Fix TTML ruby decoding to resolve styles by ID

The current code only works if the tts:ruby attributes are defined
directly on the in-line elements. This changes that so we also consider
tts:ruby attributes on `style` nodes referenced by ID.
PiperOrigin-RevId: 319515177

library/core/src/main/java/com/google/android/exoplayer2/text/ttml/TtmlNode.java

@@ -379,7 +379,7 @@ import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
       regionOutput.setText(text);
     }
     if (resolvedStyle != null) {
-      TtmlRenderUtil.applyStylesToSpan(text, start, end, resolvedStyle, parent);
+      TtmlRenderUtil.applyStylesToSpan(text, start, end, resolvedStyle, parent, globalStyles);
       regionOutput
           .setTextAlignment(resolvedStyle.getTextAlign())
           .setVerticalType(resolvedStyle.getVerticalType());

library/core/src/main/java/com/google/android/exoplayer2/text/ttml/TtmlRenderUtil.java

@@ -78,7 +78,12 @@ import java.util.Map;
   }
 
   public static void applyStylesToSpan(
-      Spannable builder, int start, int end, TtmlStyle style, @Nullable TtmlNode parent) {
+      Spannable builder,
+      int start,
+      int end,
+      TtmlStyle style,
+      @Nullable TtmlNode parent,
+      Map<String, TtmlStyle> globalStyles) {
 
     if (style.getStyle() != TtmlStyle.UNSPECIFIED) {
       builder.setSpan(new StyleSpan(style.getStyle()), start, end,
@@ -117,12 +122,12 @@ import java.util.Map;
     switch (style.getRubyType()) {
       case TtmlStyle.RUBY_TYPE_BASE:
         // look for the sibling RUBY_TEXT and add it as span between start & end.
-        @Nullable TtmlNode containerNode = findRubyContainerNode(parent);
+        @Nullable TtmlNode containerNode = findRubyContainerNode(parent, globalStyles);
         if (containerNode == null) {
           // No matching container node
           break;
         }
-        @Nullable TtmlNode textNode = findRubyTextNode(containerNode);
+        @Nullable TtmlNode textNode = findRubyTextNode(containerNode, globalStyles);
         if (textNode == null) {
           // no matching text node
           break;
@@ -200,12 +205,15 @@ import java.util.Map;
   }
 
   @Nullable
-  private static TtmlNode findRubyTextNode(TtmlNode rubyContainerNode) {
+  private static TtmlNode findRubyTextNode(
+      TtmlNode rubyContainerNode, Map<String, TtmlStyle> globalStyles) {
     Deque<TtmlNode> childNodesStack = new ArrayDeque<>();
     childNodesStack.push(rubyContainerNode);
     while (!childNodesStack.isEmpty()) {
       TtmlNode childNode = childNodesStack.pop();
-      if (childNode.style != null && childNode.style.getRubyType() == TtmlStyle.RUBY_TYPE_TEXT) {
+      @Nullable
+      TtmlStyle style = resolveStyle(childNode.style, childNode.getStyleIds(), globalStyles);
+      if (style != null && style.getRubyType() == TtmlStyle.RUBY_TYPE_TEXT) {
         return childNode;
       }
       for (int i = childNode.getChildCount() - 1; i >= 0; i--) {
@@ -217,9 +225,10 @@ import java.util.Map;
   }
 
   @Nullable
-  private static TtmlNode findRubyContainerNode(@Nullable TtmlNode node) {
+  private static TtmlNode findRubyContainerNode(
+      @Nullable TtmlNode node, Map<String, TtmlStyle> globalStyles) {
     while (node != null) {
-      @Nullable TtmlStyle style = node.style;
+      @Nullable TtmlStyle style = resolveStyle(node.style, node.getStyleIds(), globalStyles);
       if (style != null && style.getRubyType() == TtmlStyle.RUBY_TYPE_CONTAINER) {
         return node;
       }

library/core/src/test/java/com/google/android/exoplayer2/text/ttml/TtmlDecoderTest.java

