refactor transcript serializer

Author: adrienpoly

Gemfile

@@ -115,6 +115,4 @@ gem "dry-types", "~> 1.7"
 
 gem "google-protobuf", require: false
 
-gem "webvtt-ruby"
-
 gem "active_job-performs", "~> 0.3.1"

Gemfile.lock

@@ -441,7 +441,6 @@ GEM
     websocket-driver (0.7.6)
       websocket-extensions (>= 0.1.0)
     websocket-extensions (0.1.5)
-    webvtt-ruby (0.4.2)
     xpath (3.2.0)
       nokogiri (~> 1.8)
     zeitwerk (2.6.16)
@@ -501,7 +500,6 @@ DEPENDENCIES
   vite_rails
   web-console
   webmock
-  webvtt-ruby
 
 RUBY VERSION
    ruby 3.3.1p55

app/clients/youtube/transcript.rb

@@ -4,7 +4,7 @@ module Youtube
   class Transcript
     attr_reader :response
 
-    def get_vtt(video_id)
+    def get(video_id)
       message = {one: "asr", two: "en"}
       typedef = MessageType
       two = get_base64_protobuf(message, typedef)
@@ -25,11 +25,11 @@ def get_vtt(video_id)
       }
 
       @response = HTTParty.post(url, headers: headers, body: body.to_json)
-      convert_to_vtt(JSON.parse(response.body))
+      JSON.parse(@response.body)
     end
 
-    def self.get_vtt(video_id)
-      new.get_vtt(video_id)
+    def self.get(video_id)
+      new.get(video_id)
     end
 
     private
@@ -43,32 +43,5 @@ def get_base64_protobuf(message, typedef)
       encoded_data = encode_message(message, typedef)
       Base64.encode64(encoded_data).delete("\n")
     end
-
-    def convert_to_vtt(transcript)
-      vtt_content = "WEBVTT\n\n"
-      events = transcript.dig("actions", 0, "updateEngagementPanelAction", "content", "transcriptRenderer", "content", "transcriptSearchPanelRenderer", "body", "transcriptSegmentListRenderer", "initialSegments")
-      if events
-        events.each_with_index do |event, index|
-          segment = event["transcriptSegmentRenderer"]
-          start_time = format_time(segment["startMs"].to_i)
-          end_time = format_time(segment["endMs"].to_i)
-          text = segment.dig("snippet", "runs")&.map { |run| run["text"] }&.join || ""
-          vtt_content += "#{index + 1}\n"
-          vtt_content += "#{start_time} --> #{end_time}\n"
-          vtt_content += "#{text}\n\n"
-        end
-      else
-        vtt_content += "NOTE No transcript data available\n"
-      end
-      vtt_content
-    end
-
-    def format_time(ms)
-      hours = ms / (1000 * 60 * 60)
-      minutes = (ms % (1000 * 60 * 60)) / (1000 * 60)
-      seconds = (ms % (1000 * 60)) / 1000
-      milliseconds = ms % 1000
-      format("%02d:%02d:%02d.%03d", hours, minutes, seconds, milliseconds)
-    end
   end
 end

app/models/cue.rb

@@ -0,0 +1,21 @@
+class Cue
+  attr_reader :start_time, :end_time, :text
+
+  def initialize(start_time, end_time, text)
+    @start_time = start_time
+    @end_time = end_time
+    @text = text
+  end
+
+  def to_s
+    "#{start_time} --> #{end_time}\n#{text}"
+  end
+
+  def to_h
+    {
+      start_time: start_time,
+      end_time: end_time,
+      text: text
+    }
+  end
+end

app/models/talk.rb

@@ -38,7 +38,7 @@ class Talk < ApplicationRecord
   has_many :speaker_talks, dependent: :destroy, inverse_of: :talk, foreign_key: :talk_id
   has_many :speakers, through: :speaker_talks
 
-  serialize :transcript, coder: WebVTTSerializer
+  serialize :transcript, coder: TranscriptSerializer
 
   # validations
   validates :title, presence: true
@@ -129,6 +129,7 @@ def related_talks(limit: 6)
   end
 
   def update_transcript!
-    update!(transcript: Youtube::Transcript.get_vtt(video_id))
+    youtube_transcript = Youtube::Transcript.get(video_id)
+    update!(transcript: Transcript.create_from_youtube_transcript(youtube_transcript))
   end
 end

app/models/transcript.rb

@@ -0,0 +1,61 @@
+class Transcript
+  include Enumerable
+
+  attr_reader :cues
+
+  def initialize
+    @cues = []
+  end
+
+  def add_cue(cue)
+    @cues << cue
+  end
+
+  def to_h
+    @cues.map { |cue| cue.to_h }
+  end
+
+  def to_json
+    to_h.to_json
+  end
+
+  def to_vtt
+    vtt_content = "WEBVTT\n\n"
+    @cues.each_with_index do |cue, index|
+      vtt_content += "#{index + 1}\n"
+      vtt_content += "#{cue}\n\n"
+    end
+    vtt_content
+  end
+
+  def each(&)
+    @cues.each(&)
+  end
+
+  class << self
+    def create_from_youtube_transcript(youtube_transcript)
+      transcript = Transcript.new
+      events = youtube_transcript.dig("actions", 0, "updateEngagementPanelAction", "content", "transcriptRenderer", "content", "transcriptSearchPanelRenderer", "body", "transcriptSegmentListRenderer", "initialSegments")
+      if events
+        events.each do |event|
+          segment = event["transcriptSegmentRenderer"]
+          start_time = format_time(segment["startMs"].to_i)
+          end_time = format_time(segment["endMs"].to_i)
+          text = segment.dig("snippet", "runs")&.map { |run| run["text"] }&.join || ""
+          transcript.add_cue(Cue.new(start_time, end_time, text))
+        end
+      else
+        transcript.add_cue(Cue.new("00:00:00.000", "00:00:00.000", "NOTE No transcript data available"))
+      end
+      transcript
+    end
+
+    def format_time(ms)
+      hours = ms / (1000 * 60 * 60)
+      minutes = (ms % (1000 * 60 * 60)) / (1000 * 60)
+      seconds = (ms % (1000 * 60)) / 1000
+      milliseconds = ms % 1000
+      format("%02d:%02d:%02d.%03d", hours, minutes, seconds, milliseconds)
+    end
+  end
+end

