454 lines
		
	
	
		
			13 KiB
		
	
	
	
		
			JavaScript
		
	
	
	
	
	
		
		
			
		
	
	
			454 lines
		
	
	
		
			13 KiB
		
	
	
	
		
			JavaScript
		
	
	
	
	
	
|   | /** | ||
|  |  * mux.js | ||
|  |  * | ||
|  |  * Copyright (c) Brightcove | ||
|  |  * Licensed Apache-2.0 https://github.com/videojs/mux.js/blob/master/LICENSE
 | ||
|  |  */ | ||
|  | 'use strict'; | ||
|  | 
 | ||
|  | var Stream = require('../utils/stream.js'); | ||
|  | var FlvTag = require('./flv-tag.js'); | ||
|  | var m2ts = require('../m2ts/m2ts.js'); | ||
|  | var AdtsStream = require('../codecs/adts.js'); | ||
|  | var H264Stream = require('../codecs/h264').H264Stream; | ||
|  | var CoalesceStream = require('./coalesce-stream.js'); | ||
|  | var TagList = require('./tag-list.js'); | ||
|  | 
 | ||
|  | var | ||
|  |   Transmuxer, | ||
|  |   VideoSegmentStream, | ||
|  |   AudioSegmentStream, | ||
|  |   collectTimelineInfo, | ||
|  |   metaDataTag, | ||
|  |   extraDataTag; | ||
|  | 
 | ||
|  | /** | ||
|  |  * Store information about the start and end of the tracka and the | ||
|  |  * duration for each frame/sample we process in order to calculate | ||
|  |  * the baseMediaDecodeTime | ||
|  |  */ | ||
|  | collectTimelineInfo = function(track, data) { | ||
|  |   if (typeof data.pts === 'number') { | ||
|  |     if (track.timelineStartInfo.pts === undefined) { | ||
|  |       track.timelineStartInfo.pts = data.pts; | ||
|  |     } else { | ||
|  |       track.timelineStartInfo.pts = | ||
|  |         Math.min(track.timelineStartInfo.pts, data.pts); | ||
|  |     } | ||
|  |   } | ||
|  | 
 | ||
|  |   if (typeof data.dts === 'number') { | ||
|  |     if (track.timelineStartInfo.dts === undefined) { | ||
|  |       track.timelineStartInfo.dts = data.dts; | ||
|  |     } else { | ||
|  |       track.timelineStartInfo.dts = | ||
|  |         Math.min(track.timelineStartInfo.dts, data.dts); | ||
|  |     } | ||
|  |   } | ||
|  | }; | ||
|  | 
 | ||
|  | metaDataTag = function(track, pts) { | ||
|  |   var | ||
|  |     tag = new FlvTag(FlvTag.METADATA_TAG); // :FlvTag
 | ||
|  | 
 | ||
|  |   tag.dts = pts; | ||
|  |   tag.pts = pts; | ||
|  | 
 | ||
|  |   tag.writeMetaDataDouble('videocodecid', 7); | ||
|  |   tag.writeMetaDataDouble('width', track.width); | ||
|  |   tag.writeMetaDataDouble('height', track.height); | ||
|  | 
 | ||
|  |   return tag; | ||
|  | }; | ||
|  | 
 | ||
|  | extraDataTag = function(track, pts) { | ||
|  |   var | ||
|  |     i, | ||
|  |     tag = new FlvTag(FlvTag.VIDEO_TAG, true); | ||
|  | 
 | ||
|  |   tag.dts = pts; | ||
|  |   tag.pts = pts; | ||
|  | 
 | ||
|  |   tag.writeByte(0x01);// version
 | ||
|  |   tag.writeByte(track.profileIdc);// profile
 | ||
|  |   tag.writeByte(track.profileCompatibility);// compatibility
 | ||
|  |   tag.writeByte(track.levelIdc);// level
 | ||
|  |   tag.writeByte(0xFC | 0x03); // reserved (6 bits), NULA length size - 1 (2 bits)
 | ||
|  |   tag.writeByte(0xE0 | 0x01); // reserved (3 bits), num of SPS (5 bits)
 | ||
|  |   tag.writeShort(track.sps[0].length); // data of SPS
 | ||
|  |   tag.writeBytes(track.sps[0]); // SPS
 | ||
|  | 
 | ||
|  |   tag.writeByte(track.pps.length); // num of PPS (will there ever be more that 1 PPS?)
 | ||
|  |   for (i = 0; i < track.pps.length; ++i) { | ||
|  |     tag.writeShort(track.pps[i].length); // 2 bytes for length of PPS
 | ||
|  |     tag.writeBytes(track.pps[i]); // data of PPS
 | ||
|  |   } | ||
|  | 
 | ||
|  |   return tag; | ||
|  | }; | ||
|  | 
 | ||
|  | /** | ||
|  |  * Constructs a single-track, media segment from AAC data | ||
|  |  * events. The output of this stream can be fed to flash. | ||
