static uint64_t libavsmash_video_fetch_media_duration
(
libavsmash_video_decode_handler_t *vdhp
)
{
if( !vdhp )
return 0;
vdhp->media_duration = lsmash_get_media_duration_from_media_timeline( vdhp->root, vdhp->track_id );
return vdhp->media_duration;
}
static uint64_t libavsmash_audio_fetch_media_duration
(
libavsmash_audio_decode_handler_t *adhp
)
{
if( !adhp )
return 0;
adhp->media_duration = lsmash_get_media_duration_from_media_timeline( adhp->root, adhp->track_id );
return adhp->media_duration;
}
int libavsmash_video_setup_timestamp_info
(
libavsmash_video_decode_handler_t *vdhp,
libavsmash_video_output_handler_t *vohp,
int64_t *framerate_num,
int64_t *framerate_den
)
{
int err = -1;
uint64_t media_timescale = lsmash_get_media_timescale( vdhp->root, vdhp->track_id );
uint64_t media_duration = lsmash_get_media_duration_from_media_timeline( vdhp->root, vdhp->track_id );
if( media_duration == 0 )
media_duration = INT32_MAX;
if( vdhp->sample_count == 1 )
{
/* Calculate average framerate. */
reduce_fraction( &media_timescale, &media_duration );
*framerate_num = (int64_t)media_timescale;
*framerate_den = (int64_t)media_duration;
err = 0;
goto setup_finish;
}
lw_log_handler_t *lhp = &vdhp->config.lh;
lsmash_media_ts_list_t ts_list;
if( lsmash_get_media_timestamps( vdhp->root, vdhp->track_id, &ts_list ) < 0 )
{
lw_log_show( lhp, LW_LOG_ERROR, "Failed to get timestamps." );
goto setup_finish;
}
if( ts_list.sample_count != vdhp->sample_count )
{
lw_log_show( lhp, LW_LOG_ERROR, "Failed to count number of video samples." );
goto setup_finish;
}
uint32_t composition_sample_delay;
if( lsmash_get_max_sample_delay( &ts_list, &composition_sample_delay ) < 0 )
{
lsmash_delete_media_timestamps( &ts_list );
lw_log_show( lhp, LW_LOG_ERROR, "Failed to get composition delay." );
goto setup_finish;
}
if( composition_sample_delay )
{
/* Consider composition order for keyframe detection.
* Note: sample number for L-SMASH is 1-origin. */
vdhp->order_converter = (order_converter_t *)lw_malloc_zero( (ts_list.sample_count + 1) * sizeof(order_converter_t) );
if( !vdhp->order_converter )
{
lsmash_delete_media_timestamps( &ts_list );
lw_log_show( lhp, LW_LOG_ERROR, "Failed to allocate memory." );
goto setup_finish;
}
for( uint32_t i = 0; i < ts_list.sample_count; i++ )
ts_list.timestamp[i].dts = i + 1;
lsmash_sort_timestamps_composition_order( &ts_list );
for( uint32_t i = 0; i < ts_list.sample_count; i++ )
vdhp->order_converter[i + 1].composition_to_decoding = (uint32_t)ts_list.timestamp[i].dts;
}
/* Calculate average framerate. */
uint64_t largest_cts = ts_list.timestamp[0].cts;
uint64_t second_largest_cts = 0;
uint64_t first_duration = ts_list.timestamp[1].cts - ts_list.timestamp[0].cts;
uint64_t composition_timebase = first_duration;
int strict_cfr = 1;
for( uint32_t i = 1; i < ts_list.sample_count; i++ )
{
uint64_t duration = ts_list.timestamp[i].cts - ts_list.timestamp[i - 1].cts;
if( duration == 0 )
{
lsmash_delete_media_timestamps( &ts_list );
lw_log_show( lhp, LW_LOG_WARNING, "Detected CTS duplication at frame %" PRIu32, i );
err = 0;
goto setup_finish;
}
if( strict_cfr && duration != first_duration )
strict_cfr = 0;
composition_timebase = get_gcd( composition_timebase, duration );
second_largest_cts = largest_cts;
largest_cts = ts_list.timestamp[i].cts;
}
uint64_t reduce = reduce_fraction( &media_timescale, &composition_timebase );
uint64_t composition_duration = ((largest_cts - ts_list.timestamp[0].cts) + (largest_cts - second_largest_cts)) / reduce;
lsmash_delete_media_timestamps( &ts_list );
double avg_frame_rate = (vdhp->sample_count * ((double)media_timescale / composition_duration));
if( strict_cfr || !lw_try_rational_framerate( avg_frame_rate, framerate_num, framerate_den, composition_timebase ) )
{
uint64_t num = (uint64_t)(avg_frame_rate * composition_timebase + 0.5);
uint64_t den = composition_timebase;
if( num && den )
reduce_fraction( &num, &den );
else
{
num = 1;
den = 1;
}
*framerate_num = (int64_t)num;
