Mercurial > audlegacy-plugins
view src/aac/mp4ff/mp4ff.c @ 12:3da1b8942b8b trunk
[svn] - remove src/Input src/Output src/Effect src/General src/Visualization src/Container
author | nenolod |
---|---|
date | Mon, 18 Sep 2006 03:14:20 -0700 |
parents | src/Input/aac/mp4ff/mp4ff.c@13389e613d67 |
children | 357de698dab2 |
line wrap: on
line source
/* ** FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding ** Copyright (C) 2003-2004 M. Bakker, Ahead Software AG, http://www.nero.com ** ** This program is free software; you can redistribute it and/or modify ** it under the terms of the GNU General Public License as published by ** the Free Software Foundation; either version 2 of the License, or ** (at your option) any later version. ** ** This program is distributed in the hope that it will be useful, ** but WITHOUT ANY WARRANTY; without even the implied warranty of ** MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ** GNU General Public License for more details. ** ** You should have received a copy of the GNU General Public License ** along with this program; if not, write to the Free Software ** Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. ** ** Any non-GPL usage of this software or parts of this software is strictly ** forbidden. ** ** Commercial non-GPL licensing of this software is possible. ** For more info contact Ahead Software through Mpeg4AAClicense@nero.com. ** ** $Id: mp4ff.c,v 1.15 2004/01/11 15:52:18 menno Exp $ **/ #include <stdlib.h> #include <string.h> #include "mp4ffint.h" #include "drms.h" mp4ff_t *mp4ff_open_read(mp4ff_callback_t *f) { mp4ff_t *ff = malloc(sizeof(mp4ff_t)); memset(ff, 0, sizeof(mp4ff_t)); ff->stream = f; parse_atoms(ff); return ff; } void mp4ff_close(mp4ff_t *ff) { int32_t i; for (i = 0; i < ff->total_tracks; i++) { if (ff->track[i]) { if (ff->track[i]->stsz_table) free(ff->track[i]->stsz_table); if (ff->track[i]->stts_sample_count) free(ff->track[i]->stts_sample_count); if (ff->track[i]->stts_sample_delta) free(ff->track[i]->stts_sample_delta); if (ff->track[i]->stsc_first_chunk) free(ff->track[i]->stsc_first_chunk); if (ff->track[i]->stsc_samples_per_chunk) free(ff->track[i]->stsc_samples_per_chunk); if (ff->track[i]->stsc_sample_desc_index) free(ff->track[i]->stsc_sample_desc_index); if (ff->track[i]->stco_chunk_offset) free(ff->track[i]->stco_chunk_offset); if (ff->track[i]->decoderConfig) free(ff->track[i]->decoderConfig); if (ff->track[i]->ctts_sample_count) free(ff->track[i]->ctts_sample_count); if (ff->track[i]->ctts_sample_offset) free(ff->track[i]->ctts_sample_offset); #ifdef ITUNES_DRM if (ff->track[i]->p_drms) drms_free(ff->track[i]->p_drms); #endif free(ff->track[i]); } } #ifdef USE_TAGGING mp4ff_tag_delete(&(ff->tags)); #endif if (ff) free(ff); } void mp4ff_track_add(mp4ff_t *f) { f->total_tracks++; f->track[f->total_tracks - 1] = malloc(sizeof(mp4ff_track_t)); memset(f->track[f->total_tracks - 1], 0, sizeof(mp4ff_track_t)); } /* parse atoms that are sub atoms of other atoms */ int32_t parse_sub_atoms(mp4ff_t *f, const uint64_t total_size) { uint64_t size; uint8_t atom_type = 0; uint64_t counted_size = 0; uint8_t header_size = 0; while (counted_size < total_size) { size = mp4ff_atom_read_header(f, &atom_type, &header_size); counted_size += size; /* check for end of file */ if (size == 0) break; /* we're starting to read a new track, update