app/serializers/transcript_serializer.rb

@@ -0,0 +1,16 @@
+class TranscriptSerializer
+  def self.dump(transcript)
+    transcript.to_json
+  end
+
+  def self.load(transcript_json)
+    return nil if transcript_json.nil? || transcript_json.empty?
+
+    cues_array = JSON.parse(transcript_json, symbolize_names: true)
+    transcript = Transcript.new
+    cues_array.each do |cue_hash|
+      transcript.add_cue(Cue.new(cue_hash[:start_time], cue_hash[:end_time], cue_hash[:text]))
+    end
+    transcript
+  end
+end

app/serializers/webvtt_serializer.rb

@@ -12,5 +12,4 @@
 
 # These inflection rules are supported but not enabled by default:
 ActiveSupport::Inflector.inflections(:en) do |inflect|
-  inflect.acronym "WebVTT"
 end

config/initializers/inflections.rb

@@ -1,42 +1,20 @@
 require "test_helper"
-require "webvtt"
 
-module Youtube
-  class TranscriptTest < ActiveSupport::TestCase
-    def setup
-      @client = Youtube::Transcript.new
-    end
-
-    test "fetch the trasncript from a video in vtt format" do
-      video_id = "9LfmrkyP81M"
-
-      VCR.use_cassette("youtube_video_transcript", match_requests_on: [:method]) do
-        transcript = @client.get_vtt(video_id)
-        assert_not_nil transcript
-
-        # Save the VTT content to a temporary file to parse it using WebVTT gem
-        Tempfile.create(["transcript", ".vtt"]) do |file|
-          file.write(transcript)
-          file.rewind
-
-          # Parse the VTT file
-          webvtt = WebVTT.read(file.path)
+class Youtube::TranscriptTest < ActiveSupport::TestCase
+  def setup
+    @client = Youtube::Transcript.new
+  end
 
-          # Ensure it has the correct headers
-          assert_match(/^WEBVTT/, transcript)
+  test "fetch the trasncript from a video in vtt format" do
+    video_id = "9LfmrkyP81M"
 
-          # Ensure it has at least one cue
-          assert_not_empty webvtt.cues
+    VCR.use_cassette("youtube_video_transcript", match_requests_on: [:method]) do
+      transcript = @client.get(video_id)
+      assert_not_nil transcript
 
-          # Validate each cue
-          webvtt.cues.each do |cue|
-            assert_not_nil cue.start
-            assert_not_nil cue.end
-            assert_not_nil cue.text
-            assert_match(/^\d{2}:\d{2}:\d{2}\.\d{3} --> \d{2}:\d{2}:\d{2}\.\d{3}$/, "#{cue.start} --> #{cue.end}")
-          end
-        end
-      end
+      transcript = Transcript.create_from_youtube_transcript(transcript)
+      assert_not_empty transcript.cues
+      assert transcript.cues.first.is_a?(Cue)
     end
   end
 end

test/clients/youtube/transcript_test.rb

@@ -27,48 +27,13 @@
 
 class TalkTest < ActiveSupport::TestCase
   include ActiveJob::TestHelper
-  test "should serialize and deserialize transcript correctly" do
-    vtt_string = <<~VTT
-      WEBVTT
-
-      00:00.000 --> 00:05.000
-      Welcome to the talk.
-
-      00:06.000 --> 00:10.000
-      Let's get started.
-    VTT
-
-    talk = Talk.new(title: "Sample Talk", transcript: vtt_string)
-    assert talk.save
-
-    loaded_talk = Talk.find(talk.id)
-    assert_equal WebVTTSerializer.load(vtt_string), loaded_talk.transcript
-  end
-
-  test "should convert transcript to WebVTT format correctly" do
-    vtt_string = <<~VTT
-      WEBVTT
-
-      00:00.000 --> 00:05.000
-      Welcome to the talk.
-
-      00:06.000 --> 00:10.000
-      Let's get started.
-    VTT
-
-    cues = WebVTTSerializer.load(vtt_string)
-    talk = Talk.new(title: "Sample Talk", transcript: cues)
-
-    expected_vtt = WebVTTSerializer.dump(talk.transcript)
-    assert_equal vtt_string.strip, expected_vtt.strip
-  end
-
   test "should handle empty transcript" do
-    talk = Talk.new(title: "Sample Talk", transcript: [])
+    talk = Talk.new(title: "Sample Talk", transcript: Transcript.new)
     assert talk.save
 
     loaded_talk = Talk.find(talk.id)
-    assert_empty loaded_talk.transcript
+    assert_equal loaded_talk.transcript.cues, []
+    assert_equal "Sample Talk", loaded_talk.title
   end
 
   test "should update transcript" do
@@ -80,7 +45,8 @@ class TalkTest < ActiveSupport::TestCase
       end
     end
 
-    assert_not_empty @talk.transcript
-    assert @talk.transcript.length > 100
+    assert @talk.transcript.is_a?(Transcript)
+    assert @talk.transcript.cues.first.is_a?(Cue)
+    assert @talk.transcript.cues.length > 100
   end
 end