fix(parser): handle collab lockup metadata rows

spec/invidious/yt_backend/lockup_view_model_parser_spec.cr

@@ -0,0 +1,99 @@
+require "../../parsers_helper.cr"
+
+Spectator.describe "LockupViewModelParser channel videos" do
+  def lockup_video(metadata_rows : Array(JSON::Any)) : JSON::Any
+    JSON.parse({
+      "lockupViewModel" => {
+        "contentType"  => "LOCKUP_CONTENT_TYPE_VIDEO",
+        "contentId"    => "video123",
+        "contentImage" => {
+          "thumbnailViewModel" => {
+            "image" => {
+              "sources" => [{"url" => "https://example.invalid/thumb.jpg"}],
+            },
+            "overlays" => [{
+              "thumbnailBottomOverlayViewModel" => {
+                "badges" => [{
+                  "thumbnailBadgeViewModel" => {"text" => "12:34"},
+                }],
+              },
+            }],
+          },
+        },
+        "metadata" => {
+          "lockupMetadataViewModel" => {
+            "title"    => {"content" => "Collaboration video"},
+            "metadata" => {
+              "contentMetadataViewModel" => {
+                "metadataRows" => metadata_rows.map(&.raw),
+              },
+            },
+          },
+        },
+      },
+    }.to_json)
+  end
+
+  it "parses a normal first-row metadata layout" do
+    rows = [
+      JSON.parse({
+        "metadataParts" => [
+          {"text" => {"content" => "1.5M views"}},
+          {"text" => {"content" => "3 days ago"}},
+        ],
+      }.to_json),
+    ]
+
+    result = parse_item(lockup_video(rows), "Example", "UC123")
+    expect(result).to be_a(SearchVideo)
+
+    video = result.as(SearchVideo)
+    expect(video.views).to eq(1_500_000)
+    expect(video.published).to be_close(Time.utc - 3.days, 2.seconds)
+    expect(video.length_seconds).to eq(754)
+  end
+
+  it "parses metadata after a collaboration author row" do
+    rows = [
+      JSON.parse({
+        "metadataParts" => [
+          {"text" => {"content" => "Example and Collaborator"}},
+        ],
+      }.to_json),
+      JSON.parse({
+        "metadataParts" => [
+          {"text" => {"content" => "1.5M views"}},
+          {"text" => {"content" => "3 days ago"}},
+        ],
+      }.to_json),
+    ]
+
+    result = parse_item(lockup_video(rows), "Example", "UC123")
+    expect(result).to be_a(SearchVideo)
+
+    video = result.as(SearchVideo)
+    expect(video.views).to eq(1_500_000)
+    expect(video.published).to be_close(Time.utc - 3.days, 2.seconds)
+    expect(video.author).to eq("Example")
+    expect(video.ucid).to eq("UC123")
+  end
+
+  it "skips metadata rows that do not contain metadataParts" do
+    rows = [
+      JSON.parse("{}"),
+      JSON.parse({
+        "metadataParts" => [
+          {"text" => {"content" => "1.5M views"}},
+          {"text" => {"content" => "3 days ago"}},
+        ],
+      }.to_json),
+    ]
+
+    result = parse_item(lockup_video(rows), "Example", "UC123")
+    expect(result).to be_a(SearchVideo)
+
+    video = result.as(SearchVideo)
+    expect(video.views).to eq(1_500_000)
+    expect(video.published).to be_close(Time.utc - 3.days, 2.seconds)
+  end
+end

src/invidious/yt_backend/extractors.cr

@@ -662,7 +662,12 @@ private module Parsers
         metadata = item_contents.dig("metadata", "lockupMetadataViewModel")
         title = metadata.dig("title", "content").as_s
         # Contains the views of the video and the published time of the video.
-        metadata_parts = metadata.dig("metadata", "contentMetadataViewModel", "metadataRows", 0, "metadataParts").try &.as_a
+        # Collaboration videos can prepend author rows before the actual video metadata,
+        # so collect metadata parts from every row and skip rows without metadataParts.
+        metadata_rows = metadata.dig?("metadata", "contentMetadataViewModel", "metadataRows").try &.as_a
+        metadata_parts = metadata_rows.try &.compact_map { |row|
+          row["metadataParts"]?.try &.as_a
+        }.try &.flatten
 
         view_count_text = metadata_parts.try &.find { |item| item["icon"]?.nil? && item.dig?("text", "content").try &.as_s.includes?("views") }
           .try &.dig("text", "content").as_s