media/formats/mp4/mp4_stream_parser.h

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #ifndef MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_
   6 #define MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_
   7
   8 #include <set>
   9 #include <vector>
  10
  11 #include "base/basictypes.h"
  12 #include "base/callback.h"
  13 #include "base/compiler_specific.h"
  14 #include "base/memory/scoped_ptr.h"
  15 #include "media/base/media_export.h"
  16 #include "media/base/stream_parser.h"
  17 #include "media/formats/common/offset_byte_queue.h"
  18 #include "media/formats/mp4/track_run_iterator.h"
  19
  20 namespace media {
  21 namespace mp4 {
  22
  23 struct Movie;
  24 class BoxReader;
  25
  26 class MEDIA_EXPORT MP4StreamParser : public StreamParser {
  27  public:
  28   MP4StreamParser(const std::set<int>& audio_object_types, bool has_sbr);
  29   ~MP4StreamParser() override;
  30
  31   void Init(
  32       const InitCB& init_cb,
  33       const NewConfigCB& config_cb,
  34       const NewBuffersCB& new_buffers_cb,
  35       bool ignore_text_tracks,
  36       const EncryptedMediaInitDataCB& encrypted_media_init_data_cb,
  37       const NewMediaSegmentCB& new_segment_cb,
  38       const base::Closure& end_of_segment_cb,
  39       const LogCB& log_cb) override;
  40   void Flush() override;
  41   bool Parse(const uint8* buf, int size) override;
  42
  43  private:
  44   enum State {
  45     kWaitingForInit,
  46     kParsingBoxes,
  47     kWaitingForSampleData,
  48     kEmittingSamples,
  49     kError
  50   };
  51
  52   bool ParseBox(bool* err);
  53   bool ParseMoov(mp4::BoxReader* reader);
  54   bool ParseMoof(mp4::BoxReader* reader);
  55
  56   void OnEncryptedMediaInitData(
  57       const std::vector<ProtectionSystemSpecificHeader>& headers);
  58
  59   // To retain proper framing, each 'mdat' atom must be read; to limit memory
  60   // usage, the atom's data needs to be discarded incrementally as frames are
  61   // extracted from the stream. This function discards data from the stream up
  62   // to |max_clear_offset|, updating the |mdat_tail_| value so that framing can
  63   // be retained after all 'mdat' information has been read. |max_clear_offset|
  64   // is the upper bound on what can be removed from |queue_|. Anything below
  65   // this offset is no longer needed by the parser.
  66   // Returns 'true' on success, 'false' if there was an error.
  67   bool ReadAndDiscardMDATsUntil(int64 max_clear_offset);
  68
  69   void ChangeState(State new_state);
  70
  71   bool EmitConfigs();
  72   bool PrepareAVCBuffer(const AVCDecoderConfigurationRecord& avc_config,
  73                         std::vector<uint8>* frame_buf,
  74                         std::vector<SubsampleEntry>* subsamples) const;
  75   bool PrepareAACBuffer(const AAC& aac_config,
  76                         std::vector<uint8>* frame_buf,
  77                         std::vector<SubsampleEntry>* subsamples) const;
  78   bool EnqueueSample(BufferQueue* audio_buffers,
  79                      BufferQueue* video_buffers,
  80                      bool* err);
  81   bool SendAndFlushSamples(BufferQueue* audio_buffers,
  82                            BufferQueue* video_buffers);
  83
  84   void Reset();
  85
  86   // Checks to see if we have enough data in |queue_| to transition to
  87   // kEmittingSamples and start enqueuing samples.
  88   bool HaveEnoughDataToEnqueueSamples();
  89
  90   // Sets |highest_end_offset_| based on the data in |moov_|
  91   // and |moof|. Returns true if |highest_end_offset_| was successfully
  92   // computed.
  93   bool ComputeHighestEndOffset(const MovieFragment& moof);
  94
  95   State state_;
  96   InitCB init_cb_;
  97   NewConfigCB config_cb_;
  98   NewBuffersCB new_buffers_cb_;
  99   EncryptedMediaInitDataCB encrypted_media_init_data_cb_;
 100   NewMediaSegmentCB new_segment_cb_;
 101   base::Closure end_of_segment_cb_;
 102   LogCB log_cb_;
 103
 104   OffsetByteQueue queue_;
 105
 106   // These two parameters are only valid in the |kEmittingSegments| state.
 107   //
 108   // |moof_head_| is the offset of the start of the most recently parsed moof
 109   // block. All byte offsets in sample information are relative to this offset,
 110   // as mandated by the Media Source spec.
 111   int64 moof_head_;
 112   // |mdat_tail_| is the stream offset of the end of the current 'mdat' box.
 113   // Valid iff it is greater than the head of the queue.
 114   int64 mdat_tail_;
 115
 116   // The highest end offset in the current moof. This offset is
 117   // relative to |moof_head_|. This value is used to make sure we have collected
 118   // enough bytes to parse all samples and aux_info in the current moof.
 119   int64 highest_end_offset_;
 120
 121   scoped_ptr<mp4::Movie> moov_;
 122   scoped_ptr<mp4::TrackRunIterator> runs_;
 123
 124   bool has_audio_;
 125   bool has_video_;
 126   uint32 audio_track_id_;
 127   uint32 video_track_id_;
 128   // The object types allowed for audio tracks.
 129   std::set<int> audio_object_types_;
 130   bool has_sbr_;
 131   bool is_audio_track_encrypted_;
 132   bool is_video_track_encrypted_;
 133
 134   // Tracks the number of MEDIA_LOGs for skipping top level boxes. Useful to
 135   // prevent log spam.
 136   int num_top_level_box_skipped_;
 137
 138   DISALLOW_COPY_AND_ASSIGN(MP4StreamParser);
 139 };
 140
 141 }  // namespace mp4
 142 }  // namespace media
 143
 144 #endif  // MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_