|  |  */ | ||
|  | AudioSegmentStream = function(track) { | ||
|  |   var | ||
|  |     adtsFrames = [], | ||
|  |     videoKeyFrames = [], | ||
|  |     oldExtraData; | ||
|  | 
 | ||
|  |   AudioSegmentStream.prototype.init.call(this); | ||
|  | 
 | ||
|  |   this.push = function(data) { | ||
|  |     collectTimelineInfo(track, data); | ||
|  | 
 | ||
|  |     if (track) { | ||
|  |       track.audioobjecttype = data.audioobjecttype; | ||
|  |       track.channelcount = data.channelcount; | ||
|  |       track.samplerate = data.samplerate; | ||
|  |       track.samplingfrequencyindex = data.samplingfrequencyindex; | ||
|  |       track.samplesize = data.samplesize; | ||
|  |       track.extraData = (track.audioobjecttype << 11) | | ||
|  |                         (track.samplingfrequencyindex << 7) | | ||
|  |                         (track.channelcount << 3); | ||
|  |     } | ||
|  | 
 | ||
|  |     data.pts = Math.round(data.pts / 90); | ||
|  |     data.dts = Math.round(data.dts / 90); | ||
|  | 
 | ||
|  |     // buffer audio data until end() is called
 | ||
|  |     adtsFrames.push(data); | ||
|  |   }; | ||
|  | 
 | ||
|  |   this.flush = function() { | ||
|  |     var currentFrame, adtsFrame, lastMetaPts, tags = new TagList(); | ||
|  |     // return early if no audio data has been observed
 | ||
|  |     if (adtsFrames.length === 0) { | ||
|  |       this.trigger('done', 'AudioSegmentStream'); | ||
|  |       return; | ||
|  |     } | ||
|  | 
 | ||
|  |     lastMetaPts = -Infinity; | ||
|  | 
 | ||
|  |     while (adtsFrames.length) { | ||
|  |       currentFrame = adtsFrames.shift(); | ||
|  | 
 | ||
|  |       // write out a metadata frame at every video key frame
 | ||
|  |       if (videoKeyFrames.length && currentFrame.pts >= videoKeyFrames[0]) { | ||
|  |         lastMetaPts = videoKeyFrames.shift(); | ||
|  |         this.writeMetaDataTags(tags, lastMetaPts); | ||
|  |       } | ||
|  | 
 | ||
|  |       // also write out metadata tags every 1 second so that the decoder
 | ||
|  |       // is re-initialized quickly after seeking into a different
 | ||
|  |       // audio configuration.
 | ||
|  |       if (track.extraData !== oldExtraData || currentFrame.pts - lastMetaPts >= 1000) { | ||
|  |         this.writeMetaDataTags(tags, currentFrame.pts); | ||
|  |         oldExtraData = track.extraData; | ||
|  |         lastMetaPts = currentFrame.pts; | ||
|  |       } | ||
|  | 
 | ||
|  |       adtsFrame = new FlvTag(FlvTag.AUDIO_TAG); | ||
|  |       adtsFrame.pts = currentFrame.pts; | ||
|  |       adtsFrame.dts = currentFrame.dts; | ||
|  | 
 | ||
|  |       adtsFrame.writeBytes(currentFrame.data); | ||
|  | 
 | ||
|  |       tags.push(adtsFrame.finalize()); | ||
|  |     } | ||
|  | 
 | ||
|  |     videoKeyFrames.length = 0; | ||
|  |     oldExtraData = null; | ||
|  |     this.trigger('data', {track: track, tags: tags.list}); | ||
|  | 
 | ||
|  |     this.trigger('done', 'AudioSegmentStream'); | ||
|  |   }; | ||
|  | 
 | ||
|  |   this.writeMetaDataTags = function(tags, pts) { | ||
|  |     var adtsFrame; | ||
|  | 
 | ||
|  |     adtsFrame = new FlvTag(FlvTag.METADATA_TAG); | ||
|  |     // For audio, DTS is always the same as PTS. We want to set the DTS
 | ||
|  |     // however so we can compare with video DTS to determine approximate
 | ||
|  |     // packet order
 | ||
|  |     adtsFrame.pts = pts; | ||
|  |     adtsFrame.dts = pts; | ||
|  | 
 | ||
|  |     // AAC is always 10
 | ||
|  |     adtsFrame.writeMetaDataDouble('audiocodecid', 10); | ||
|  |     adtsFrame.writeMetaDataBoolean('stereo', track.channelcount === 2); | ||