@@ -656,15 +656,19 @@ public final class TtmlDecoderTest {
 
     Spanned thirdCue = getOnlyCueTextAtTimeUs(subtitle, 30_000_000);
     assertThat(thirdCue.toString()).isEqualTo("Cue with annotated text.");
-    assertThat(thirdCue).hasNoRubySpanBetween(0, thirdCue.length());
+    assertThat(thirdCue).hasRubySpanBetween("Cue with ".length(), "Cue with annotated".length());
 
     Spanned fourthCue = getOnlyCueTextAtTimeUs(subtitle, 40_000_000);
-    assertThat(fourthCue.toString()).isEqualTo("Cue with text.");
+    assertThat(fourthCue.toString()).isEqualTo("Cue with annotated text.");
     assertThat(fourthCue).hasNoRubySpanBetween(0, fourthCue.length());
 
     Spanned fifthCue = getOnlyCueTextAtTimeUs(subtitle, 50_000_000);
-    assertThat(fifthCue.toString()).isEqualTo("Cue with annotated text.");
+    assertThat(fifthCue.toString()).isEqualTo("Cue with text.");
     assertThat(fifthCue).hasNoRubySpanBetween(0, fifthCue.length());
+
+    Spanned sixthCue = getOnlyCueTextAtTimeUs(subtitle, 60_000_000);
+    assertThat(sixthCue.toString()).isEqualTo("Cue with annotated text.");
+    assertThat(sixthCue).hasNoRubySpanBetween(0, sixthCue.length());
   }
 
   private static Spanned getOnlyCueTextAtTimeUs(Subtitle subtitle, long timeUs) {

testdata/src/test/assets/ttml/rubies.xml

@@ -16,9 +16,15 @@
   -->
 <tt xmlns:ttm="http://www.w3.org/2006/10/ttaf1#metadata"
   xmlns:ttp="http://www.w3.org/2006/10/ttaf1#parameter"
-  xmlns:tts="http://www.w3.org/2006/10/ttaf1#style"
+  xmlns:tts="http://www.w3.org/2006/10/ttaf1#style" xmlns="http://www.w3.org/ns/ttml"
-  xmlns="http://www.w3.org/ns/ttml"
   xmlns="http://www.w3.org/2006/10/ttaf1">
+  <head>
+    <styling>
+      <style id="cont" tts:ruby="container" />
+      <style id="base" tts:ruby="base" />
+      <style id="text" tts:ruby="text" />
+    </styling>
+  </head>
   <body>
     <div>
       <!-- Base before and after text, one with explicit position -->
@@ -47,17 +53,29 @@
       </p>
     </div>
     <div>
-      <!-- No text section -> no span -->
+      <!-- ruby info in style block -->
       <p begin="30s" end="38s">
+        Cue with
+        <span style="cont">
+          <span style="base">annotated</span>
+          <span style="text">rubies</span>
+          text.
+        </span>
+      </p>
+    </div>
+    <div>
+      <!-- No text section -> no span -->
+      <p begin="40s" end="48s">
         Cue with
         <span tts:ruby="container" tts:rubyPosition="before">
           <span tts:ruby="base">annotated</span>
         </span>
-        text.</p>
+        text.
+      </p>
     </div>
     <div>
       <!-- No base section -> text still stripped-->
-      <p begin="40s" end="48s">
+      <p begin="50s" end="58s">
         Cue with
         <span tts:ruby="container" tts:rubyPosition="before">
           <span tts:ruby="text">rubies</span>
@@ -67,7 +85,7 @@
     </div>
     <div>
       <!-- No container section -> text still stripped-->
-      <p begin="50s" end="58s">
+      <p begin="60s" end="68s">
         Cue with
         <span tts:ruby="text">rubies</span>
         <span tts:ruby="base">annotated</span>