Media: Add thumbnail extraction to AudioVideoMetadataExtractor.

Spinoff of patch https://codereview.chromium.org/250143002/

For trivial change in chrome/utility/media_galleries/media_metadata_parser.cc:
TBR=vandebo

BUG=304290

Review URL: https://codereview.chromium.org/270873003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@269588 0039d316-1c4b-4281-b951-d872f2087c98

chrome/utility/media_galleries/media_metadata_parser.cc

@@ -56,7 +56,8 @@ scoped_ptr<MediaMetadataParser::MediaMetadata> ParseAudioVideoMetadata(
   DCHECK(metadata.get());
   media::AudioVideoMetadataExtractor extractor;
 
-  if (!extractor.Extract(source))
+  // TODO(tommycli): Add attached picture extraction.
+  if (!extractor.Extract(source, false /* extract_attached_pics */))
     return metadata.Pass();
 
   if (extractor.duration() >= 0)

media/base/audio_video_metadata_extractor.cc

@@ -47,6 +47,9 @@ bool ExtractInt(AVDictionaryEntry* tag, const char* expected_key,
   return true;
 }
 
+// Set attached image size limit to 4MB. Chosen arbitrarily.
+const int kAttachedImageSizeLimit = 4 * 1024 * 1024;
+
 }  // namespace
 
 AudioVideoMetadataExtractor::StreamInfo::StreamInfo() {}
@@ -66,7 +69,8 @@ AudioVideoMetadataExtractor::AudioVideoMetadataExtractor()
 AudioVideoMetadataExtractor::~AudioVideoMetadataExtractor() {
 }
 
-bool AudioVideoMetadataExtractor::Extract(DataSource* source) {
+bool AudioVideoMetadataExtractor::Extract(DataSource* source,
+                                          bool extract_attached_images) {
   DCHECK(!extracted_);
 
   bool read_ok = true;
@@ -111,12 +115,24 @@ bool AudioVideoMetadataExtractor::Extract(DataSource* source) {
 
     info.type = avcodec_get_name(stream->codec->codec_id);
 
-    // Extract dimensions of largest stream that's not an attached picture.
+    // Extract dimensions of largest stream that's not an attached image.
     if (stream->codec->width > 0 && stream->codec->width > width_ &&
         stream->codec->height > 0 && stream->codec->height > height_) {
       width_ = stream->codec->width;
       height_ = stream->codec->height;
     }
+
+    // Extract attached image if requested.
+    if (extract_attached_images &&
+        stream->disposition == AV_DISPOSITION_ATTACHED_PIC &&
+        stream->attached_pic.size > 0 &&
+        stream->attached_pic.size <= kAttachedImageSizeLimit &&
+        stream->attached_pic.data != NULL) {
+      attached_images_bytes_.push_back(std::string());
+      attached_images_bytes_.back().assign(
+          reinterpret_cast<const char*>(stream->attached_pic.data),
+          stream->attached_pic.size);
+    }
   }
 
   extracted_ = true;
@@ -209,6 +225,12 @@ AudioVideoMetadataExtractor::stream_infos() const {
   return stream_infos_;
 }
 
+const std::vector<std::string>&
+AudioVideoMetadataExtractor::attached_images_bytes() const {
+  DCHECK(extracted_);
+  return attached_images_bytes_;
+}
+
 void AudioVideoMetadataExtractor::ExtractDictionary(
     AVDictionary* metadata, TagDictionary* raw_tags) {
   if (!metadata)

media/base/audio_video_metadata_extractor.h

@@ -38,7 +38,7 @@ class MEDIA_EXPORT AudioVideoMetadataExtractor {
 
   // Returns whether or not the fields were successfully extracted. Should only
   // be called once.
-  bool Extract(DataSource* source);
+  bool Extract(DataSource* source, bool extract_attached_pics);
 
   // Returns -1 if we cannot extract the duration. In seconds.
   double duration() const;
@@ -67,6 +67,10 @@ class MEDIA_EXPORT AudioVideoMetadataExtractor {
   // First element is the container. Subsequent elements are the child streams.
   const StreamInfoVector& stream_infos() const;
 
+  // Empty if Extract call did not request attached images, or if no attached
+  // images were found.
+  const std::vector<std::string>& attached_images_bytes() const;
+
  private:
   void ExtractDictionary(AVDictionary* metadata, TagDictionary* raw_tags);
 
@@ -92,6 +96,8 @@ class MEDIA_EXPORT AudioVideoMetadataExtractor {
 
   StreamInfoVector stream_infos_;
 
+  std::vector<std::string> attached_images_bytes_;
+
   DISALLOW_COPY_AND_ASSIGN(AudioVideoMetadataExtractor);
 };
 

media/base/audio_video_metadata_extractor_unittest.cc

@@ -4,6 +4,7 @@
 
 #include "base/logging.h"
 #include "base/memory/scoped_ptr.h"
+#include "base/sha1.h"
 #include "build/build_config.h"
 #include "media/base/audio_video_metadata_extractor.h"
 #include "media/base/test_data_util.h"
@@ -14,6 +15,7 @@ namespace media {
 
 scoped_ptr<AudioVideoMetadataExtractor> GetExtractor(
     const std::string& filename,
+    bool extract_attached_images,
     bool expected_result,
     double expected_duration,
     int expected_width,
@@ -23,7 +25,7 @@ scoped_ptr<AudioVideoMetadataExtractor> GetExtractor(
 
   scoped_ptr<AudioVideoMetadataExtractor> extractor(
       new AudioVideoMetadataExtractor);
-  bool extracted = extractor->Extract(&source);
+  bool extracted = extractor->Extract(&source, extract_attached_images);
   EXPECT_EQ(expected_result, extracted);
 
   if (!extracted)
@@ -38,12 +40,12 @@ scoped_ptr<AudioVideoMetadataExtractor> GetExtractor(
 }
 
