perf: Improve performance when parsing EMSG (#7557)

With the change, we reuse MediaSource's MP4 parsing code to avoid
parsing everything twice.
It will also help with the implementation of
#7556 in the future.

demo/config.js

@@ -487,8 +487,6 @@ shakaDemo.Config = class {
         .addNumberInput_('Update interval seconds',
             'streaming.updateIntervalSeconds',
             /* canBeDecimal= */ true)
-        .addBoolInput_('Dispatch all emsg boxes',
-            'streaming.dispatchAllEmsgBoxes')
         .addBoolInput_('Observe media quality changes',
             'streaming.observeQualityChanges')
         .addNumberInput_('Max Variant Disabled Time',
@@ -645,7 +643,9 @@ shakaDemo.Config = class {
             'Codec Switching Strategy',
             'mediaSource.codecSwitchingStrategy',
             strategyOptions,
-            strategyOptionsNames);
+            strategyOptionsNames)
+        .addBoolInput_('Dispatch all emsg boxes',
+            'mediaSource.dispatchAllEmsgBoxes');
   }
 
   /** @private */

externs/shaka/player.js

@@ -1557,7 +1557,6 @@ shaka.extern.LiveSyncConfiguration;
  *   minBytesForProgressEvents: number,
  *   preferNativeHls: boolean,
  *   updateIntervalSeconds: number,
- *   dispatchAllEmsgBoxes: boolean,
  *   observeQualityChanges: boolean,
  *   maxDisabledTime: number,
  *   segmentPrefetchLimit: number,
@@ -1739,10 +1738,6 @@ shaka.extern.LiveSyncConfiguration;
  *   The minimum number of seconds to see if the manifest has changes.
  *   <br>
  *   Defaults to <code>1</code>.
- * @property {boolean} dispatchAllEmsgBoxes
- *   If true, all emsg boxes are parsed and dispatched.
- *   <br>
- *   Defaults to <code>false</code>.
  * @property {boolean} observeQualityChanges
  *   If true, monitor media quality changes and emit
  *   <code>shaka.Player.MediaQualityChangedEvent</code>.
@@ -1850,7 +1845,8 @@ shaka.extern.StreamingConfiguration;
  *   addExtraFeaturesToSourceBuffer: function(string): string,
  *   forceTransmux: boolean,
  *   insertFakeEncryptionInInit: boolean,
- *   modifyCueCallback: shaka.extern.TextParser.ModifyCueCallback
+ *   modifyCueCallback: shaka.extern.TextParser.ModifyCueCallback,
+ *   dispatchAllEmsgBoxes: boolean
  * }}
  *
  * @description
@@ -1889,6 +1885,10 @@ shaka.extern.StreamingConfiguration;
  *    A callback called for each cue after it is parsed, but right before it
  *    is appended to the presentation.
  *    Gives a chance for client-side editing of cue text, cue timing, etc.
+ * @property {boolean} dispatchAllEmsgBoxes
+ *   If true, all emsg boxes are parsed and dispatched.
+ *   <br>
+ *   Defaults to <code>false</code>.
  * @exportDoc
  */
 shaka.extern.MediaSourceConfiguration;

lib/media/media_source_engine.js

@@ -859,15 +859,18 @@ shaka.media.MediaSourceEngine = class {
   }
 
   /**
+   * This method is only public for testing.
+   *
    * @param {shaka.util.ManifestParserUtils.ContentType} contentType
    * @param {!BufferSource} data
-   * @param {?shaka.media.SegmentReference} reference The segment reference
-   *   we are appending, or null for init segments
+   * @param {!shaka.media.SegmentReference} reference The segment reference
+   *   we are appending
+   * @param {shaka.extern.Stream} stream
    * @param {!string} mimeType
    * @return {{timestamp: ?number, metadata: !Array.<shaka.extern.ID3Metadata>}}
-   * @private
    */
-  getTimestampAndDispatchMetadata_(contentType, data, reference, mimeType) {
+  getTimestampAndDispatchMetadata(contentType, data, reference, stream,
+      mimeType) {
     let timestamp = null;
     let metadata = [];
 
