feat: improve remux format detection

Author: bajankristof

lib/ffmpeg.rb

@@ -101,6 +101,7 @@ def ffmpeg_binary=(path)
 
       @ffmpeg_binary = path&.to_s
       @ffmpeg_version = nil
+      @muxers = nil
     end
 
     # Get the path to the ffmpeg binary.
@@ -132,6 +133,16 @@ def ffmpeg_version?(pattern)
       ffmpeg_version.start_with?(pattern.to_s)
     end
 
+    # Get the set of available muxer names for the ffmpeg binary.
+    #
+    # @return [Set<String>]
+    def muxers
+      @muxers ||= begin
+        stdout, = ffmpeg_capture3('-muxers', '-v', 'quiet')
+        stdout.scan(/^\s*E\S*\s+(\S+)/).flatten.to_set
+      end
+    end
+
     # Safely captures the standard output and the standard error of the ffmpeg command.
     #
     # @param args [Array<String>] The arguments to pass to ffmpeg.

lib/ffmpeg/media.rb

@@ -34,6 +34,9 @@ module FFMPEG
   #  media.load!
   #  media.video? # => true
   class Media
+    WEBM_CODEC_NAMES = Set.new(%w[vp8 vp9 av1 opus vorbis]).freeze
+    private_constant :WEBM_CODEC_NAMES
+
     # Raised if media metadata cannot be loaded.
     class LoadError < Error
       attr_reader :output
@@ -67,6 +70,41 @@ def initialize(message, output)
                          :format_name, :format_long_name,
                          :start_time, :bit_rate, :duration
 
+    # Returns the file extension that best represents this media's container format.
+    #
+    # @return [String] e.g. ".mp4", ".mkv", ".webm", ".ts", ".m3u8"
+    autoload def extname
+      case format_name
+      when /\Adash\b/   then '.mpd'
+      when /\bhls\b/    then '.m3u8'
+      when /\bmpegts\b/ then '.ts'
+      when /\b(mov|mp4)\b/
+        case major_brand
+        when /\Aqt\b/i  then '.mov'
+        when /\Am4a\b/i then '.m4a'
+        when /\Am4v\b/i then '.m4v'
+        when /\Am4s\b/i then '.m4s'
+        else                 '.mp4'
+        end
+      when /\bmatroska\b/
+        if streams
+           .select { _1.video? || _1.audio? }
+           .reject(&:attached_pic?)
+           .all? { WEBM_CODEC_NAMES.include?(_1.codec_name) }
+          '.webm'
+        else
+          '.mkv'
+        end
+      else
+        muxer =
+          format_name
+          .split(',')
+          .find { FFMPEG.muxers.include?(_1) }
+          .then { _1 || format_name.split(',').first }
+        ".#{muxer}"
+      end
+    end
+
     # @param path [String, Pathname, URI] The local path or remote URL to a multimedia file.
     # @param ffprobe_args [Array<String>] Additional arguments to pass to ffprobe.
     # @param load [Boolean] Whether to load the metadata immediately.
@@ -95,7 +133,7 @@ def remux(output_path = nil, timeout: nil, &block)
       raise ArgumentError if remote?
 
       Dir.mktmpdir do |tmpdir|
-        output_path = File.join(tmpdir, File.basename(@path))
+        output_path = File.join(tmpdir, "#{File.basename(@path, '.*')}#{extname}")
 
         status = Remuxer.new(timeout:).process(self, output_path, &block)
 

spec/ffmpeg/media_spec.rb

@@ -594,6 +594,62 @@ module FFMPEG
       end
     end
 
+    describe '#extname' do
+      context 'when the media is an MP4' do
+        it 'returns .mp4' do
+          expect(subject.extname).to eq('.mp4')
+        end
+      end
+
+      context 'when the media is a MOV (QuickTime)' do
+        let(:path) { fixture_media_file('rotated@0.mov') }
+
+        it 'returns .mov' do
+          expect(subject.extname).to eq('.mov')
+        end
+      end
+
+      context 'when the media is a WAV' do
+        let(:path) { fixture_media_file('hello.wav') }
+
+        it 'returns .wav' do
+          expect(subject.extname).to eq('.wav')
+        end
+      end
+
+      context 'when the media is an MP3' do
+        let(:path) { fixture_media_file('napoleon.mp3') }
+
+        it 'returns .mp3' do
+          expect(subject.extname).to eq('.mp3')
+        end
+      end
+
+      context 'when the media is a Matroska container with H.264 video named .webm' do
+        let(:path) { fixture_media_file('mkvh264.webm') }
+
+        it 'returns .mkv' do
+          expect(subject.extname).to eq('.mkv')
+        end
+      end
+
+      context 'when the media is a WebM container with VP9 video and Opus audio' do
+        let(:path) { fixture_media_file('landscape@smol.webm') }
+
+        it 'returns .webm' do
+          expect(subject.extname).to eq('.webm')
+        end
+      end
+
+      context 'when the media is a Matroska container with VP9 video and AAC audio named .webm' do
+        let(:path) { fixture_media_file('mkvaac.webm') }
+
+        it 'returns .mkv' do
+          expect(subject.extname).to eq('.mkv')
+        end
+      end
+    end
+
     describe '#remux' do
       context 'with an output_path' do
         let(:output_path) { tmp_file(ext: 'mp4') }
@@ -664,6 +720,23 @@ module FFMPEG
             end
           end
         end
+
+        context 'when the media is incorrectly named' do
+          let(:path) do
+            path = tmp_file(ext: 'webm')
+            FileUtils.cp(fixture_media_file('mkvh264.webm'), path)
+            path
+          end
+
+          it 'remuxes the file in place' do
+            status = subject.remux
+
+            expect(status).to be_a(Transcoder::Status)
+            expect(status.success?).to be(true)
+            expect(File.exist?(path)).to be(true)
+            expect(File.size(path)).to be > 0
+          end
+        end
       end
     end
 

spec/ffmpeg_spec.rb

@@ -55,6 +55,13 @@
       expect(described_class.instance_variable_get(:@ffmpeg_version)).to be_nil
     end
 
+    it 'clears the cached muxers' do
+      expect(File).to receive(:executable?).with('/path/to/ffmpeg').and_return(true)
+      described_class.instance_variable_set(:@muxers, Set.new(['mp4']))
+      described_class.ffmpeg_binary = '/path/to/ffmpeg'
+      expect(described_class.instance_variable_get(:@muxers)).to be_nil
+    end
+
     context 'when the assigned value is nil' do
       it 'clears the ffmpeg binary and version' do
         described_class.instance_variable_set(:@ffmpeg_binary, '/path/to/ffmpeg')
@@ -125,6 +132,22 @@
     end
   end
 
+  describe '.muxers' do
+    before { described_class.instance_variable_set(:@muxers, nil) }
+    after  { described_class.instance_variable_set(:@muxers, nil) }
+
+    it 'returns a set of available muxer names' do
+      expect(described_class.muxers).to be_a(Set)
+      expect(described_class.muxers).to include('mp4', 'matroska', 'webm')
+    end
+
+    it 'caches the result' do
+      expect(described_class).to receive(:ffmpeg_capture3).once.and_call_original
+      described_class.muxers
+      described_class.muxers
+    end
+  end
+
   describe '.ffmpeg_execute' do
     it 'returns the process status' do
       args = ['-i', fixture_media_file('hello.wav'), '-f', 'null', '-']