Add support for TTML origin and extent attributes defined in style instead of region

RELEASENOTES.md

@@ -24,6 +24,8 @@
         `DefaultAudioSink` to stall.
 *   Video:
 *   Text:
+    *   TTML: Add support for referencing `tts:origin` and `tts:extent` via
+        `style` ([#2953](https://github.com/google/ExoPlayer/issues/2953)).
 *   Metadata:
 *   Image:
 *   DataSource:

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/text/DelegatingSubtitleDecoderTtmlParserTest.java

@@ -308,13 +308,9 @@ public void multipleRegions() throws IOException {
 
     cue = getOnlyCueAtTimeUs(subtitle, 9_000_000);
     assertThat(cue.text.toString()).isEqualTo("dolor");
-    assertThat(cue.position).isEqualTo(Cue.DIMEN_UNSET);
-    assertThat(cue.line).isEqualTo(Cue.DIMEN_UNSET);
-    assertThat(cue.size).isEqualTo(Cue.DIMEN_UNSET);
-    // TODO: Should be as below, once https://github.com/google/ExoPlayer/issues/2953 is fixed.
-    // assertEquals(10f / 100f, cue.position);
-    // assertEquals(80f / 100f, cue.line);
-    // assertEquals(1f, cue.size);
+    assertThat(cue.position).isEqualTo(10f / 100f);
+    assertThat(cue.line).isEqualTo(80f / 100f);
+    assertThat(cue.size).isEqualTo(1f);
 
     cue = getOnlyCueAtTimeUs(subtitle, 21_000_000);
     assertThat(cue.text.toString()).isEqualTo("They first said this");

libraries/extractor/src/main/java/androidx/media3/extractor/text/ttml/TtmlNode.java

@@ -53,6 +53,7 @@
 
   public static final String ANONYMOUS_REGION_ID = "";
   public static final String ATTR_ID = "id";
+  public static final String ATTR_STYLE = "style";
   public static final String ATTR_TTS_ORIGIN = "origin";
   public static final String ATTR_TTS_EXTENT = "extent";
   public static final String ATTR_TTS_DISPLAY_ALIGN = "displayAlign";

libraries/extractor/src/main/java/androidx/media3/extractor/text/ttml/TtmlParser.java

@@ -315,7 +315,8 @@ private static Map<String, TtmlStyle> parseHeader(
           globalStyles.put(styleId, style);
         }
       } else if (XmlPullParserUtil.isStartTag(xmlParser, TtmlNode.TAG_REGION)) {
-        @Nullable TtmlRegion ttmlRegion = parseRegionAttributes(xmlParser, cellRows, ttsExtent);
+        @Nullable
+        TtmlRegion ttmlRegion = parseRegionAttributes(xmlParser, cellRows, ttsExtent, globalStyles);
         if (ttmlRegion != null) {
           globalRegions.put(ttmlRegion.id, ttmlRegion);
         }
@@ -350,7 +351,10 @@ private static void parseMetadata(XmlPullParser xmlParser, Map<String, String> i
    */
   @Nullable
   private static TtmlRegion parseRegionAttributes(
-      XmlPullParser xmlParser, int cellRows, @Nullable TtsExtent ttsExtent) {
+      XmlPullParser xmlParser,
+      int cellRows,
+      @Nullable TtsExtent ttsExtent,
+      Map<String, TtmlStyle> globalStyles) {
     @Nullable String regionId = XmlPullParserUtil.getAttributeValue(xmlParser, TtmlNode.ATTR_ID);
     if (regionId == null) {
       return null;
@@ -361,6 +365,15 @@ private static TtmlRegion parseRegionAttributes(
 
     @Nullable
     String regionOrigin = XmlPullParserUtil.getAttributeValue(xmlParser, TtmlNode.ATTR_TTS_ORIGIN);
+    if (regionOrigin == null) {
+      String styleId = XmlPullParserUtil.getAttributeValue(xmlParser, TtmlNode.ATTR_STYLE);
+      if (styleId != null) {
+        TtmlStyle style = globalStyles.get(styleId);
+        if (style != null) {
+          regionOrigin = style.getOrigin();
+        }
+      }
+    }
     if (regionOrigin != null) {
       Matcher originPercentageMatcher = PERCENTAGE_COORDINATES.matcher(regionOrigin);
       Matcher originPixelMatcher = PIXEL_COORDINATES.matcher(regionOrigin);
@@ -393,19 +406,24 @@ private static TtmlRegion parseRegionAttributes(
         return null;
       }
     } else {
-      Log.w(TAG, "Ignoring region without an origin");
-      return null;
-      // TODO: Should default to top left as below in this case, but need to fix
-      // https://github.com/google/ExoPlayer/issues/2953 first.
       // Origin is omitted. Default to top left.
-      // position = 0;
-      // line = 0;
+       position = 0;
+       line = 0;
     }
 