@@ -894,16 +897,23 @@ shaka.media.MediaSourceEngine = class {
             [id3Metadata], /* offset= */ 0, reference.endTime);
       }
     } else if (mimeType.includes('/mp4') &&
-        reference && reference.timestampOffset == 0 &&
+        reference &&
         reference.initSegmentReference &&
         reference.initSegmentReference.timescale) {
       const timescale = reference.initSegmentReference.timescale;
       if (!isNaN(timescale)) {
+        const hasEmsg = ((stream.emsgSchemeIdUris != null &&
+            stream.emsgSchemeIdUris.length > 0) ||
+            this.config_.dispatchAllEmsgBoxes);
         const Mp4Parser = shaka.util.Mp4Parser;
         let startTime = 0;
         let parsedMedia = false;
-        new Mp4Parser()
-            .fullBox('prft', (box) => this.parsePrft_(timescale, box))
+        const parser = new Mp4Parser();
+        if (hasEmsg) {
+          parser.fullBox('emsg', (box) =>
+            this.parseEMSG_(reference, stream.emsgSchemeIdUris, box));
+        }
+        parser.fullBox('prft', (box) => this.parsePrft_(timescale, box))
             .box('moof', Mp4Parser.children)
             .box('traf', Mp4Parser.children)
             .fullBox('tfdt', (box) => {
@@ -916,7 +926,7 @@ shaka.media.MediaSourceEngine = class {
               parsedMedia = true;
               box.parser.stop();
             }).parse(data, /* partialOkay= */ true);
-        if (parsedMedia) {
+        if (parsedMedia && reference.timestampOffset == 0) {
           timestamp = startTime;
         }
       }
@@ -937,6 +947,130 @@ shaka.media.MediaSourceEngine = class {
     return {timestamp, metadata};
   }
 
