last_error;
wav_io_context_t io;
void *io_cookie;
};
pcm_io_context_t io;
} wav_reader_t;
static const uint8_t WAV_GUID_PCM[] = {
1, 0, 0, 0, 0, 0, 0x10, 0, 0x80, 0, 0, 0xaa, 0, 0x38, 0x9b, 0x71
@ -74,144 +56,11 @@ static void wav_teardown(pcm_reader_t **reader)
*reader = 0;
}
static
int riff_read(wav_reader_t *reader, void *buffer, uint32_t size)
{
int rc;
uint32_t count = 0;
if (reader->last_error)
return -1;
do {
rc = reader->io.read(reader->io_cookie, buffer, size - count);
if (rc > 0)
count += rc;
else if (rc < 0)
reader->last_error = WAV_IO_ERROR;
} while (rc > 0 && count < size);
return count > 0 ? count : rc;
}
static
int riff_skip(wav_reader_t *reader, int64_t count)
{
char buff[8192];
int rc;
if (reader->last_error)
return -1;
if (count == 0)
return 0;
if (reader->io.seek &&
reader->io.seek(reader->io_cookie, count, SEEK_CUR) >= 0)
return 0;
do {
if ((rc = riff_read(reader, buff, count > 8192 ? 8192 : count)) > 0)
count -= rc;
} while (rc > 0 && count > 0);
if (count > 0)
reader->last_error = WAV_IO_ERROR;
return reader->last_error ? -1 : 0;
}
static
int riff_seek(wav_reader_t *reader, int64_t off, int whence)
{
int rc;
if (reader->last_error)
return -1;
if (!reader->io.seek)
goto FAIL;
if ((rc = reader->io.seek(reader->io_cookie, off, whence)) < 0)
goto FAIL;
return 0;
FAIL:
reader->last_error = WAV_IO_ERROR;
return -1;
}
static
int64_t riff_tell(wav_reader_t *reader)
{
int64_t off;
if (reader->last_error || !reader->io.tell)
return -1;
off = reader->io.tell(reader->io_cookie);
if (off < 0) {
reader->last_error = WAV_IO_ERROR;
return -1;
}
return off;
}
static
int riff_read16(wav_reader_t *reader, uint16_t *value)
{
TRY_IO(riff_read(reader, value, 2) != 2);
*value = m4af_ltoh16(*value);
return 0;
FAIL:
return -1;
}
static
int riff_read32(wav_reader_t *reader, uint32_t *value)
{
TRY_IO(riff_read(reader, value, 4) != 4);
*value = m4af_ltoh32(*value);
return 0;
FAIL:
return -1;
}
static
int riff_read64(wav_reader_t *reader, uint64_t *value)
{
TRY_IO(riff_read(reader, value, 8) != 8);
*value = m4af_ltoh64(*value);
return 0;
FAIL:
return -1;
}
static
int riff_scan(wav_reader_t *reader, const char *fmt, ...)
{
int c, count = 0;
va_list ap;
va_start(ap, fmt);
while ((c = *fmt++)) {
switch (c) {
case 'S':
TRY_IO(riff_read16(reader, va_arg(ap, uint16_t*)));
++count;
break;
case 'L':
TRY_IO(riff_read32(reader, va_arg(ap, uint32_t*)));
++count;
break;
case 'Q':
TRY_IO(riff_read64(reader, va_arg(ap, uint64_t*)));
++count;
break;
}
}
FAIL:
va_end(ap);
return count;
}
static
uint32_t riff_next_chunk(wav_reader_t *reader, uint32_t *chunk_size)
{
uint32_t fcc;
if (riff_scan(reader, "LL", &fcc, chunk_size) == 2)
return fcc;
return 0;
return (pcm_scanl(&reader->io, "LL", &fcc, chunk_size) == 2) ? fcc : 0;
}
static
@ -225,7 +74,7 @@ int wav_read_frames(pcm_reader_t *preader, void *buffer, unsigned nframes)
nframes = reader->length - reader->position;
nbytes = nframes * reader->sample_format.bytes_per_frame;
if (nbytes) {
if ((rc = riff_read(reader, buffer, nbytes)) < 0)
if ((rc = pcm_read(&reader->io, buffer, nbytes)) < 0)
return -1;
nframes = rc / reader->sample_format.bytes_per_frame;
reader->position += nframes;
@ -240,11 +89,10 @@ int riff_ds64(wav_reader_t *reader, int64_t *length)
uint64_t riff_size, sample_count;
fcc = riff_next_chunk(reader, &chunk_size);
