More improvements on WAV parser (#246)

* More improvements on WAV parser

* Calculate duration based on byte rate

* Adjust test

* Use sample_rate for determining duration

* Adjust test

* Bump version

CHANGELOG.md

@@ -1,3 +1,6 @@
+## 2.10.0
+* Improve WAV parser by focusing on performance rather than on attempting a best-effort when extracting metadata from files that do not strictly follow the format spec.
+
 ## 2.9.0
 * Improve WAV parser by performing a best-effort when extracting metadata from files that do not strictly follow the format spec.
 

lib/format_parser/version.rb

@@ -1,3 +1,3 @@
 module FormatParser
-  VERSION = '2.9.0'
+  VERSION = '2.10.0'
 end

lib/parsers/wav_parser.rb

@@ -21,30 +21,27 @@ def call(io)
     # The specification does not require the Format chunk to be the first chunk
     # after the RIFF header.
     # https://www.mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html
+    fmt_processed = false
+    data_processed = false
     fmt_data = {}
     data_size = 0
-    total_sample_frames = nil
     loop do
       chunk_type, chunk_size = safe_read(io, 8).unpack('a4l')
       case chunk_type
       when 'fmt ' # watch out: the chunk ID of the format chunk ends with a space
         fmt_data = unpack_fmt_chunk(io, chunk_size)
+        fmt_processed = true
       when 'data'
         data_size = chunk_size
-      when 'fact'
-        total_sample_frames = safe_read(io, 4).unpack('l').first
-        safe_skip(io, chunk_size - 4)
+        data_processed = true
       else
         # Skip this chunk until a known chunk is encountered
         safe_skip(io, chunk_size)
       end
-    rescue FormatParser::IOUtils::InvalidRead
-      # We've reached EOF, so it's time to make the most out of the metadata we
-      # managed to parse
-      break
+      break if fmt_processed && data_processed
     end
 
-    file_info(fmt_data, data_size, total_sample_frames)
+    file_info(fmt_data, data_size)
   end
 
   def unpack_fmt_chunk(io, chunk_size)
@@ -70,10 +67,10 @@ def unpack_fmt_chunk(io, chunk_size)
     }
   end
 
-  def file_info(fmt_data, data_size, sample_frames)
+  def file_info(fmt_data, data_size)
     # NOTE: Each sample includes information for each channel
-    sample_frames ||= data_size / (fmt_data[:channels] * fmt_data[:bits_per_sample] / 8) if fmt_data[:channels] > 0 && fmt_data[:bits_per_sample] > 0
-    duration_in_seconds = sample_frames / fmt_data[:sample_rate].to_f if fmt_data[:sample_rate] > 0
+    sample_frames = data_size / (fmt_data[:channels] * fmt_data[:bits_per_sample] / 8) if fmt_data[:channels] > 0 && fmt_data[:bits_per_sample] > 0
+    duration_in_seconds = sample_frames / fmt_data[:sample_rate].to_f if sample_frames && fmt_data[:byte_rate] > 0
     FormatParser::Audio.new(
       format: :wav,
       num_audio_channels: fmt_data[:channels],

spec/parsers/wav_parser_spec.rb

@@ -20,8 +20,9 @@
     expect(parse_result.format).to eq(:wav)
     expect(parse_result.num_audio_channels).to eq(1)
     expect(parse_result.audio_sample_rate_hz).to eq(8000)
-    expect(parse_result.media_duration_frames).to eq(110488)
-    expect(parse_result.media_duration_seconds).to be_within(0.01).of(13.81)
+    # Fixture does not define bits_per_sample in the fmt chunk
+    expect(parse_result.media_duration_frames).to be_nil
+    expect(parse_result.media_duration_seconds).to be_nil
   end
 
   it 'returns correct info about pcm files with more channels' do