MSE: Add StreamParser buffer remuxing utility and tests

media/base/stream_parser_unittest.cc

+// Copyright 2014 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include <algorithm>
+#include <sstream>
+
+#include "base/basictypes.h"
+#include "media/base/stream_parser.h"
+#include "media/base/stream_parser_buffer.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace media {
+
+typedef StreamParser::BufferQueue BufferQueue;
+typedef StreamParser::TextBufferQueueMap TextBufferQueueMap;
+
+const int kEnd = -1;
+const uint8 kFakeData[] = { 0xFF };
+
+static bool IsAudio(scoped_refptr<StreamParserBuffer> buffer) {
+  return buffer->type() == DemuxerStream::AUDIO;
+}
+
+static bool IsVideo(scoped_refptr<StreamParserBuffer> buffer) {
+  return buffer->type() == DemuxerStream::VIDEO;
+}
+
+static bool IsText(scoped_refptr<StreamParserBuffer> buffer) {
+  return buffer->type() == DemuxerStream::TEXT;
+}
+
+// Creates and appends a sequence of StreamParserBuffers to the provided
+// |queue|. |decode_timestamps| determines the number of appended buffers and
+// their sequence of decode timestamps; a |kEnd| timestamp indicates the
+// end of the sequence and no buffer is appended for it. Each new buffer's
+// type will be |type|, and if text, its text track number will be set to
+// |text_track_number|.
+static void GenerateBuffers(const int* decode_timestamps,
+                            StreamParserBuffer::Type type,
+                            int text_track_number,
+                            BufferQueue* queue) {
+  DCHECK(decode_timestamps);
+  DCHECK(queue);
+  DCHECK_NE(type, DemuxerStream::UNKNOWN);
+  DCHECK_LT(type, DemuxerStream::NUM_TYPES);
+  for (int i = 0; decode_timestamps[i] != kEnd; ++i) {
+    scoped_refptr<StreamParserBuffer> buffer =
+        StreamParserBuffer::CopyFrom(kFakeData, sizeof(kFakeData),
+                                     true, type);
+    if (type == DemuxerStream::TEXT)
+      buffer->set_text_track_number(text_track_number);
+    buffer->SetDecodeTimestamp(
+        base::TimeDelta::FromMicroseconds(decode_timestamps[i]));
+    queue->push_back(buffer);
+  }
+}
+
+class StreamParserTest : public testing::Test {
+ protected:
+  StreamParserTest() {}
+
+  // Returns the number of buffers in |merged_buffers_| for which |predicate|
+  // returns true.
+  size_t CountMatchingMergedBuffers(
+      bool (*predicate)(scoped_refptr<StreamParserBuffer> buffer)) {
+    return static_cast<size_t>(count_if(merged_buffers_.begin(),
+                                        merged_buffers_.end(),
+                                        predicate));
+  }
+
+  // Appends test audio buffers in the sequence described by |decode_timestamps|
+  // to |audio_buffers_|. See GenerateBuffers() for |decode_timestamps| format.
+  void GenerateAudioBuffers(const int* decode_timestamps) {
+    GenerateBuffers(decode_timestamps, DemuxerStream::AUDIO, -1,
+                    &audio_buffers_);
+  }
+
+  // Appends test video buffers in the sequence described by |decode_timestamps|
+  // to |video_buffers_|. See GenerateBuffers() for |decode_timestamps| format.
+  void GenerateVideoBuffers(const int* decode_timestamps) {
+    GenerateBuffers(decode_timestamps, DemuxerStream::VIDEO, -1,
+                    &video_buffers_);
+  }
+