ASSERT_FORMAT(reader,
fcc == RIFF_FOURCC('d','s','6','4') && chunk_size >= 28);
TRY_IO(riff_scan(reader, "QQQL",
ENSURE(fcc == RIFF_FOURCC('d','s','6','4') && chunk_size >= 28);
TRY_IO(pcm_scanl(&reader->io, "QQQL",
&riff_size, length, &sample_count, &table_size) != 4);
TRY_IO(riff_skip(reader, (chunk_size - 27) & ~1));
TRY_IO(pcm_skip(&reader->io, (chunk_size - 27) & ~1));
reader->data_offset += (chunk_size + 9) & ~1;
FAIL:
return -1;
@ -259,41 +107,34 @@ int wav_fmt(wav_reader_t *reader, uint32_t size)
uint8_t guid[16];
int is_float = 0;
ASSERT_FORMAT(reader, size >= 16);
TRY_IO(riff_scan(reader, "SSLLSS", &wFormatTag, &nChannels,
ENSURE(size >= 16);
TRY_IO(pcm_scanl(&reader->io, "SSLLSS", &wFormatTag, &nChannels,
&nSamplesPerSec, &nAvgBytesPerSec, &nBlockAlign,
&wBitsPerSample) != 6);
wValidBitsPerSample = wBitsPerSample;
if (wFormatTag != 1 && wFormatTag != 3 && wFormatTag != 0xfffe) {
reader->last_error = WAV_UNSUPPORTED_FORMAT;
goto FAIL;
}
ASSERT_FORMAT(reader,
nChannels && nSamplesPerSec && nAvgBytesPerSec &&
nBlockAlign && wBitsPerSample && !(wBitsPerSample & 7) &&
nBlockAlign == nChannels * wBitsPerSample / 8);
ENSURE(wFormatTag == 1 || wFormatTag == 3 || wFormatTag == 0xfffe);
ENSURE(nChannels && nSamplesPerSec && nAvgBytesPerSec &&
nBlockAlign && wBitsPerSample && !(wBitsPerSample & 7) &&
nBlockAlign == nChannels * wBitsPerSample / 8);
if (wFormatTag == 3)
is_float = 1;
if (wFormatTag != 0xfffe)
TRY_IO(riff_skip(reader, (size - 15) & ~1));
TRY_IO(pcm_skip(&reader->io, (size - 15) & ~1));
else {
ASSERT_FORMAT(reader, size >= 40);
TRY_IO(riff_scan(reader, "SSL",
ENSURE(size >= 40);
TRY_IO(pcm_scanl(&reader->io, "SSL",
&cbSize, &wValidBitsPerSample, &dwChannelMask) != 3);
TRY_IO(riff_read(reader, guid, 16) != 16);
TRY_IO(pcm_read(&reader->io, guid, 16) != 16);
if (memcmp(guid, WAV_GUID_FLOAT, 16) == 0)
is_float = 1;
else if (memcmp(guid, WAV_GUID_PCM, 16) != 0) {
reader->last_error = WAV_UNSUPPORTED_FORMAT;
else if (memcmp(guid, WAV_GUID_PCM, 16) != 0)
goto FAIL;
}
ASSERT_FORMAT(reader,
wValidBitsPerSample &&
wValidBitsPerSample <= wBitsPerSample);
TRY_IO(riff_skip(reader, (size - 39) & ~1));
ENSURE(wValidBitsPerSample && wValidBitsPerSample <= wBitsPerSample);
TRY_IO(pcm_skip(&reader->io, (size - 39) & ~1));
}
reader->sample_format.sample_rate = nSamplesPerSec;
reader->sample_format.bits_per_channel = wValidBitsPerSample;
@ -318,12 +159,10 @@ int wav_parse(wav_reader_t *reader, int64_t *data_length)
*data_length = 0;
container = riff_next_chunk(reader, &chunk_size);
if (container != RIFF_FOURCC('R','I','F','F') &&
container != RIFF_FOURCC('R','F','6','4'))
goto FAIL;
TRY_IO(riff_read32(reader, &fcc));
if (fcc != RIFF_FOURCC('W','A','V','E'))
goto FAIL;
ENSURE(container == RIFF_FOURCC('R','I','F','F') ||
container == RIFF_FOURCC('R','F','6','4'));
TRY_IO(pcm_read32le(&reader->io, &fcc));
ENSURE(fcc == RIFF_FOURCC('W','A','V','E'));
reader->data_offset = 12;
if (container == RIFF_FOURCC('R','F','6','4'))
@ -338,7 +177,7 @@ int wav_parse(wav_reader_t *reader, int64_t *data_length)
reader->data_offset += 8;
break;
} else {
TRY_IO(riff_skip(reader, (chunk_size + 1) & ~1));
TRY_IO(pcm_skip(&reader->io, (chunk_size + 1) & ~1));
}
reader->data_offset += (chunk_size + 9) & ~1;
}
@ -356,8 +195,7 @@ static pcm_reader_vtbl_t wav_vtable = {
