feat: Add originalLanguage to the Track structure (shaka-project#5409)

Shaka in most of places normalizes tracks' language code to be compliant with ISO 639-1 when possible. However, it does not do that all the time (i.e. normalization is missing in MSS parser) and there is no way to get value that has been explicitly set in a manifest. Moreover, documentation is misleading, as it claims that value is taken directly from a manifest.
Normalization should take place, specifically to easify PeriodCombiner algorithm and also to not break existing applications.
However, original value can be desired for some implementations.
This PR introduces new field to get original language value from the manifest.

externs/shaka/manifest.js

@@ -344,6 +344,7 @@ shaka.extern.FetchCryptoKeysFunction;
  *   drmInfos: !Array.<shaka.extern.DrmInfo>,
  *   keyIds: !Set.<string>,
  *   language: string,
+ *   originalLanguage: ?string,
  *   label: ?string,
  *   type: string,
  *   primary: boolean,
@@ -429,6 +430,9 @@ shaka.extern.FetchCryptoKeysFunction;
  *   The Stream's language, specified as a language code. <br>
  *   Audio stream's language must be identical to the language of the containing
  *   Variant.
+ * @property {?string} originalLanguage
+ *   <i>Optional.</i> <br>
+ *   The original language, if any, that appeared in the manifest.
  * @property {?string} label
  *   The Stream's label, unique text that should describe the audio/text track.
  * @property {string} type

externs/shaka/offline.js

@@ -123,6 +123,7 @@ shaka.extern.ManifestDB;
  *   hdr: (string|undefined),
  *   kind: (string|undefined),
  *   language: string,
+ *   originalLanguage: (?string|undefined),
  *   label: ?string,
  *   width: ?number,
  *   height: ?number,
@@ -162,6 +163,8 @@ shaka.extern.ManifestDB;
  *   The kind of text stream; undefined for audio/video.
  * @property {string} language
  *   The language of the stream; '' for video.
+ * @property {(?string|undefined)} originalLanguage
+ *   The original language, if any, that appeared in the manifest.
  * @property {?string} label
  *   The label of the stream; '' for video.
  * @property {?number} width

externs/shaka/player.js

@@ -238,7 +238,8 @@ shaka.extern.BufferedInfo;
  *   originalVideoId: ?string,
  *   originalAudioId: ?string,
  *   originalTextId: ?string,
- *   originalImageId: ?string
+ *   originalImageId: ?string,
+ *   originalLanguage: ?string
  * }}
  *
  * @description
@@ -260,8 +261,10 @@ shaka.extern.BufferedInfo;
  *   The bandwidth required to play the track, in bits/sec.
  *
  * @property {string} language
- *   The language of the track, or <code>'und'</code> if not given.  This is the
- *   exact value provided in the manifest; it may need to be normalized.
+ *   The language of the track, or <code>'und'</code> if not given.  This value
+ *   is normalized as follows - language part is always lowercase and translated
+ *   to ISO-639-1 when possible, locale part is always uppercase,
+ *   i.e. <code>'en-US'</code>.
  * @property {?string} label
  *   The track label, which is unique text that should describe the track.
  * @property {?string} kind
@@ -340,6 +343,10 @@ shaka.extern.BufferedInfo;
  * @property {?string} originalImageId
  *   (image tracks only) The original ID of the image track, if any, as it
  *   appeared in the original manifest.
+ * @property {?string} originalLanguage
+ *   The original language of the track, if any, as it appeared in the original
+ *   manifest.  This is the exact value provided in the manifest; for normalized
+ *   value use <code>language</code> property.
  * @exportDoc
  */
 shaka.extern.Track;

lib/dash/dash_parser.js

@@ -1108,8 +1108,8 @@ shaka.dash.DashParser = class {
         this.config_.dash.ignoreDrmInfo,
         this.config_.dash.keySystemsByURI);
 
-    const language =
-        shaka.util.LanguageUtils.normalize(elem.getAttribute('lang') || 'und');
+    const language = shaka.util.LanguageUtils.normalize(
+        context.adaptationSet.language || 'und');
 