     float width;
     float height;
     @Nullable
     String regionExtent = XmlPullParserUtil.getAttributeValue(xmlParser, TtmlNode.ATTR_TTS_EXTENT);
+    if (regionExtent == null) {
+      String styleId = XmlPullParserUtil.getAttributeValue(xmlParser, TtmlNode.ATTR_STYLE);
+      if (styleId != null) {
+        TtmlStyle style = globalStyles.get(styleId);
+        if (style != null) {
+          regionExtent = style.getExtent();
+        }
+      }
+    }
     if (regionExtent != null) {
       Matcher extentPercentageMatcher = PERCENTAGE_COORDINATES.matcher(regionExtent);
       Matcher extentPixelMatcher = PIXEL_COORDINATES.matcher(regionExtent);
@@ -439,13 +457,9 @@ private static TtmlRegion parseRegionAttributes(
         return null;
       }
     } else {
-      Log.w(TAG, "Ignoring region without an extent");
-      return null;
-      // TODO: Should default to extent of parent as below in this case, but need to fix
-      // https://github.com/google/ExoPlayer/issues/2953 first.
       // Extent is omitted. Default to extent of parent.
-      // width = 1;
-      // height = 1;
+       width = 1;
+       height = 1;
     }
 
     @Cue.AnchorType int lineAnchor = Cue.ANCHOR_TYPE_START;
@@ -626,6 +640,12 @@ private static String[] parseStyleIds(String parentStyleIds) {
         case TtmlNode.ATTR_TTS_SHEAR:
           style = createIfNull(style).setShearPercentage(parseShear(attributeValue));
           break;
+        case TtmlNode.ATTR_TTS_ORIGIN:
+          style = createIfNull(style).setOrigin(attributeValue);
+          break;
+        case TtmlNode.ATTR_TTS_EXTENT:
+          style = createIfNull(style).setExtent(attributeValue);
+          break;
         default:
           // ignore
           break;

libraries/extractor/src/main/java/androidx/media3/extractor/text/ttml/TtmlStyle.java

@@ -96,6 +96,8 @@
   private @OptionalBoolean int textCombine;
   @Nullable private TextEmphasis textEmphasis;
   private float shearPercentage;
+  @Nullable private String origin;
+  @Nullable private String extent;
 
   public TtmlStyle() {
     linethrough = UNSPECIFIED;
@@ -277,6 +279,12 @@ private TtmlStyle inherit(@Nullable TtmlStyle ancestor, boolean chaining) {
       if (shearPercentage == UNSPECIFIED_SHEAR) {
         shearPercentage = ancestor.shearPercentage;
       }
+      if (origin == null) {
+        origin = ancestor.origin;
+      }
+      if (extent == null) {
+        extent = ancestor.extent;
+      }
       // attributes not inherited as of http://www.w3.org/TR/ttml1/
       if (chaining && !hasBackgroundColor && ancestor.hasBackgroundColor) {
         setBackgroundColor(ancestor.backgroundColor);
@@ -382,4 +390,26 @@ public TtmlStyle setFontSizeUnit(int fontSizeUnit) {
   public float getFontSize() {
     return fontSize;
   }
+
+  @CanIgnoreReturnValue
+  public TtmlStyle setOrigin(@Nullable String origin) {
+    this.origin = origin;
+    return this;
+  }
+
+  @Nullable
+  public String getOrigin() {
+    return origin;
+  }
+
+  @CanIgnoreReturnValue
+  public TtmlStyle setExtent(@Nullable String extent) {
+    this.extent = extent;
+    return this;
+  }
+
+  @Nullable
+  public String getExtent() {
+    return extent;
+  }
 }

libraries/extractor/src/test/java/androidx/media3/extractor/text/ttml/TtmlParserTest.java