index, * so that all data and tables get written in the right place */ if (atom_type == ATOM_TRAK) { mp4ff_track_add(f); } /* parse subatoms */ if (atom_type < SUBATOMIC) { parse_sub_atoms(f, size-header_size); } else { mp4ff_atom_read(f, (uint32_t)size, atom_type); } } return 0; } /* parse root atoms */ int32_t parse_atoms(mp4ff_t *f) { uint64_t size; uint8_t atom_type = 0; uint8_t header_size = 0; f->file_size = 0; while ((size = mp4ff_atom_read_header(f, &atom_type, &header_size)) != 0) { f->file_size += size; f->last_atom = atom_type; if (atom_type == ATOM_MDAT && f->moov_read) { /* moov atom is before mdat, we can stop reading when mdat is encountered */ /* file position will stay at beginning of mdat data */ // break; } if (atom_type == ATOM_MOOV && size > header_size) { f->moov_read = 1; f->moov_offset = mp4ff_position(f)-header_size; f->moov_size = size; } /* parse subatoms */ if (atom_type < SUBATOMIC) { parse_sub_atoms(f, size-header_size); } else { /* skip this atom */ mp4ff_set_position(f, mp4ff_position(f)+size-header_size); } } return 0; } int32_t mp4ff_get_decoder_config(const mp4ff_t *f, const int32_t track, uint8_t** ppBuf, uint32_t* pBufSize) { if (track < 0 || track >= f->total_tracks) { *ppBuf = NULL; *pBufSize = 0; return 1; } if (f->track[track]->decoderConfig == NULL || f->track[track]->decoderConfigLen == 0) { *ppBuf = NULL; *pBufSize = 0; } else { *ppBuf = malloc(f->track[track]->decoderConfigLen); if (*ppBuf == NULL) { *pBufSize = 0; return 1; } memcpy(*ppBuf, f->track[track]->decoderConfig, f->track[track]->decoderConfigLen); *pBufSize = f->track[track]->decoderConfigLen; } return 0; } int32_t mp4ff_get_track_type(const mp4ff_t *f, const int track) { if (track < 0) return -1; return f->track[track]->type; } int32_t mp4ff_total_tracks(const mp4ff_t *f) { return f->total_tracks; } int32_t mp4ff_time_scale(const mp4ff_t *f, const int32_t track) { if (track < 0) return -1; return f->track[track]->timeScale; } uint32_t mp4ff_get_avg_bitrate(const mp4ff_t *f, const int32_t track) { if (track < 0) return -1; return f->track[track]->avgBitrate; } uint32_t mp4ff_get_max_bitrate(const mp4ff_t *f, const int32_t track) { if (track < 0) return -1; return f->track[track]->maxBitrate; } int64_t mp4ff_get_track_duration(const mp4ff_t *f, const int32_t track) { if (track < 0) return -1; return f->track[track]->duration; } int64_t mp4ff_get_track_duration_use_offsets(const mp4ff_t *f, const int32_t track) { int64_t duration = mp4ff_get_track_duration(f,track); if (duration!=-1) { int64_t offset = mp4ff_get_sample_offset(f,track,0); if (offset > duration) duration = 0; else duration -= offset; } return duration; } int32_t mp4ff_num_samples(const mp4ff_t *f, const int32_t track) { int32_t i; int32_t total = 0; if (track < 0) return -1; for (i = 0; i < f->track[track]->stts_entry_count; i++) { total += f->track[track]->stts_sample_count[i]; } return total; } uint32_t mp4ff_get_sample_rate(const mp4ff_t *f, const int32_t track) { return f->track[track]->sampleRate; } uint32_t mp4ff_get_channel_count(const mp4ff_t * f,const int32_t track) { return f->track[track]->channelCount; } uint32_t mp4ff_get_audio_type(const mp4ff_t * f,const int32_t track) { return f->track[track]->audioType; } int32_t mp4ff_get_sample_duration_use_offsets(const mp4ff_t *f, const int32_t track, const int32_t sample) { int32_t d,o; d = mp4ff_get_sample_duration(f,track,sample); if (d!