     // This attribute is currently non-standard, but it is supported by Kaltura.
     let label = elem.getAttribute('label');
@@ -1380,6 +1380,7 @@ shaka.dash.DashParser = class {
         drmInfos: contentProtection.drmInfos,
         keyIds,
         language,
+        originalLanguage: context.adaptationSet.language,
         label,
         type: context.adaptationSet.contentType,
         primary: isPrimary,
@@ -1569,6 +1570,7 @@ shaka.dash.DashParser = class {
       pixelAspectRatio: pixelAspectRatio,
       emsgSchemeIdUris: emsgSchemeIdUris,
       id: elem.getAttribute('id'),
+      language: elem.getAttribute('lang'),
       numChannels: numChannels,
       audioSamplingRate: audioSamplingRate,
       availabilityTimeOffset: availabilityTimeOffset,
@@ -1979,6 +1981,7 @@ shaka.dash.DashParser.RequestSegmentCallback;
  *   pixelAspectRatio: (string|undefined),
  *   emsgSchemeIdUris: !Array.<string>,
  *   id: ?string,
+ *   language: ?string,
  *   numChannels: ?number,
  *   audioSamplingRate: ?number,
  *   availabilityTimeOffset: number
@@ -2014,6 +2017,8 @@ shaka.dash.DashParser.RequestSegmentCallback;
  *   emsg registered schemeIdUris.
  * @property {?string} id
  *   The ID of the element.
+ * @property {?string} language
+ *   The original language of the element.
  * @property {?number} numChannels
  *   The number of audio channels, or null if unknown.
  * @property {?number} audioSamplingRate

lib/hls/hls_parser.js

@@ -666,7 +666,7 @@ shaka.hls.HlsParser = class {
       const type = basicInfo.type;
       const mimeType = basicInfo.mimeType;
       const codecs = basicInfo.codecs;
-      const language = basicInfo.language || 'und';
+      const languageValue = basicInfo.language;
       const height = basicInfo.height;
       const width = basicInfo.width;
       const channelsCount = basicInfo.channelCount;
@@ -685,7 +685,7 @@ shaka.hls.HlsParser = class {
 
       // Make the stream info, with those values.
       const streamInfo = await this.convertParsedPlaylistIntoStreamInfo_(
-          playlist, uri, uri, codecs, type, language, primary, name,
+          playlist, uri, uri, codecs, type, languageValue, primary, name,
           channelsCount, closedCaptions, characteristics, forced, sampleRate,
           spatialAudio, mimeType);
       this.uriToStreamInfosMap_.set(uri, streamInfo);
@@ -698,7 +698,7 @@ shaka.hls.HlsParser = class {
       // Wrap the stream from that stream info with a variant.
       variants.push({
         id: 0,
-        language: language,
+        language: this.getLanguage_(languageValue),
         disabledUntilTime: 0,
         primary: true,
         audio: type == 'audio' ? streamInfo.stream : null,
@@ -1609,16 +1609,15 @@ shaka.hls.HlsParser = class {
   }
 
   /**
-   * Get the language value.
+   * Get the normalized language value.
    *
-   * @param {!shaka.hls.Tag} tag
+   * @param {?string} languageValue
    * @return {string}
    * @private
    */
-  getLanguage_(tag) {
+  getLanguage_(languageValue) {
     const LanguageUtils = shaka.util.LanguageUtils;
-    const languageValue = tag.getAttributeValue('LANGUAGE') || 'und';
-    return LanguageUtils.normalize(languageValue);
+    return LanguageUtils.normalize(languageValue || 'und');
   }
 
   /**
@@ -1751,7 +1750,8 @@ shaka.hls.HlsParser = class {
     for (const tag of closedCaptionsTags) {
       goog.asserts.assert(tag.name == 'EXT-X-MEDIA',
           'Should only be called on media tags!');
-      const language = this.getLanguage_(tag);
+      const languageValue = tag.getAttributeValue('LANGUAGE');
+      const language = this.getLanguage_(languageValue);
 
       // The GROUP-ID value is a quoted-string that specifies the group to which
       // the Rendition belongs.
@@ -1797,7 +1797,7 @@ shaka.hls.HlsParser = class {
       return this.uriToStreamInfosMap_.get(verbatimMediaPlaylistUri);
     }
 
-    const language = this.getLanguage_(tag);
+    const language = tag.getAttributeValue('LANGUAGE');
     const name = tag.getAttributeValue('NAME');
 
     // NOTE: According to the HLS spec, "DEFAULT=YES" requires "AUTOSELECT=YES".
@@ -1860,7 +1860,7 @@ shaka.hls.HlsParser = class {
       return this.uriToStreamInfosMap_.get(verbatimImagePlaylistUri);
     }
 