@@ -74,6 +74,8 @@ public final class TtmlParserTest {
   private static final String RUBIES_FILE = "media/ttml/rubies.xml";
   private static final String TEXT_EMPHASIS_FILE = "media/ttml/text_emphasis.xml";
   private static final String SHEAR_FILE = "media/ttml/shear.xml";
+  private static final String REGION_ATTRS_FROM_STYLE_FILE =
+      "media/ttml/inherit_region_attributes_from_style.xml";
 
   @Test
   public void simple_allCues() throws Exception {
@@ -485,13 +487,9 @@ public void multipleRegions() throws Exception {
 
     cue = Iterables.getOnlyElement(allCues.get(2).cues);
     assertThat(cue.text.toString()).isEqualTo("dolor");
-    assertThat(cue.position).isEqualTo(Cue.DIMEN_UNSET);
-    assertThat(cue.line).isEqualTo(Cue.DIMEN_UNSET);
-    assertThat(cue.size).isEqualTo(Cue.DIMEN_UNSET);
-    // TODO: Should be as below, once https://github.com/google/ExoPlayer/issues/2953 is fixed.
-    // assertEquals(10f / 100f, cue.position);
-    // assertEquals(80f / 100f, cue.line);
-    // assertEquals(1f, cue.size);
+    assertThat(cue.position).isEqualTo(10f / 100f);
+    assertThat(cue.line).isEqualTo(80f / 100f);
+    assertThat(cue.size).isEqualTo(1f);
 
     cue = Iterables.getOnlyElement(allCues.get(3).cues);
     assertThat(cue.text.toString()).isEqualTo("They first said this");
@@ -1093,6 +1091,26 @@ public void shear() throws Exception {
     assertThat(eighthCue.shearDegrees).isWithin(0.01f).of(90f);
   }
 
+  @Test
+  public void regionAttrsFromStyle() throws Exception {
+    ImmutableList<CuesWithTiming> allCues = getAllCues(REGION_ATTRS_FROM_STYLE_FILE);
+
+    Cue firstCue = Iterables.getOnlyElement(allCues.get(0).cues);
+    assertThat(firstCue.position).isEqualTo(10f / 100f);
+    assertThat(firstCue.line).isEqualTo(10f / 100f);
+    assertThat(firstCue.size).isEqualTo(20f / 100f);
+
+    Cue secondCue = Iterables.getOnlyElement(allCues.get(1).cues);
+    assertThat(secondCue.position).isEqualTo(30f / 100f);
+    assertThat(secondCue.line).isEqualTo(30f / 100f);
+    assertThat(secondCue.size).isEqualTo(40f / 100f);
+
+    Cue thirdCue = Iterables.getOnlyElement(allCues.get(2).cues);
+    assertThat(thirdCue.position).isEqualTo(30f / 100f);
+    assertThat(thirdCue.line).isEqualTo(30f / 100f);
+    assertThat(thirdCue.size).isEqualTo(20f / 100f);
+  }
+
   private static Spanned getOnlyCueTextAtIndex(List<CuesWithTiming> allCues, int index) {
     Cue cue = getOnlyCueAtIndex(allCues, index);
     assertThat(cue.text).isInstanceOf(Spanned.class);

libraries/test_data/src/test/assets/media/ttml/inherit_region_attributes_from_style.xml

@@ -0,0 +1,22 @@
+<tt xmlns="http://www.w3.org/ns/ttml"
+  xmlns:tts="http://www.w3.org/2006/10/ttaf1#style">
+  <head>
+    <styling>
+      <style xml:id="style1" tts:extent="20% 20%" />
+      <style xml:id="style2" tts:origin="30% 30%" />
+      <style xml:id="style3" style="style1 style2" />
+    </styling>
+    <layout>
+      <region xml:id="region1" style="style1" tts:origin="10% 10%"  />
+      <region xml:id="region2" style="style2" tts:extent="40% 40%" />
+      <region xml:id="region3" style="style3" />
+    </layout>
+  </head>
+  <body>
+      <div>
+        <p begin="2s" end="4s" region="region1">text 1</p>
+        <p begin="4s" end="6s" region="region2">text 2</p>
+        <p begin="8s" end="10s" region="region3">text 3</p>
+      </div>
+  </body>
+</tt>