 TEST(AudioVideoMetadataExtractorTest, InvalidFile) {
-  GetExtractor("ten_byte_file", false, 0, -1, -1);
+  GetExtractor("ten_byte_file", true, false, 0, -1, -1);
 }
 
 TEST(AudioVideoMetadataExtractorTest, AudioOGG) {
   scoped_ptr<AudioVideoMetadataExtractor> extractor =
-      GetExtractor("9ch.ogg", true, 0, -1, -1);
+      GetExtractor("9ch.ogg", true, true, 0, -1, -1);
   EXPECT_EQ("Processed by SoX", extractor->comment());
 
   EXPECT_EQ("ogg", extractor->stream_infos()[0].type);
@@ -55,11 +57,13 @@ TEST(AudioVideoMetadataExtractorTest, AudioOGG) {
   EXPECT_EQ("vorbis", extractor->stream_infos()[1].type);
   EXPECT_EQ("Processed by SoX",
             extractor->stream_infos()[1].tags.find("COMMENT")->second);
+
+  EXPECT_EQ(0u, extractor->attached_images_bytes().size());
 }
 
 TEST(AudioVideoMetadataExtractorTest, AudioWAV) {
   scoped_ptr<AudioVideoMetadataExtractor> extractor =
-      GetExtractor("sfx_u8.wav", true, 0, -1, -1);
+      GetExtractor("sfx_u8.wav", true, true, 0, -1, -1);
   EXPECT_EQ("Lavf54.37.100", extractor->encoder());
   EXPECT_EQ("Amadeus Pro", extractor->encoded_by());
 
@@ -74,11 +78,13 @@ TEST(AudioVideoMetadataExtractorTest, AudioWAV) {
 
   EXPECT_EQ("pcm_u8", extractor->stream_infos()[1].type);
   EXPECT_EQ(0u, extractor->stream_infos()[1].tags.size());
+
+  EXPECT_EQ(0u, extractor->attached_images_bytes().size());
 }
 
 TEST(AudioVideoMetadataExtractorTest, VideoWebM) {
   scoped_ptr<AudioVideoMetadataExtractor> extractor =
-      GetExtractor("bear-320x240-multitrack.webm", true, 2, 320, 240);
+      GetExtractor("bear-320x240-multitrack.webm", true, true, 2, 320, 240);
   EXPECT_EQ("Lavf53.9.0", extractor->encoder());
 
   EXPECT_EQ(6u, extractor->stream_infos().size());
@@ -104,12 +110,14 @@ TEST(AudioVideoMetadataExtractorTest, VideoWebM) {
   EXPECT_EQ(1u, extractor->stream_infos()[5].tags.size());
   EXPECT_EQ("Lavc52.32.0",
             extractor->stream_infos()[5].tags.find("ENCODER")->second);
+
+  EXPECT_EQ(0u, extractor->attached_images_bytes().size());
 }
 
 #if defined(USE_PROPRIETARY_CODECS)
 TEST(AudioVideoMetadataExtractorTest, AndroidRotatedMP4Video) {
   scoped_ptr<AudioVideoMetadataExtractor> extractor =
-      GetExtractor("90rotation.mp4", true, 0, 1920, 1080);
+      GetExtractor("90rotation.mp4", true, true, 0, 1920, 1080);
 
   EXPECT_EQ(90, extractor->rotation());
 
@@ -144,11 +152,13 @@ TEST(AudioVideoMetadataExtractorTest, AndroidRotatedMP4Video) {
   EXPECT_EQ("SoundHandle",
             extractor->stream_infos()[2].tags.find("handler_name")->second);
   EXPECT_EQ("eng", extractor->stream_infos()[2].tags.find("language")->second);
+
+  EXPECT_EQ(0u, extractor->attached_images_bytes().size());
 }
 
 TEST(AudioVideoMetadataExtractorTest, AudioMP3) {
   scoped_ptr<AudioVideoMetadataExtractor> extractor =
-      GetExtractor("id3_png_test.mp3", true, 1, -1, -1);
+      GetExtractor("id3_png_test.mp3", true, true, 1, -1, -1);
 
   EXPECT_EQ("Airbag", extractor->title());
   EXPECT_EQ("Radiohead", extractor->artist());
@@ -181,6 +191,17 @@ TEST(AudioVideoMetadataExtractorTest, AudioMP3) {
   EXPECT_EQ(2u, extractor->stream_infos()[2].tags.size());
   EXPECT_EQ("Other", extractor->stream_infos()[2].tags.find("comment")->second);
   EXPECT_EQ("", extractor->stream_infos()[2].tags.find("title")->second);
+
+  EXPECT_EQ(1u, extractor->attached_images_bytes().size());
+  EXPECT_EQ(155752u, extractor->attached_images_bytes()[0].size());
+
+  EXPECT_EQ("\x89PNG\r\n\x1a\n",
+            extractor->attached_images_bytes()[0].substr(0, 8));
+  EXPECT_EQ("IEND\xae\x42\x60\x82",
+            extractor->attached_images_bytes()[0].substr(
+                extractor->attached_images_bytes()[0].size() - 8, 8));
+  EXPECT_EQ("\xF3\xED\x8F\xC7\xC7\x98\xB9V|p\xC0u!\xB5\x82\xCF\x95\xF0\xCD\xCE",
+            base::SHA1HashString(extractor->attached_images_bytes()[0]));
 }
 #endif