mp4_stream_parser.h revision 116680a4aac90f2aa7413d9095a592090648e557 
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
20namespace media {
21namespace mp4 {
22
23struct Movie;
24class BoxReader;
25
26class MEDIA_EXPORT MP4StreamParser : public StreamParser {
27 public:
28  MP4StreamParser(const std::set<int>& audio_object_types, bool has_sbr);
29  virtual ~MP4StreamParser();
30
31  virtual void Init(const InitCB& init_cb, const NewConfigCB& config_cb,
32                    const NewBuffersCB& new_buffers_cb,
33                    bool ignore_text_tracks,
34                    const NeedKeyCB& need_key_cb,
35                    const NewMediaSegmentCB& new_segment_cb,
36                    const base::Closure& end_of_segment_cb,
37                    const LogCB& log_cb) OVERRIDE;
38  virtual void Flush() OVERRIDE;
39  virtual bool Parse(const uint8* buf, int size) OVERRIDE;
40
41 private:
42  enum State {
43    kWaitingForInit,
44    kParsingBoxes,
45    kWaitingForSampleData,
46    kEmittingSamples,
47    kError
48  };
49
50  bool ParseBox(bool* err);
51  bool ParseMoov(mp4::BoxReader* reader);
52  bool ParseMoof(mp4::BoxReader* reader);
53
54  void EmitNeedKeyIfNecessary(
55      const std::vector<ProtectionSystemSpecificHeader>& headers);
56
57  // To retain proper framing, each 'mdat' atom must be read; to limit memory
58  // usage, the atom's data needs to be discarded incrementally as frames are
59  // extracted from the stream. This function discards data from the stream up
60  // to |max_clear_offset|, updating the |mdat_tail_| value so that framing can
61  // be retained after all 'mdat' information has been read. |max_clear_offset|
62  // is the upper bound on what can be removed from |queue_|. Anything below
63  // this offset is no longer needed by the parser.
64  // Returns 'true' on success, 'false' if there was an error.
65  bool ReadAndDiscardMDATsUntil(int64 max_clear_offset);
66
67  void ChangeState(State new_state);
68
69  bool EmitConfigs();
70  bool PrepareAVCBuffer(const AVCDecoderConfigurationRecord& avc_config,
71                        std::vector<uint8>* frame_buf,
72                        std::vector<SubsampleEntry>* subsamples) const;
73  bool PrepareAACBuffer(const AAC& aac_config,
74                        std::vector<uint8>* frame_buf,
75                        std::vector<SubsampleEntry>* subsamples) const;
76  bool EnqueueSample(BufferQueue* audio_buffers,
77                     BufferQueue* video_buffers,
78                     bool* err);
79  bool SendAndFlushSamples(BufferQueue* audio_buffers,
80                           BufferQueue* video_buffers);
81
82  void Reset();
83
84  // Checks to see if we have enough data in |queue_| to transition to
85  // kEmittingSamples and start enqueuing samples.
86  bool HaveEnoughDataToEnqueueSamples();
87
88  // Sets |highest_end_offset_| based on the data in |moov_|
89  // and |moof|. Returns true if |highest_end_offset_| was successfully
90  // computed.
91  bool ComputeHighestEndOffset(const MovieFragment& moof);
92
93  State state_;
94  InitCB init_cb_;
95  NewConfigCB config_cb_;
96  NewBuffersCB new_buffers_cb_;
97  NeedKeyCB need_key_cb_;
98  NewMediaSegmentCB new_segment_cb_;
99  base::Closure end_of_segment_cb_;
100  LogCB log_cb_;
101
102  OffsetByteQueue queue_;
103
104  // These two parameters are only valid in the |kEmittingSegments| state.
105  //
106  // |moof_head_| is the offset of the start of the most recently parsed moof
107  // block. All byte offsets in sample information are relative to this offset,
108  // as mandated by the Media Source spec.
109  int64 moof_head_;
110  // |mdat_tail_| is the stream offset of the end of the current 'mdat' box.
111  // Valid iff it is greater than the head of the queue.
112  int64 mdat_tail_;
113
114  // The highest end offset in the current moof. This offset is
115  // relative to |moof_head_|. This value is used to make sure we have collected
116  // enough bytes to parse all samples and aux_info in the current moof.
117  int64 highest_end_offset_;
118
119  scoped_ptr<mp4::Movie> moov_;
120  scoped_ptr<mp4::TrackRunIterator> runs_;
121
122  bool has_audio_;
123  bool has_video_;
124  uint32 audio_track_id_;
125  uint32 video_track_id_;
126  // The object types allowed for audio tracks.
127  std::set<int> audio_object_types_;
128  bool has_sbr_;
129  bool is_audio_track_encrypted_;
130  bool is_video_track_encrypted_;
131
132  DISALLOW_COPY_AND_ASSIGN(MP4StreamParser);
133};
134
135}  // namespace mp4
136}  // namespace media
137
138#endif  // MEDIA_FORMATS_MP4_MP4_STREAM_PARSER_H_
139