wav_teardown
};
pcm_reader_t *wav_open(wav_io_context_t *io_ctx, void *io_cookie,
int ignore_length)
pcm_reader_t *wav_open(pcm_io_context_t *io, int ignore_length)
{
wav_reader_t *reader = 0;
int64_t data_length;
@ -365,8 +203,7 @@ pcm_reader_t *wav_open(wav_io_context_t *io_ctx, void *io_cookie,
if ((reader = calloc(1, sizeof(wav_reader_t))) == 0)
return 0;
memcpy(&reader->io, io_ctx, sizeof(wav_io_context_t));
reader->io_cookie = io_cookie;
memcpy(&reader->io, io, sizeof(pcm_io_context_t));
reader->ignore_length = ignore_length;
if (wav_parse(reader, &data_length) < 0) {
free(reader);
@ -378,39 +215,34 @@ pcm_reader_t *wav_open(wav_io_context_t *io_ctx, void *io_cookie,
else
reader->length = data_length / bpf;
if (reader->length == INT64_MAX && reader->io.seek && reader->io.tell) {
if (reader->io.seek(reader->io_cookie, 0, SEEK_END) >= 0) {
int64_t size = reader->io.tell(reader->io_cookie);
if (reader->length == INT64_MAX) {
if (pcm_seek(&reader->io, 0, SEEK_END) >= 0) {
int64_t size = pcm_tell(&reader->io);
if (size > 0)
reader->length = (size - reader->data_offset) / bpf;
reader->io.seek(reader->io_cookie, reader->data_offset, SEEK_SET);
pcm_seek(&reader->io, reader->data_offset, SEEK_SET);
}
}
reader->vtbl = &wav_vtable;
return (pcm_reader_t *)reader;
}
pcm_reader_t *raw_open(wav_io_context_t *io_ctx, void *io_cookie,
pcm_reader_t *raw_open(pcm_io_context_t *io,
const pcm_sample_description_t *desc)
{
wav_reader_t *reader = 0;
if ((reader = calloc(1, sizeof(wav_reader_t))) == 0)
return 0;
memcpy(&reader->io, io_ctx, sizeof(wav_io_context_t));
memcpy(&reader->io, io, sizeof(pcm_io_context_t));
memcpy(&reader->sample_format, desc, sizeof(pcm_sample_description_t));
reader->io_cookie = io_cookie;
if (io_ctx->seek && io_ctx->tell) {
if (reader->io.seek(reader->io_cookie, 0, SEEK_END) >= 0) {
int64_t size = reader->io.tell(reader->io_cookie);
if (size > 0)
reader->length = size / desc->bytes_per_frame;
reader->io.seek(reader->io_cookie, reader->data_offset, SEEK_SET);
}
if (pcm_seek(&reader->io, 0, SEEK_END) >= 0) {
int64_t size = pcm_tell(&reader->io);
if (size > 0)
reader->length = size / reader->sample_format.bytes_per_frame;
pcm_seek(&reader->io, 0, SEEK_SET);
} else
reader->length = INT64_MAX;
reader->vtbl = &wav_vtable;
return (pcm_reader_t *)reader;
}

View File

@ -8,28 +8,8 @@
#include "lpcm.h"
#include "pcm_reader.h"
enum wav_error_code {
WAV_IO_ERROR = 1,
WAV_NO_MEMORY,
WAV_INVALID_FORMAT,
WAV_UNSUPPORTED_FORMAT
};
typedef int (*wav_read_callback)(void *cookie, void *data, uint32_t size);
typedef int (*wav_seek_callback)(void *cookie, int64_t off, int whence);
typedef int64_t (*wav_tell_callback)(void *cookie);
typedef struct wav_io_context_t {
wav_read_callback read;
wav_seek_callback seek;
wav_tell_callback tell;
} wav_io_context_t;
typedef struct wav_reader_t wav_reader_t;
pcm_reader_t *wav_open(wav_io_context_t *io_ctx, void *io_cookie,
int ignore_length);
pcm_reader_t *raw_open(wav_io_context_t *io_ctx, void *io_cookie,
pcm_reader_t *wav_open(pcm_io_context_t *io, int ignore_length);
pcm_reader_t *raw_open(pcm_io_context_t *io,
const pcm_sample_description_t *desc);
#endif

View File

@ -1,4 +1,4 @@
#ifndef VERSION_H
#define VERSION_H
const char *fdkaac_version = "0.3.3";
const char *fdkaac_version = "0.4.0";
#endif