fix(TTML): Fix support of urls in smpte:backgroundImage

externs/shaka/text.js

@@ -36,11 +36,13 @@ shaka.extern.TextParser = class {
    * @param {shaka.extern.TextParser.TimeContext} timeContext
    *    The time information that should be used to adjust the times values
    *    for each cue.
+   * @param {?string} uri
+   *    The media uri.
    * @return {!Array.<!shaka.text.Cue>}
    *
    * @exportDoc
    */
-  parseMedia(data, timeContext) {}
+  parseMedia(data, timeContext, uri) {}
 
   /**
    * Notifies the stream if the manifest is in sequence mode or not.

lib/media/media_source_engine.js

@@ -830,7 +830,8 @@ shaka.media.MediaSourceEngine = class {
       await this.textEngine_.appendBuffer(
           data,
           reference ? reference.startTime : null,
-          reference ? reference.endTime : null);
+          reference ? reference.endTime : null,
+          reference ? reference.getUris()[0] : null);
       return;
     }
 

lib/player.js

@@ -4705,7 +4705,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
       vttOffset: 0,
     };
     const data = shaka.util.BufferUtils.toUint8(buffer);
-    const cues = TextParser.parseMedia(data, time);
+    const cues = TextParser.parseMedia(data, time, uri);
 
     const references = [];
     for (const cue of cues) {
@@ -4968,7 +4968,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
         vttOffset: 0,
       };
       const data = shaka.util.BufferUtils.toUint8(buffer);
-      const cues = obj.parseMedia(data, time);
+      const cues = obj.parseMedia(data, time, /* uri= */ null);
       return shaka.text.WebVttGenerator.convert(cues, adCuePoints);
     }
     throw new shaka.util.Error(

lib/text/mp4_ttml_parser.js

@@ -75,7 +75,7 @@ shaka.text.Mp4TtmlParser = class {
    * @override
    * @export
    */
-  parseMedia(data, time) {
+  parseMedia(data, time, uri) {
     const Mp4Parser = shaka.util.Mp4Parser;
 
     let sawMDAT = false;
@@ -86,7 +86,7 @@ shaka.text.Mp4TtmlParser = class {
           sawMDAT = true;
           // Join this to any previous payload, in case the mp4 has multiple
           // mdats.
-          payload = payload.concat(this.parser_.parseMedia(data, time));
+          payload = payload.concat(this.parser_.parseMedia(data, time, uri));
         }));
     parser.parse(data, /* partialOkay= */ false);
 

lib/text/srt_text_parser.js

@@ -55,7 +55,7 @@ shaka.text.SrtTextParser = class {
    * @override
    * @export
    */
-  parseMedia(data, time) {
+  parseMedia(data, time, uri) {
     const SrtTextParser = shaka.text.SrtTextParser;
     const BufferUtils = shaka.util.BufferUtils;
     const StringUtils = shaka.util.StringUtils;
@@ -67,7 +67,7 @@ shaka.text.SrtTextParser = class {
 
     const newData = BufferUtils.toUint8(StringUtils.toUTF8(vvtText));
 
-    return this.parser_.parseMedia(newData, time);
+    return this.parser_.parseMedia(newData, time, uri);
   }
 
   /**

lib/text/text_engine.js

@@ -166,9 +166,10 @@ shaka.text.TextEngine = class {
    * @param {BufferSource} buffer
    * @param {?number} startTime relative to the start of the presentation
    * @param {?number} endTime relative to the start of the presentation
+   * @param {?string=} uri
    * @return {!Promise}
    */
-  async appendBuffer(buffer, startTime, endTime) {
+  async appendBuffer(buffer, startTime, endTime, uri) {
     goog.asserts.assert(
         this.parser_, 'The parser should already be initialized');
 
@@ -196,9 +197,14 @@ shaka.text.TextEngine = class {
       vttOffset: vttOffset,
     };
 
+    let finalUri = null;
+    if (uri) {
+      finalUri = uri;
+    }
+
     // Parse the buffer and add the new cues.
     const allCues = this.parser_.parseMedia(
-        shaka.util.BufferUtils.toUint8(buffer), time);
+        shaka.util.BufferUtils.toUint8(buffer), time, finalUri);
     const cuesToAppend = allCues.filter((cue) => {
       return cue.startTime >= this.appendWindowStart_ &&
           cue.startTime < this.appendWindowEnd_;

lib/text/ttml_text_parser.js

@@ -7,6 +7,7 @@
 goog.provide('shaka.text.TtmlTextParser');
 