*framerate_den = (int64_t)den;
}
err = 0;
setup_finish:;
if( vohp->vfr2cfr )
{
/* Override average framerate by specified output constant framerate. */
*framerate_num = (int64_t)vohp->cfr_num;
*framerate_den = (int64_t)vohp->cfr_den;
vohp->frame_count = ((double)vohp->cfr_num / vohp->cfr_den)
* ((double)media_duration / media_timescale)
+ 0.5;
}
else
vohp->frame_count = libavsmash_video_get_sample_count( vdhp );
uint32_t min_cts_sample_number = get_decoding_sample_number( vdhp->order_converter, 1 );
vdhp->config.error = lsmash_get_cts_from_media_timeline( vdhp->root, vdhp->track_id, min_cts_sample_number, &vdhp->min_cts );
return err;
}
void LSMASHAudioSource::get_audio_track( const char *source, uint32_t track_number, bool skip_priming, IScriptEnvironment *env )
{
uint32_t number_of_tracks = open_file( source, env );
if( track_number && track_number > number_of_tracks )
env->ThrowError( "LSMASHAudioSource: the number of tracks equals %I32u.", number_of_tracks );
/* L-SMASH */
uint32_t i;
lsmash_media_parameters_t media_param;
if( track_number == 0 )
{
/* Get the first audio track. */
for( i = 1; i <= number_of_tracks; i++ )
{
adh.track_ID = lsmash_get_track_ID( adh.root, i );
if( adh.track_ID == 0 )
env->ThrowError( "LSMASHAudioSource: failed to find audio track." );
lsmash_initialize_media_parameters( &media_param );
if( lsmash_get_media_parameters( adh.root, adh.track_ID, &media_param ) )
env->ThrowError( "LSMASHAudioSource: failed to get media parameters." );
if( media_param.handler_type == ISOM_MEDIA_HANDLER_TYPE_AUDIO_TRACK )
break;
}
if( i > number_of_tracks )
env->ThrowError( "LSMASHAudioSource: failed to find audio track." );
}
else
{
/* Get the desired audio track. */
adh.track_ID = lsmash_get_track_ID( adh.root, track_number );
if( adh.track_ID == 0 )
env->ThrowError( "LSMASHAudioSource: failed to find audio track." );
lsmash_initialize_media_parameters( &media_param );
if( lsmash_get_media_parameters( adh.root, adh.track_ID, &media_param ) )
env->ThrowError( "LSMASHAudioSource: failed to get media parameters." );
if( media_param.handler_type != ISOM_MEDIA_HANDLER_TYPE_AUDIO_TRACK )
env->ThrowError( "LSMASHAudioSource: the track you specified is not an audio track." );
}
if( lsmash_construct_timeline( adh.root, adh.track_ID ) )
env->ThrowError( "LSMASHAudioSource: failed to get construct timeline." );
if( get_summaries( adh.root, adh.track_ID, &adh.config ) )
env->ThrowError( "LSMASHAudioSource: failed to get summaries." );
adh.frame_count = lsmash_get_sample_count_in_media_timeline( adh.root, adh.track_ID );
vi.num_audio_samples = lsmash_get_media_duration_from_media_timeline( adh.root, adh.track_ID );
if( skip_priming )
{
uint32_t itunes_metadata_count = lsmash_count_itunes_metadata( adh.root );
for( i = 1; i <= itunes_metadata_count; i++ )
{
lsmash_itunes_metadata_t metadata;
if( lsmash_get_itunes_metadata( adh.root, i, &metadata ) < 0 )
continue;
if( metadata.item != ITUNES_METADATA_ITEM_CUSTOM
|| (metadata.type != ITUNES_METADATA_TYPE_STRING && metadata.type != ITUNES_METADATA_TYPE_BINARY)
|| !metadata.meaning || !metadata.name
|| memcmp( "com.apple.iTunes", metadata.meaning, strlen( metadata.meaning ) )
|| memcmp( "iTunSMPB", metadata.name, strlen( metadata.name ) ) )
{
lsmash_cleanup_itunes_metadata( &metadata );
continue;
}
char *value = NULL;
if( metadata.type == ITUNES_METADATA_TYPE_STRING )
{
int length = strlen( metadata.value.string );
if( length >= 116 )
value = duplicate_as_string( metadata.value.string, length );
}
else /* metadata.type == ITUNES_METADATA_TYPE_BINARY */
{
if( metadata.value.binary.size >= 116 )
value = duplicate_as_string( metadata.value.binary.data, metadata.value.binary.size );
}
lsmash_cleanup_itunes_metadata( &metadata );
if( !value )
continue;
uint32_t dummy[9];
uint32_t priming_samples;
uint32_t padding;
uint64_t duration;
if( 12 != sscanf( value, " %I32x %I32x %I32x %I64x %I32x %I32x %I32x %I32x %I32x %I32x %I32x %I32x",
&dummy[0], &priming_samples, &padding, &duration, &dummy[1], &dummy[2],