-    const language = this.getLanguage_(tag);
+    const language = tag.getAttributeValue('LANGUAGE');
     const name = tag.getAttributeValue('NAME');
 
     const characteristics = tag.getAttributeValue('CHARACTERISTICS');
@@ -1930,7 +1930,7 @@ shaka.hls.HlsParser = class {
     const closedCaptions = this.getClosedCaptions_(tag, type);
     const codecs = shaka.util.ManifestParserUtils.guessCodecs(type, allCodecs);
     const streamInfo = this.createStreamInfo_(verbatimMediaPlaylistUri,
-        codecs, type, /* language= */ 'und', /* primary= */ false,
+        codecs, type, /* language= */ null, /* primary= */ false,
         /* name= */ null, /* channelcount= */ null, closedCaptions,
         /* characteristics= */ null, /* forced= */ false,
         /* sampleRate= */ null, /* spatialAudio= */ false);
@@ -1949,7 +1949,7 @@ shaka.hls.HlsParser = class {
    * @param {string} verbatimMediaPlaylistUri
    * @param {string} codecs
    * @param {string} type
-   * @param {string} language
+   * @param {?string} languageValue
    * @param {boolean} primary
    * @param {?string} name
    * @param {?number} channelsCount
@@ -1961,7 +1961,7 @@ shaka.hls.HlsParser = class {
    * @return {!shaka.hls.HlsParser.StreamInfo}
    * @private
    */
-  createStreamInfo_(verbatimMediaPlaylistUri, codecs, type, language,
+  createStreamInfo_(verbatimMediaPlaylistUri, codecs, type, languageValue,
       primary, name, channelsCount, closedCaptions, characteristics, forced,
       sampleRate, spatialAudio) {
     // TODO: Refactor, too many parameters
@@ -1971,9 +1971,9 @@ shaka.hls.HlsParser = class {
     // This stream is lazy-loaded inside the createSegmentIndex function.
     // So we start out with a stream object that does not contain the actual
     // segment index, then download when createSegmentIndex is called.
-    const stream = this.makeStreamObject_(codecs, type, language, primary, name,
-        channelsCount, closedCaptions, characteristics, forced, sampleRate,
-        spatialAudio);
+    const stream = this.makeStreamObject_(codecs, type, languageValue, primary,
+        name, channelsCount, closedCaptions, characteristics, forced,
+        sampleRate, spatialAudio);
     const streamInfo = {
       stream,
       type,
@@ -2013,7 +2013,7 @@ shaka.hls.HlsParser = class {
       const wasLive = this.isLive_();
       const realStreamInfo = await this.convertParsedPlaylistIntoStreamInfo_(
           playlist, verbatimMediaPlaylistUri, absoluteMediaPlaylistUri, codecs,
-          type, language, primary, name, channelsCount, closedCaptions,
+          type, languageValue, primary, name, channelsCount, closedCaptions,
           characteristics, forced, sampleRate, spatialAudio);
       if (abortSignal.aborted) {
         return;
@@ -2250,7 +2250,7 @@ shaka.hls.HlsParser = class {
    * @param {string} absoluteMediaPlaylistUri
    * @param {string} codecs
    * @param {string} type
-   * @param {string} language
+   * @param {?string} languageValue
    * @param {boolean} primary
    * @param {?string} name
    * @param {?number} channelsCount
@@ -2264,7 +2264,7 @@ shaka.hls.HlsParser = class {
    * @private
    */
   async convertParsedPlaylistIntoStreamInfo_(playlist, verbatimMediaPlaylistUri,
-      absoluteMediaPlaylistUri, codecs, type, language, primary, name,
+      absoluteMediaPlaylistUri, codecs, type, languageValue, primary, name,
       channelsCount, closedCaptions, characteristics, forced, sampleRate,
       spatialAudio, mimeType = undefined) {
     if (playlist.type != shaka.hls.PlaylistType.MEDIA) {
@@ -2335,9 +2335,9 @@ shaka.hls.HlsParser = class {
     const {nextMediaSequence, nextPart} =
         this.getNextMediaSequenceAndPart_(mediaSequenceNumber, segments);
 