+  // Current tests only need up to two distinct text BufferQueues. This helper
+  // conditionally appends buffers to the underlying |text_buffers_a_| and
+  // |text_buffers_b_| and conditionally inserts these BufferQueues into
+  // |text_map_| keyed by the respective |text_track_number_{a,b}|. If
+  // |decode_timestamps_{a,b}| is NULL, then the corresponding BufferQueue is
+  // neither appended to nor inserted into |text_map_| (though it may previously
+  // have been inserted under either the same or different key). Note that
+  // key collision on map insertion does not replace the previous value.
+  void GenerateTextBuffers(const int* decode_timestamps_a,
+                           int text_track_number_a,
+                           const int* decode_timestamps_b,
+                           int text_track_number_b) {
+    if (decode_timestamps_a) {
+      GenerateBuffers(decode_timestamps_a, DemuxerStream::TEXT,
+                      text_track_number_a, &text_buffers_a_);
+      text_map_.insert(std::make_pair(text_track_number_a, text_buffers_a_));
+    }
+
+    if (decode_timestamps_b) {
+      GenerateBuffers(decode_timestamps_b, DemuxerStream::TEXT,
+                      text_track_number_b, &text_buffers_b_);
+      text_map_.insert(std::make_pair(text_track_number_b, text_buffers_b_));
+    }
+  }
+
+  // Returns a string that describes the sequence of buffers in
+  // |merged_buffers_|. The string is a concatenation of space-delimited buffer
+  // descriptors in the same sequence as |merged_buffers_|. Each descriptor is
+  // the concatenation of
+  // 1) a single character that describes the buffer's type(), e.g. A, V, or T
+  //    for audio, video, or text, respectively
+  // 2) if text, the buffer's text_track_number(), otherwise nothing
+  // 3) ":"
+  // 4) the buffer's decode timestamp.
+  // If |include_type_and_text_track| is false, then items 1, 2, and 3 are
+  // not included in descriptors. This is useful when buffers with different
+  // media types but the same decode timestamp are expected, and the exact
+  // sequence of media types for the tying timestamps is not subject to
+  // verification.
+  std::string MergedBufferQueueString(bool include_type_and_text_track) {
+    std::stringstream results_stream;
+    for (BufferQueue::const_iterator itr = merged_buffers_.begin();
+         itr != merged_buffers_.end();
+         ++itr) {
+      if (itr != merged_buffers_.begin())
+        results_stream << " ";
+      const StreamParserBuffer& buffer = *(*itr);
+      if (include_type_and_text_track) {
+        switch (buffer.type()) {
+          case DemuxerStream::AUDIO:
+            results_stream << "A";
+            break;
+          case DemuxerStream::VIDEO:
+            results_stream << "V";
+            break;
+          case DemuxerStream::TEXT:
+            results_stream << "T";
+            results_stream << buffer.text_track_number();
+            break;
+          default:
+            NOTREACHED();
+        }
+        results_stream << ":";
+      }
+      results_stream << buffer.GetDecodeTimestamp().InMicroseconds();
+    }
+
+    return results_stream.str();
+  }
+
+  // Verifies that MergeBufferQueues() of the current |audio_buffers_|,
+  // |video_buffers_|, |text_map_|, and |merged_buffers_| returns true and