|  |     adtsFrame.writeMetaDataDouble('audiosamplerate', track.samplerate); | ||
|  |     // Is AAC always 16 bit?
 | ||
|  |     adtsFrame.writeMetaDataDouble('audiosamplesize', 16); | ||
|  | 
 | ||
|  |     tags.push(adtsFrame.finalize()); | ||
|  | 
 | ||
|  |     adtsFrame = new FlvTag(FlvTag.AUDIO_TAG, true); | ||
|  |     // For audio, DTS is always the same as PTS. We want to set the DTS
 | ||
|  |     // however so we can compare with video DTS to determine approximate
 | ||
|  |     // packet order
 | ||
|  |     adtsFrame.pts = pts; | ||
|  |     adtsFrame.dts = pts; | ||
|  | 
 | ||
|  |     adtsFrame.view.setUint16(adtsFrame.position, track.extraData); | ||
|  |     adtsFrame.position += 2; | ||
|  |     adtsFrame.length = Math.max(adtsFrame.length, adtsFrame.position); | ||
|  | 
 | ||
|  |     tags.push(adtsFrame.finalize()); | ||
|  |   }; | ||
|  | 
 | ||
|  |   this.onVideoKeyFrame = function(pts) { | ||
|  |     videoKeyFrames.push(pts); | ||
|  |   }; | ||
|  | }; | ||
|  | AudioSegmentStream.prototype = new Stream(); | ||
|  | 
 | ||
|  | /** | ||
|  |  * Store FlvTags for the h264 stream | ||
|  |  * @param track {object} track metadata configuration | ||
|  |  */ | ||
|  | VideoSegmentStream = function(track) { | ||
|  |   var | ||
|  |     nalUnits = [], | ||
|  |     config, | ||
|  |     h264Frame; | ||
|  |   VideoSegmentStream.prototype.init.call(this); | ||
|  | 
 | ||
|  |   this.finishFrame = function(tags, frame) { | ||
|  |     if (!frame) { | ||
|  |       return; | ||
|  |     } | ||
|  |     // Check if keyframe and the length of tags.
 | ||
|  |     // This makes sure we write metadata on the first frame of a segment.
 | ||
|  |     if (config && track && track.newMetadata && | ||
|  |         (frame.keyFrame || tags.length === 0)) { | ||
|  |       // Push extra data on every IDR frame in case we did a stream change + seek
 | ||
|  |       var metaTag = metaDataTag(config, frame.dts).finalize(); | ||
|  |       var extraTag = extraDataTag(track, frame.dts).finalize(); | ||
|  | 
 | ||
|  |       metaTag.metaDataTag = extraTag.metaDataTag = true; | ||
|  | 
 | ||
|  |       tags.push(metaTag); | ||
|  |       tags.push(extraTag); | ||
|  |       track.newMetadata = false; | ||
|  | 
 | ||
|  |       this.trigger('keyframe', frame.dts); | ||
|  |     } | ||
|  | 
 | ||
|  |     frame.endNalUnit(); | ||
|  |     tags.push(frame.finalize()); | ||
|  |     h264Frame = null; | ||
|  |   }; | ||
|  | 
 | ||
|  |   this.push = function(data) { | ||
|  |     collectTimelineInfo(track, data); | ||
|  | 
 | ||
|  |     data.pts = Math.round(data.pts / 90); | ||
|  |     data.dts = Math.round(data.dts / 90); | ||
|  | 
 | ||
|  |     // buffer video until flush() is called
 | ||
|  |     nalUnits.push(data); | ||
|  |   }; | ||
|  | 
 | ||
|  |   this.flush = function() { | ||
|  |     var | ||
|  |       currentNal, | ||
|  |       tags = new TagList(); | ||
|  | 
 | ||
|  |     // Throw away nalUnits at the start of the byte stream until we find
 | ||
|  |     // the first AUD
 | ||
|  |     while (nalUnits.length) { | ||
|  |       if (nalUnits[0].nalUnitType === 'access_unit_delimiter_rbsp') { | ||
|  |         break; | ||
|  |       } | ||
|  |       nalUnits.shift(); | ||
|  |     } | ||
|  | 
 | ||
|  |     // return early if no video data has been observed
 | ||
|  |     if (nalUnits.length === 0) { | ||
|  |       this.trigger('done', 'VideoSegmentStream'); | ||
|  |       return; | ||
|  |     } | ||
|  | 
 | ||
|  |     while (nalUnits.length) { | ||
|  |       currentNal = nalUnits.shift(); | ||
|  | 
 | ||
|  |       // record the track config
 | ||