 goog.require('goog.asserts');
+goog.require('goog.Uri');
 goog.require('shaka.log');
 goog.require('shaka.text.Cue');
 goog.require('shaka.text.CueRegion');
@@ -50,7 +51,7 @@ shaka.text.TtmlTextParser = class {
    * @override
    * @export
    */
-  parseMedia(data, time) {
+  parseMedia(data, time, uri) {
     const TtmlTextParser = shaka.text.TtmlTextParser;
     const XmlUtils = shaka.util.XmlUtils;
     const ttpNs = TtmlTextParser.parameterNs_;
@@ -147,7 +148,7 @@ shaka.text.TtmlTextParser = class {
         body, time, rateInfo, metadataElements, styles,
         regionElements, cueRegions, whitespaceTrim,
         cellResolutionInfo, /* parentCueElement= */ null,
-        /* isContent= */ false);
+        /* isContent= */ false, uri);
     if (cue) {
       // According to the TTML spec, backgrounds default to transparent.
       // So default the background of the top-level element to transparent.
@@ -175,12 +176,14 @@ shaka.text.TtmlTextParser = class {
    * @param {?{columns: number, rows: number}} cellResolution
    * @param {?Element} parentCueElement
    * @param {boolean} isContent
+   * @param {?string} uri
    * @return {shaka.text.Cue}
    * @private
    */
   static parseCue_(
       cueNode, timeContext, rateInfo, metadataElements, styles, regionElements,
-      cueRegions, whitespaceTrim, cellResolution, parentCueElement, isContent) {
+      cueRegions, whitespaceTrim, cellResolution, parentCueElement, isContent,
+      uri) {
     /** @type {Element} */
     let cueElement;
     /** @type {Element} */
@@ -222,7 +225,21 @@ shaka.text.TtmlTextParser = class {
       }
     }
 
-    if (cueNode.nodeName == 'p' || imageElement) {
+    let imageUri = null;
+    const backgroundImage = shaka.util.XmlUtils.getAttributeNSList(
+        cueElement,
+        shaka.text.TtmlTextParser.smpteNsList_,
+        'backgroundImage');
+    if (uri && backgroundImage && !backgroundImage.startsWith('#')) {
+      const baseUri = new goog.Uri(uri);
+      const relativeUri = new goog.Uri(backgroundImage);
+      const newUri = baseUri.resolve(relativeUri).toString();
+      if (newUri) {
+        imageUri = newUri;
+      }
+    }
+
+    if (cueNode.nodeName == 'p' || imageElement || imageUri) {
       isContent = true;
     }
 
@@ -255,6 +272,7 @@ shaka.text.TtmlTextParser = class {
             cellResolution,
             cueElement,
             isContent,
+            uri,
         );
 
         // This node may or may not generate a nested cue.
@@ -388,6 +406,7 @@ shaka.text.TtmlTextParser = class {
         cueElement,
         regionElementForStyle,
         imageElement,
+        imageUri,
         styles,
         /** isNested= */ parentIsContent, // "nested in a <div>" doesn't count.
         /** isLeaf= */ (nestedCues.length == 0));
@@ -490,13 +509,15 @@ shaka.text.TtmlTextParser = class {
    * @param {!Element} cueElement
    * @param {Element} region
    * @param {Element} imageElement
+   * @param {?string} imageUri
    * @param {!Array.<!Element>} styles
    * @param {boolean} isNested
    * @param {boolean} isLeaf
    * @private
    */
   static addStyle_(
-      cue, cueElement, region, imageElement, styles, isNested, isLeaf) {
+      cue, cueElement, region, imageElement, imageUri, styles,
+      isNested, isLeaf) {
     const TtmlTextParser = shaka.text.TtmlTextParser;
     const Cue = shaka.text.Cue;
 
@@ -664,6 +685,8 @@ shaka.text.TtmlTextParser = class {
           backgroundImageData) {
         cue.backgroundImage = 'data:image/png;base64,' + backgroundImageData;
       }
+    } else if (imageUri) {
+      cue.backgroundImage = imageUri;
     }
 
     const textOutline = TtmlTextParser.getStyleAttribute_(

test/media/media_source_engine_unit.js

@@ -631,7 +631,7 @@ describe('MediaSourceEngine', () => {
           ContentType.TEXT, data, reference, fakeStream,
           /* hasClosedCaptions= */ false);
       expect(mockTextEngine.appendBuffer).toHaveBeenCalledWith(
-          data, 0, 10);
+          data, 0, 10, 'foo://bar');
     });
 
     it('appends transmuxed data', async () => {

test/text/mp4_ttml_parser_unit.js

@@ -43,7 +43,7 @@ describe('Mp4TtmlParser', () => {
     parser.parseInit(ttmlInitSegment);
     const time =
         {periodStart: 0, segmentStart: 0, segmentEnd: 60, vttOffset: 0};
-    const ret = parser.parseMedia(ttmlSegmentMultipleMDAT, time);
+    const ret = parser.parseMedia(ttmlSegmentMultipleMDAT, time, null);
     // Bodies.
     expect(ret.length).toBe(2);
     // Divs.
@@ -63,10 +63,10 @@ describe('Mp4TtmlParser', () => {
     const parser = new shaka.text.Mp4TtmlParser();
     parser.parseInit(ttmlInitSegment);
 
-    const ret1 = parser.parseMedia(ttmlSegment, time1);
+    const ret1 = parser.parseMedia(ttmlSegment, time1, null);
     expect(ret1.length).toBeGreaterThan(0);
 