+  // results in an updated |merged_buffers_| that matches expectation. The
+  // expectation, specified in |expected|, is compared to the string resulting
+  // from MergedBufferQueueString() (see comments for that method) with
+  // |verify_type_and_text_track_sequence| passed. |merged_buffers_| is appended
+  // to by the merge, and may be setup by the caller to have some pre-existing
+  // buffers; it is both an input and output of this method.
+  // Regardless of |verify_type_and_text_track_sequence|, the marginal number
+  // of buffers of each type (audio, video, text) resulting from the merge is
+  // also verified to match the number of buffers in |audio_buffers_|,
+  // |video_buffers_|, and |text_map_|, respectively.
+  void VerifyMergeSuccess(const std::string& expected,
+                          bool verify_type_and_text_track_sequence) {
+    // |merged_buffers| may already have some buffers. Count them by type for
+    // later inclusion in verification.
+    size_t original_audio_in_merged = CountMatchingMergedBuffers(IsAudio);
+    size_t original_video_in_merged = CountMatchingMergedBuffers(IsVideo);
+    size_t original_text_in_merged = CountMatchingMergedBuffers(IsText);
+
+    EXPECT_TRUE(MergeBufferQueues(audio_buffers_, video_buffers_, text_map_,
+                                  &merged_buffers_));
+
+    // Verify resulting contents of |merged_buffers| matches |expected|.
+    EXPECT_EQ(expected,
+              MergedBufferQueueString(verify_type_and_text_track_sequence));
+
+    // Verify that the correct number of each type of buffer is in the merge
+    // result.
+    size_t audio_in_merged = CountMatchingMergedBuffers(IsAudio);
+    size_t video_in_merged = CountMatchingMergedBuffers(IsVideo);
+    size_t text_in_merged = CountMatchingMergedBuffers(IsText);
+
+    EXPECT_GE(audio_in_merged, original_audio_in_merged);
+    EXPECT_GE(video_in_merged, original_video_in_merged);
+    EXPECT_GE(text_in_merged, original_text_in_merged);
+
+    EXPECT_EQ(audio_buffers_.size(),
+              audio_in_merged - original_audio_in_merged);
+    EXPECT_EQ(video_buffers_.size(),
+              video_in_merged - original_video_in_merged);
+
+    size_t expected_text_buffer_count = 0;
+    for (TextBufferQueueMap::const_iterator itr = text_map_.begin();
+         itr != text_map_.end();
+         ++itr) {
+      expected_text_buffer_count += itr->second.size();
+    }
+    EXPECT_EQ(expected_text_buffer_count,
+              text_in_merged - original_text_in_merged);
+  }
+
+  // Verifies that MergeBufferQueues() of the current |audio_buffers_|,
+  // |video_buffers_|, |text_map_|, and |merged_buffers_| returns false.
+  void VerifyMergeFailure() {
+    EXPECT_FALSE(MergeBufferQueues(audio_buffers_, video_buffers_, text_map_,
+                                   &merged_buffers_));
+  }
+
+  // Helper to allow tests to clear all the input BufferQueues (except
+  // |merged_buffers_|) and the TextBufferQueueMap that are used in
+  // VerifyMerge{Success/Failure}().
+  void ClearQueuesAndTextMapButKeepAnyMergedBuffers() {
+    audio_buffers_.clear();
+    video_buffers_.clear();
+    text_buffers_a_.clear();
+    text_buffers_b_.clear();
+    text_map_.clear();
+  }
+
+ private:
+  BufferQueue audio_buffers_, video_buffers_, text_buffers_a_, text_buffers_b_;