|  |       if (currentNal.nalUnitType === 'seq_parameter_set_rbsp') { | ||
|  |         track.newMetadata = true; | ||
|  |         config = currentNal.config; | ||
|  |         track.width = config.width; | ||
|  |         track.height = config.height; | ||
|  |         track.sps = [currentNal.data]; | ||
|  |         track.profileIdc = config.profileIdc; | ||
|  |         track.levelIdc = config.levelIdc; | ||
|  |         track.profileCompatibility = config.profileCompatibility; | ||
|  |         h264Frame.endNalUnit(); | ||
|  |       } else if (currentNal.nalUnitType === 'pic_parameter_set_rbsp') { | ||
|  |         track.newMetadata = true; | ||
|  |         track.pps = [currentNal.data]; | ||
|  |         h264Frame.endNalUnit(); | ||
|  |       } else if (currentNal.nalUnitType === 'access_unit_delimiter_rbsp') { | ||
|  |         if (h264Frame) { | ||
|  |           this.finishFrame(tags, h264Frame); | ||
|  |         } | ||
|  |         h264Frame = new FlvTag(FlvTag.VIDEO_TAG); | ||
|  |         h264Frame.pts = currentNal.pts; | ||
|  |         h264Frame.dts = currentNal.dts; | ||
|  |       } else { | ||
|  |         if (currentNal.nalUnitType === 'slice_layer_without_partitioning_rbsp_idr') { | ||
|  |           // the current sample is a key frame
 | ||
|  |           h264Frame.keyFrame = true; | ||
|  |         } | ||
|  |         h264Frame.endNalUnit(); | ||
|  |       } | ||
|  |       h264Frame.startNalUnit(); | ||
|  |       h264Frame.writeBytes(currentNal.data); | ||
|  |     } | ||
|  |     if (h264Frame) { | ||
|  |       this.finishFrame(tags, h264Frame); | ||
|  |     } | ||
|  | 
 | ||
|  |     this.trigger('data', {track: track, tags: tags.list}); | ||
|  | 
 | ||
|  |     // Continue with the flush process now
 | ||
|  |     this.trigger('done', 'VideoSegmentStream'); | ||
|  |   }; | ||
|  | }; | ||
|  | 
 | ||
|  | VideoSegmentStream.prototype = new Stream(); | ||
|  | 
 | ||
|  | /** | ||
|  |  * An object that incrementally transmuxes MPEG2 Trasport Stream | ||
|  |  * chunks into an FLV. | ||
|  |  */ | ||
|  | Transmuxer = function(options) { | ||
|  |   var | ||
|  |     self = this, | ||
|  | 
 | ||
|  |     packetStream, parseStream, elementaryStream, | ||
|  |     videoTimestampRolloverStream, audioTimestampRolloverStream, | ||
|  |     timedMetadataTimestampRolloverStream, | ||
|  |     adtsStream, h264Stream, | ||
|  |     videoSegmentStream, audioSegmentStream, captionStream, | ||
|  |     coalesceStream; | ||
|  | 
 | ||
|  |   Transmuxer.prototype.init.call(this); | ||
|  | 
 | ||
|  |   options = options || {}; | ||
|  | 
 | ||
|  |   // expose the metadata stream
 | ||
|  |   this.metadataStream = new m2ts.MetadataStream(); | ||
|  | 
 | ||
|  |   options.metadataStream = this.metadataStream; | ||
|  | 
 | ||
|  |   // set up the parsing pipeline
 | ||
|  |   packetStream = new m2ts.TransportPacketStream(); | ||
|  |   parseStream = new m2ts.TransportParseStream(); | ||
|  |   elementaryStream = new m2ts.ElementaryStream(); | ||
|  |   videoTimestampRolloverStream = new m2ts.TimestampRolloverStream('video'); | ||
|  |   audioTimestampRolloverStream = new m2ts.TimestampRolloverStream('audio'); | ||
|  |   timedMetadataTimestampRolloverStream = new m2ts.TimestampRolloverStream('timed-metadata'); | ||
|  | 
 | ||
|  |   adtsStream = new AdtsStream(); | ||
|  |   h264Stream = new H264Stream(); | ||
|  |   coalesceStream = new CoalesceStream(options); | ||
|  | 
 | ||
|  |   // disassemble MPEG2-TS packets into elementary streams
 | ||
|  |   packetStream | ||
|  |     .pipe(parseStream) | ||
|  |     .pipe(elementaryStream); | ||
|  | 
 | ||
|  |   // !!THIS ORDER IS IMPORTANT!!