+
+  /**
+   * Parse the EMSG box from a MP4 container.
+   *
+   * @param {!shaka.media.SegmentReference} reference
+   * @param {?Array.<string>} emsgSchemeIdUris Array of emsg
+   *     scheme_id_uri for which emsg boxes should be parsed.
+   * @param {!shaka.extern.ParsedBox} box
+   * @private
+   * https://dashif-documents.azurewebsites.net/Events/master/event.html#emsg-format
+   * aligned(8) class DASHEventMessageBox
+   *    extends FullBox(‘emsg’, version, flags = 0){
+   * if (version==0) {
+   *   string scheme_id_uri;
+   *   string value;
+   *   unsigned int(32) timescale;
+   *   unsigned int(32) presentation_time_delta;
+   *   unsigned int(32) event_duration;
+   *   unsigned int(32) id;
+   * } else if (version==1) {
+   *   unsigned int(32) timescale;
+   *   unsigned int(64) presentation_time;
+   *   unsigned int(32) event_duration;
+   *   unsigned int(32) id;
+   *   string scheme_id_uri;
+   *   string value;
+   * }
+   * unsigned int(8) message_data[];
+   */
+  parseEMSG_(reference, emsgSchemeIdUris, box) {
+    let timescale;
+    let id;
+    let eventDuration;
+    let schemeId;
+    let startTime;
+    let presentationTimeDelta;
+    let value;
+
+    if (box.version === 0) {
+      schemeId = box.reader.readTerminatedString();
+      value = box.reader.readTerminatedString();
+      timescale = box.reader.readUint32();
+      presentationTimeDelta = box.reader.readUint32();
+      eventDuration = box.reader.readUint32();
+      id = box.reader.readUint32();
+      startTime = reference.startTime + (presentationTimeDelta / timescale);
+    } else {
+      timescale = box.reader.readUint32();
+      const pts = box.reader.readUint64();
+      startTime = (pts / timescale) + reference.timestampOffset;
+      presentationTimeDelta = startTime - reference.startTime;
+      eventDuration = box.reader.readUint32();
+      id = box.reader.readUint32();
+      schemeId = box.reader.readTerminatedString();
+      value = box.reader.readTerminatedString();
+    }
+    const messageData = box.reader.readBytes(
+        box.reader.getLength() - box.reader.getPosition());
+
+    // See DASH sec. 5.10.3.3.1
+    // If a DASH client detects an event message box with a scheme that is not
+    // defined in MPD, the client is expected to ignore it.
+    if ((emsgSchemeIdUris && emsgSchemeIdUris.includes(schemeId)) ||
+        this.config_.dispatchAllEmsgBoxes) {
+      // See DASH sec. 5.10.4.1
+      // A special scheme in DASH used to signal manifest updates.
+      if (schemeId == 'urn:mpeg:dash:event:2012') {
+        this.playerInterface_.onManifestUpdate();
+      } else {
+        // All other schemes are dispatched as a general 'emsg' event.
+        const endTime = startTime + (eventDuration / timescale);
+        /** @type {shaka.extern.EmsgInfo} */
+        const emsg = {
+          startTime: startTime,
+          endTime: endTime,
+          schemeIdUri: schemeId,
+          value: value,
+          timescale: timescale,
+          presentationTimeDelta: presentationTimeDelta,
+          eventDuration: eventDuration,
+          id: id,
+          messageData: messageData,
+        };
+
+        // Dispatch an event to notify the application about the emsg box.
+        const eventName = shaka.util.FakeEvent.EventName.Emsg;
+        const data = (new Map()).set('detail', emsg);
+        const event = new shaka.util.FakeEvent(eventName, data);
+        // A user can call preventDefault() on a cancelable event.
+        event.cancelable = true;
+
+        this.playerInterface_.onEvent(event);
+
+        if (event.defaultPrevented) {
+          // If the caller uses preventDefault() on the 'emsg' event, don't
+          // process any further, and don't generate an ID3 'metadata' event
+          // for the same data.
+          return;
+        }
+
+        // Additionally, ID3 events generate a 'metadata' event.  This is a
+        // pre-parsed version of the metadata blob already dispatched in the
+        // 'emsg' event.
+        if (schemeId == 'https://aomedia.org/emsg/ID3' ||
+            schemeId == 'https://developer.apple.com/streaming/emsg-id3') {
+          // See https://aomediacodec.github.io/id3-emsg/
+          const frames = shaka.util.Id3Utils.getID3Frames(messageData);
+          if (frames.length) {
+            /** @private {shaka.extern.ID3Metadata} */
+            const metadata = {
+              cueTime: startTime,
+              data: messageData,
+              frames: frames,
+              dts: startTime,
+              pts: startTime,
+            };
+            this.playerInterface_.onMetadata(
+                [metadata], /* offset= */ 0, endTime);
+          }
+        }
+      }
+    }
+  }
+
   /**
    * Parse PRFT box.
    * @param {number} timescale
@@ -1040,8 +1174,8 @@ shaka.media.MediaSourceEngine = class {
       mimeType = this.transmuxers_[contentType].getOriginalMimeType();
     }
     if (reference) {
-      const {timestamp, metadata} = this.getTimestampAndDispatchMetadata_(
-          contentType, data, reference, mimeType);
+      const {timestamp, metadata} = this.getTimestampAndDispatchMetadata(
+          contentType, data, reference, stream, mimeType);
       if (timestamp != null) {
         if (this.firstVideoTimestamp_ == null &&
             contentType == ContentType.VIDEO) {
@@ -2342,7 +2476,8 @@ shaka.media.MediaSourceEngine.SourceBufferMode_ = {
  * @typedef {{
  *   getKeySystem: function():?string,
  *   onMetadata: function(!Array<shaka.extern.ID3Metadata>, number, ?number),
- *   onEvent: function(!Event)
+ *   onEvent: function(!Event),
+ *   onManifestUpdate: function()
  * }}
  *
  * @summary Player interface
@@ -2353,5 +2488,7 @@ shaka.media.MediaSourceEngine.SourceBufferMode_ = {
  *   Callback to use when metadata arrives.
  * @property {function(!Event)} onEvent
  *   Called when an event occurs that should be sent to the app.
+ * @property {function()} onManifestUpdate
+ *   Called when an embedded 'emsg' box should trigger a manifest update.
  */
 shaka.media.MediaSourceEngine.PlayerInterface;