&dummy[3], &dummy[4], &dummy[5], &dummy[6], &dummy[7], &dummy[8] ) )
{
delete [] value;
continue;
}
delete [] value;
adh.implicit_preroll = 1;
aoh.skip_decoded_samples = priming_samples;
vi.num_audio_samples = duration + priming_samples;
break;
}
if( aoh.skip_decoded_samples == 0 )
{
uint32_t ctd_shift;
if( lsmash_get_composition_to_decode_shift_from_media_timeline( adh.root, adh.track_ID, &ctd_shift ) )
env->ThrowError( "LSMASHAudioSource: failed to get the timeline shift." );
aoh.skip_decoded_samples = ctd_shift + get_start_time( adh.root, adh.track_ID );
}
}
/* libavformat */
for( i = 0; i < format_ctx->nb_streams && format_ctx->streams[i]->codec->codec_type != AVMEDIA_TYPE_AUDIO; i++ );
if( i == format_ctx->nb_streams )
env->ThrowError( "LSMASHAudioSource: failed to find stream by libavformat." );
/* libavcodec */
AVStream *stream = format_ctx->streams[i];
AVCodecContext *ctx = stream->codec;
adh.config.ctx = ctx;
AVCodec *codec = avcodec_find_decoder( ctx->codec_id );
if( !codec )
env->ThrowError( "LSMASHAudioSource: failed to find %s decoder.", codec->name );
ctx->thread_count = 0;
if( avcodec_open2( ctx, codec, NULL ) < 0 )
env->ThrowError( "LSMASHAudioSource: failed to avcodec_open2." );
}
static void setup_timestamp_info( libavsmash_video_decode_handler_t *hp, VideoInfo *vi, uint64_t media_timescale, IScriptEnvironment *env )
{
if( vi->num_frames == 1 )
{
/* Calculate average framerate. */
uint64_t media_duration = lsmash_get_media_duration_from_media_timeline( hp->root, hp->track_ID );
if( media_duration == 0 )
media_duration = INT32_MAX;
reduce_fraction( &media_timescale, &media_duration );
vi->fps_numerator = (unsigned int)media_timescale;
vi->fps_denominator = (unsigned int)media_duration;
return;
}
lsmash_media_ts_list_t ts_list;
if( lsmash_get_media_timestamps( hp->root, hp->track_ID, &ts_list ) )
env->ThrowError( "LSMASHVideoSource: failed to get timestamps." );
if( ts_list.sample_count != vi->num_frames )
env->ThrowError( "LSMASHVideoSource: failed to count number of video samples." );
uint32_t composition_sample_delay;
if( lsmash_get_max_sample_delay( &ts_list, &composition_sample_delay ) )
{
lsmash_delete_media_timestamps( &ts_list );
env->ThrowError( "LSMASHVideoSource: failed to get composition delay." );
}
if( composition_sample_delay )
{
/* Consider composition order for keyframe detection.
* Note: sample number for L-SMASH is 1-origin. */
hp->order_converter = (order_converter_t *)malloc( (ts_list.sample_count + 1) * sizeof(order_converter_t) );
if( !hp->order_converter )
{
lsmash_delete_media_timestamps( &ts_list );
env->ThrowError( "LSMASHVideoSource: failed to allocate memory." );
}
for( uint32_t i = 0; i < ts_list.sample_count; i++ )
ts_list.timestamp[i].dts = i + 1;
lsmash_sort_timestamps_composition_order( &ts_list );
for( uint32_t i = 0; i < ts_list.sample_count; i++ )
hp->order_converter[i + 1].composition_to_decoding = (uint32_t)ts_list.timestamp[i].dts;
}
/* Calculate average framerate. */
uint64_t largest_cts = ts_list.timestamp[1].cts;
uint64_t second_largest_cts = ts_list.timestamp[0].cts;
uint64_t composition_timebase = ts_list.timestamp[1].cts - ts_list.timestamp[0].cts;
for( uint32_t i = 2; i < ts_list.sample_count; i++ )
{
if( ts_list.timestamp[i].cts == ts_list.timestamp[i - 1].cts )
{
lsmash_delete_media_timestamps( &ts_list );
return;
}
composition_timebase = get_gcd( composition_timebase, ts_list.timestamp[i].cts - ts_list.timestamp[i - 1].cts );
second_largest_cts = largest_cts;
largest_cts = ts_list.timestamp[i].cts;
}
uint64_t reduce = reduce_fraction( &media_timescale, &composition_timebase );
uint64_t composition_duration = ((largest_cts - ts_list.timestamp[0].cts) + (largest_cts - second_largest_cts)) / reduce;
lsmash_delete_media_timestamps( &ts_list );
vi->fps_numerator = (unsigned int)((vi->num_frames * ((double)media_timescale / composition_duration)) * composition_timebase + 0.5);
vi->fps_denominator = (unsigned int)composition_timebase;
}