-    const ret2 = parser.parseMedia(ttmlSegment, time2);
+    const ret2 = parser.parseMedia(ttmlSegment, time2, null);
     expect(ret2.length).toBeGreaterThan(0);
 
     expect(ret2[0].startTime).toBe(ret1[0].startTime + 7);
@@ -164,7 +164,7 @@ describe('Mp4TtmlParser', () => {
     parser.parseInit(ttmlInitSegment);
     const time =
         {periodStart: 0, segmentStart: 0, segmentEnd: 60, vttOffset: 0};
-    const result = parser.parseMedia(ttmlSegment, time);
+    const result = parser.parseMedia(ttmlSegment, time, null);
     shaka.test.TtmlUtils.verifyHelper(
         cues, result, {startTime: 23, endTime: 53.5});
   });

test/text/srt_text_parser_unit.js

@@ -149,7 +149,7 @@ describe('SrtTextParser', () => {
     const data = BufferUtils.toUint8(StringUtils.toUTF8(text));
 
     const parser = new shaka.text.SrtTextParser();
-    const result = parser.parseMedia(data, time);
+    const result = parser.parseMedia(data, time, null);
 
     const expected = cues.map((cue) => {
       if (cue.nestedCues) {

test/text/text_engine_unit.js

@@ -105,6 +105,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 0, segmentStart: 0, segmentEnd: 3, vttOffset: 0},
+        null,
       ]);
 
       expect(mockDisplayer.appendSpy).toHaveBeenCalledOnceMoreWith([
@@ -120,6 +121,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 0, segmentStart: 3, segmentEnd: 5, vttOffset: 0},
+        null,
       ]);
 
       expect(mockDisplayer.appendSpy).toHaveBeenCalledOnceMoreWith([
@@ -272,6 +274,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 0, segmentStart: 0, segmentEnd: 3, vttOffset: 0},
+        null,
       ]);
       expect(mockDisplayer.appendSpy).toHaveBeenCalledOnceMoreWith([
         [
@@ -286,6 +289,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 4, segmentStart: 4, segmentEnd: 7, vttOffset: 4},
+        null,
       ]);
       expect(mockDisplayer.appendSpy).toHaveBeenCalledOnceMoreWith([
         [
@@ -316,6 +320,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 0, segmentStart: 0, segmentEnd: 3, vttOffset: 0},
+        null,
       ]);
 
       textEngine.setTimestampOffset(8);
@@ -325,6 +330,7 @@ describe('TextEngine', () => {
       expect(mockParseMedia).toHaveBeenCalledOnceMoreWith([
         dummyData,
         {periodStart: 8, segmentStart: 4, segmentEnd: 7, vttOffset: 4},
+        null,
       ]);
     });
   });

test/text/ttml_text_parser_unit.js

@@ -1263,6 +1263,35 @@ describe('TtmlTextParser', () => {
         });
   });
 
+  it('supports smpte:backgroundImage attribute with url', () => {
+    verifyHelper(
+        [
+          {
+            startTime: 62.05,
+            endTime: 3723.2,
+            payload: '',
+          },
+        ],
+        '<tt ' +
+        'xmlns:ttm="http://www.w3.org/ns/ttml#metadata" ' +
+        'xmlns:smpte="http://www.smpte-ra.org/schemas/2052-1/2010/smpte-tt">' +
+        '<metadata>' +
+        '<smpte:image imageType="PNG" encoding="Base64" xml:id="img_0">' +
+        'base64EncodedImage</smpte:image>' +
+        '</metadata>' +
+        '<body><div smpte:backgroundImage="img_0.png">' +
+        '<p begin="01:02.05" end="01:02:03.200"></p>' +
+        '</div></body></tt>',
+        {periodStart: 0, segmentStart: 60, segmentEnd: 3730, vttOffset: 0},
+        {startTime: 62.05, endTime: 3723.2},
+        {
+          startTime: 62.05,
+          endTime: 3723.2,
+          backgroundImage: 'foo://bar/img_0.png',
+          isContainer: false,
+        });
+  });
+
   it('supports smpte:backgroundImage attribute in div element', () => {
     verifyHelper(
         [],
@@ -2112,7 +2141,8 @@ describe('TtmlTextParser', () => {
   function verifyHelper(cues, text, time, bodyProperties, divProperties) {
     const data =
         shaka.util.BufferUtils.toUint8(shaka.util.StringUtils.toUTF8(text));
-    const result = new shaka.text.TtmlTextParser().parseMedia(data, time);
+    const result = new shaka.text.TtmlTextParser()
+        .parseMedia(data, time, 'foo://bar');
     shaka.test.TtmlUtils.verifyHelper(
         cues, result, bodyProperties, divProperties);
   }
@@ -2139,7 +2169,8 @@ describe('TtmlTextParser', () => {
     expect(() => {
       new shaka.text.TtmlTextParser().parseMedia(
           shaka.util.BufferUtils.toUint8(data),
-          {periodStart: 0, segmentStart: 0, segmentEnd: 10, vttOffset: 0});
+          {periodStart: 0, segmentStart: 0, segmentEnd: 10, vttOffset: 0},
+          'foo://bar');
     }).toThrow(error);
   }
 });