[xhwang, 2014/02/06 00:36:32]

I don't find a rule against this, but I rarely see this style. Usually we
declare each variable on a separate line. This style tends to be tricky for
pointers (which isn't the case here though), e.g. int* a, b;

[wolenetz, 2014/02/06 23:56:03]

Done.

+  BufferQueue merged_buffers_;
+  TextBufferQueueMap text_map_;
+
+  DISALLOW_COPY_AND_ASSIGN(StreamParserTest);
+};
+
+TEST_F(StreamParserTest, MergeBufferQueues_AllEmpty) {
+  std::string expected = "";
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_SingleAudioBuffer) {
+  std::string expected = "A:100";
+  int audio_timestamps[] = { 100, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_SingleVideoBuffer) {
+  std::string expected = "V:100";
+  int video_timestamps[] = { 100, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_SingleTextBuffer) {
+  std::string expected = "T12:100";
+  int text_timestamps[] = { 100, kEnd };
+  GenerateTextBuffers(text_timestamps, 12, NULL, -1);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_OverlappingAudioVideo) {
+  std::string expected = "A:100 V:101 V:102 A:103 A:104 V:105";
+  int audio_timestamps[] = { 100, 103, 104, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 101, 102, 105, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_OverlappingMultipleText) {
+  std::string expected = "T1:100 T1:101 T2:103 T1:104 T2:105 T2:106";
+  int text_timestamps_1[] = { 100, 101, 104, kEnd };
+  int text_timestamps_2[] = { 103, 105, 106, kEnd };
+  GenerateTextBuffers(text_timestamps_1, 1, text_timestamps_2, 2);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_OverlappingAudioVideoText) {
+  std::string expected = "A:100 V:101 T1:102 V:103 T2:104 A:105 V:106 T1:107";
+  int audio_timestamps[] = { 100, 105, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 101, 103, 106, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  int text_timestamps_1[] = { 102, 107, kEnd };
+  int text_timestamps_2[] = { 104, kEnd };
+  GenerateTextBuffers(text_timestamps_1, 1, text_timestamps_2, 2);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_NonDecreasingNoCrossMediaDuplicate) {
+  std::string expected = "A:100 A:100 A:100 V:101 V:101 V:101 A:102 V:103 "
+                         "V:103";
+  int audio_timestamps[] = { 100, 100, 100, 102, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 101, 101, 101, 103, 103, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_CrossStreamDuplicates) {
+  // Interface keeps the choice undefined of which stream's buffer wins the
+  // selection when timestamps are tied. Verify at least the right number of
+  // each kind of buffer results, and that buffers are in nondecreasing order.
+  std::string expected = "100 100 100 100 100 100 102 102 102 102 102 102 102";
+  int audio_timestamps[] = { 100, 100, 100, 102, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 100, 100, 102, 102, 102, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  int text_timestamps[] = { 100, 102, 102, 102, kEnd };
+  GenerateTextBuffers(text_timestamps, 1, NULL, -1);
+  VerifyMergeSuccess(expected, false);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_InvalidDecreasingSingleStream) {
+  int audio_timestamps[] = { 101, 102, 100, 103, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  VerifyMergeFailure();
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_InvalidDecreasingMultipleStreams) {
+  int audio_timestamps[] = { 101, 102, 100, 103, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 104, 100, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  VerifyMergeFailure();
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_ValidAppendToExistingMerge) {
+  std::string expected = "A:100 V:101 T1:102 V:103 T2:104 A:105 V:106 T1:107";
+  int audio_timestamps[] = { 100, 105, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 101, 103, 106, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  int text_timestamps_1[] = { 102, 107, kEnd };
+  int text_timestamps_2[] = { 104, kEnd };
+  GenerateTextBuffers(text_timestamps_1, 1, text_timestamps_2, 2);
+  VerifyMergeSuccess(expected, true);
+
+  ClearQueuesAndTextMapButKeepAnyMergedBuffers();
+
+  expected = "A:100 V:101 T1:102 V:103 T2:104 A:105 V:106 T1:107 "
+             "A:107 V:111 T1:112 V:113 T2:114 A:115 V:116 T1:117";
+  int more_audio_timestamps[] = { 107, 115, kEnd };
+  GenerateAudioBuffers(more_audio_timestamps);
+  int more_video_timestamps[] = { 111, 113, 116, kEnd };
+  GenerateVideoBuffers(more_video_timestamps);
+  int more_text_timestamps_1[] = { 112, 117, kEnd };
+  int more_text_timestamps_2[] = { 114, kEnd };
+  GenerateTextBuffers(more_text_timestamps_1, 1, more_text_timestamps_2, 2);
+  VerifyMergeSuccess(expected, true);
+}
+
+TEST_F(StreamParserTest, MergeBufferQueues_InvalidAppendToExistingMerge) {
+  std::string expected = "A:100 V:101 T1:102 V:103 T2:104 A:105 V:106 T1:107";
+  int audio_timestamps[] = { 100, 105, kEnd };
+  GenerateAudioBuffers(audio_timestamps);
+  int video_timestamps[] = { 101, 103, 106, kEnd };
+  GenerateVideoBuffers(video_timestamps);
+  int text_timestamps_1[] = { 102, 107, kEnd };
+  int text_timestamps_2[] = { 104, kEnd };
+  GenerateTextBuffers(text_timestamps_1, 1, text_timestamps_2, 2);
+  VerifyMergeSuccess(expected, true);
+
+  // Appending empty buffers to pre-existing merge result should succeed and not
+  // change the existing result.
+  ClearQueuesAndTextMapButKeepAnyMergedBuffers();
+  VerifyMergeSuccess(expected, true);
+
+  // But appending something with a lower timestamp than the last timestamp
+  // in the pre-existing merge result should fail.
+  int more_audio_timestamps[] = { 106, kEnd };
+  GenerateAudioBuffers(more_audio_timestamps);
+  VerifyMergeFailure();
+}
+
+}  // namespace media
+