 | ||
|  |   // demux the streams
 | ||
|  |   elementaryStream | ||
|  |     .pipe(videoTimestampRolloverStream) | ||
|  |     .pipe(h264Stream); | ||
|  |   elementaryStream | ||
|  |     .pipe(audioTimestampRolloverStream) | ||
|  |     .pipe(adtsStream); | ||
|  | 
 | ||
|  |   elementaryStream | ||
|  |     .pipe(timedMetadataTimestampRolloverStream) | ||
|  |     .pipe(this.metadataStream) | ||
|  |     .pipe(coalesceStream); | ||
|  |   // if CEA-708 parsing is available, hook up a caption stream
 | ||
|  |   captionStream = new m2ts.CaptionStream(options); | ||
|  |   h264Stream.pipe(captionStream) | ||
|  |     .pipe(coalesceStream); | ||
|  | 
 | ||
|  |   // hook up the segment streams once track metadata is delivered
 | ||
|  |   elementaryStream.on('data', function(data) { | ||
|  |     var i, videoTrack, audioTrack; | ||
|  | 
 | ||
|  |     if (data.type === 'metadata') { | ||
|  |       i = data.tracks.length; | ||
|  | 
 | ||
|  |       // scan the tracks listed in the metadata
 | ||
|  |       while (i--) { | ||
|  |         if (data.tracks[i].type === 'video') { | ||
|  |           videoTrack = data.tracks[i]; | ||
|  |         } else if (data.tracks[i].type === 'audio') { | ||
|  |           audioTrack = data.tracks[i]; | ||
|  |         } | ||
|  |       } | ||
|  | 
 | ||
|  |       // hook up the video segment stream to the first track with h264 data
 | ||
|  |       if (videoTrack && !videoSegmentStream) { | ||
|  |         coalesceStream.numberOfTracks++; | ||
|  |         videoSegmentStream = new VideoSegmentStream(videoTrack); | ||
|  | 
 | ||
|  |         // Set up the final part of the video pipeline
 | ||
|  |         h264Stream | ||
|  |           .pipe(videoSegmentStream) | ||
|  |           .pipe(coalesceStream); | ||
|  |       } | ||
|  | 
 | ||
|  |       if (audioTrack && !audioSegmentStream) { | ||
|  |         // hook up the audio segment stream to the first track with aac data
 | ||
|  |         coalesceStream.numberOfTracks++; | ||
|  |         audioSegmentStream = new AudioSegmentStream(audioTrack); | ||
|  | 
 | ||
|  |         // Set up the final part of the audio pipeline
 | ||
|  |         adtsStream | ||
|  |           .pipe(audioSegmentStream) | ||
|  |           .pipe(coalesceStream); | ||
|  | 
 | ||
|  |         if (videoSegmentStream) { | ||
|  |           videoSegmentStream.on('keyframe', audioSegmentStream.onVideoKeyFrame); | ||
|  |         } | ||
|  |       } | ||
|  |     } | ||
|  |   }); | ||
|  | 
 | ||
|  |   // feed incoming data to the front of the parsing pipeline
 | ||
|  |   this.push = function(data) { | ||
|  |     packetStream.push(data); | ||
|  |   }; | ||
|  | 
 | ||
|  |   // flush any buffered data
 | ||
|  |   this.flush = function() { | ||
|  |     // Start at the top of the pipeline and flush all pending work
 | ||
|  |     packetStream.flush(); | ||
|  |   }; | ||
|  | 
 | ||
|  |   // Caption data has to be reset when seeking outside buffered range
 | ||
|  |   this.resetCaptions = function() { | ||
|  |     captionStream.reset(); | ||
|  |   }; | ||
|  | 
 | ||
|  |   // Re-emit any data coming from the coalesce stream to the outside world
 | ||
|  |   coalesceStream.on('data', function(event) { | ||
|  |     self.trigger('data', event); | ||
|  |   }); | ||
|  | 
 | ||
|  |   // Let the consumer know we have finished flushing the entire pipeline
 | ||
|  |   coalesceStream.on('done', function() { | ||
|  |     self.trigger('done'); | ||
|  |   }); | ||
|  | }; | ||
|  | Transmuxer.prototype = new Stream(); | ||
|  | 
 | ||
|  | // forward compatibility
 | ||
|  | module.exports = Transmuxer; |