=-1) { o = mp4ff_get_sample_offset(f,track,sample); if (o>d) d = 0; else d -= o; } return d; } int32_t mp4ff_get_sample_duration(const mp4ff_t *f, const int32_t track, const int32_t sample) { int32_t i, co = 0; for (i = 0; i < f->track[track]->stts_entry_count; i++) { int32_t delta = f->track[track]->stts_sample_count[i]; if (sample < co + delta) return f->track[track]->stts_sample_delta[i]; co += delta; } return (int32_t)(-1); } int64_t mp4ff_get_sample_position(const mp4ff_t *f, const int32_t track, const int32_t sample) { int32_t i, co = 0; int64_t acc = 0; for (i = 0; i < f->track[track]->stts_entry_count; i++) { int32_t delta = f->track[track]->stts_sample_count[i]; if (sample < co + delta) { acc += f->track[track]->stts_sample_delta[i] * (sample - co); return acc; } else { acc += f->track[track]->stts_sample_delta[i] * delta; } co += delta; } return (int64_t)(-1); } int32_t mp4ff_get_sample_offset(const mp4ff_t *f, const int32_t track, const int32_t sample) { int32_t i, co = 0; for (i = 0; i < f->track[track]->ctts_entry_count; i++) { int32_t delta = f->track[track]->ctts_sample_count[i]; if (sample < co + delta) return f->track[track]->ctts_sample_offset[i]; co += delta; } return 0; } int32_t mp4ff_find_sample(const mp4ff_t *f, const int32_t track, const int64_t offset,int32_t * toskip) { int32_t i, co = 0; int64_t offset_total = 0; mp4ff_track_t * p_track = f->track[track]; for (i = 0; i < p_track->stts_entry_count; i++) { int32_t sample_count = p_track->stts_sample_count[i]; int32_t sample_delta = p_track->stts_sample_delta[i]; int64_t offset_delta = (int64_t)sample_delta * (int64_t)sample_count; if (offset < offset_total + offset_delta) { int64_t offset_fromstts = offset - offset_total; if (toskip) *toskip = (int32_t)(offset_fromstts % sample_delta); return co + (int32_t)(offset_fromstts / sample_delta); } else { offset_total += offset_delta; } co += sample_count; } return (int32_t)(-1); } int32_t mp4ff_find_sample_use_offsets(const mp4ff_t *f, const int32_t track, const int64_t offset,int32_t * toskip) { return mp4ff_find_sample(f,track,offset + mp4ff_get_sample_offset(f,track,0),toskip); } int32_t mp4ff_read_sample(mp4ff_t *f, const int32_t track, const int32_t sample, uint8_t **audio_buffer, uint32_t *bytes) { int32_t result = 0; *bytes = mp4ff_audio_frame_size(f, track, sample); if (*bytes==0) return 0; *audio_buffer = (uint8_t*)malloc(*bytes); mp4ff_set_sample_position(f, track, sample); result = mp4ff_read_data(f, *audio_buffer, *bytes); if (!result) { free(*audio_buffer); *audio_buffer = 0; return 0; } #ifdef ITUNES_DRM if (f->track[track]->p_drms != NULL) { drms_decrypt(f->track[track]->p_drms, (uint32_t*)*audio_buffer, *bytes); } #endif return *bytes; } int32_t mp4ff_read_sample_v2(mp4ff_t *f, const int track, const int sample,unsigned char *buffer) { int32_t result = 0; int32_t size = mp4ff_audio_frame_size(f,track,sample); if (size<=0) return 0; mp4ff_set_sample_position(f, track, sample); result = mp4ff_read_data(f,buffer,size); #ifdef ITUNES_DRM if (f->track[track]->p_drms != NULL) { drms_decrypt(f->track[track]->p_drms, (uint32_t*)buffer, size); } #endif return result; } int32_t mp4ff_read_sample_getsize(mp4ff_t *f, const int track, const int sample) { int32_t temp = mp4ff_audio_frame_size(f, track, sample); if (temp<0) temp = 0; return temp; }