-    const stream = this.makeStreamObject_(codecs, type, language, primary, name,
-        channelsCount, closedCaptions, characteristics, forced, sampleRate,
-        spatialAudio);
+    const stream = this.makeStreamObject_(codecs, type, languageValue, primary,
+        name, channelsCount, closedCaptions, characteristics, forced,
+        sampleRate, spatialAudio);
     stream.segmentIndex = segmentIndex;
     stream.encrypted = encrypted;
     stream.drmInfos = drmInfos;
@@ -2414,7 +2414,7 @@ shaka.hls.HlsParser = class {
    * manually on the object after creation.
    * @param {string} codecs
    * @param {string} type
-   * @param {string} language
+   * @param {?string} languageValue
    * @param {boolean} primary
    * @param {?string} name
    * @param {?number} channelsCount
@@ -2426,7 +2426,7 @@ shaka.hls.HlsParser = class {
    * @return {!shaka.extern.Stream}
    * @private
    */
-  makeStreamObject_(codecs, type, language, primary, name, channelsCount,
+  makeStreamObject_(codecs, type, languageValue, primary, name, channelsCount,
       closedCaptions, characteristics, forced, sampleRate, spatialAudio) {
     // Fill out a "best-guess" mimeType, for now. It will be replaced once the
     // stream is lazy-loaded.
@@ -2445,7 +2445,8 @@ shaka.hls.HlsParser = class {
       encrypted: false,
       drmInfos: [],
       keyIds: new Set(),
-      language,
+      language: this.getLanguage_(languageValue),
+      originalLanguage: languageValue,
       label: name,  // For historical reasons, since before "originalId".
       type,
       primary,

lib/mss/mss_parser.js

@@ -17,6 +17,7 @@ goog.require('shaka.media.SegmentReference');
 goog.require('shaka.mss.ContentProtection');
 goog.require('shaka.net.NetworkingEngine');
 goog.require('shaka.util.Error');
+goog.require('shaka.util.LanguageUtils');
 goog.require('shaka.util.ManifestParserUtils');
 goog.require('shaka.util.Mp4Generator');
 goog.require('shaka.util.OperationManager');
@@ -513,7 +514,8 @@ shaka.mss.MssParser = class {
       encrypted: drmInfos.length > 0,
       drmInfos: drmInfos,
       keyIds: new Set(),
-      language: lang || 'und',
+      language: shaka.util.LanguageUtils.normalize(lang || 'und'),
+      originalLanguage: lang,
       label: '',
       type: '',
       primary: false,

lib/offline/indexeddb/v1_storage_cell.js

@@ -159,6 +159,7 @@ shaka.offline.indexeddb.V1StorageCell = class
       hdr: undefined,
       kind: old.kind,
       language: old.language,
+      originalLanguage: old.language || null,
       label: old.label,
       width: old.width,
       height: old.height,

lib/offline/indexeddb/v2_storage_cell.js

@@ -108,6 +108,7 @@ shaka.offline.indexeddb.V2StorageCell = class
       hdr: undefined,
       kind: old.kind,
       language: old.language,
+      originalLanguage: old.language || null,
       label: old.label,
       width: old.width,
       height: old.height,

lib/offline/manifest_converter.js

@@ -194,6 +194,7 @@ shaka.offline.ManifestConverter = class {
       drmInfos: [],
       keyIds: streamDB.keyIds,
       language: streamDB.language,
+      originalLanguage: streamDB.originalLanguage || null,
       label: streamDB.label,
       type: streamDB.type,
       primary: streamDB.primary,

lib/offline/storage.js

@@ -1308,6 +1308,7 @@ shaka.offline.Storage = class {
       hdr: stream.hdr,
       kind: stream.kind,
       language: stream.language,
+      originalLanguage: stream.originalLanguage,
       label: stream.label,
       width: stream.width || null,
       height: stream.height || null,

lib/player.js

@@ -2382,6 +2382,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
         drmInfos: [],  // Filled in by DrmEngine config.
         keyIds: new Set(),
         language: 'und',
+        originalLanguage: null,
         label: null,
         type: ContentType.VIDEO,
         primary: false,
@@ -4873,6 +4874,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
       drmInfos: [],
       keyIds: new Set(),
       language: language,
+      originalLanguage: language,
       label: label || null,
       type: ContentType.TEXT,
       primary: false,
@@ -5024,6 +5026,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
       drmInfos: [],
       keyIds: new Set(),
       language: 'und',
+      originalLanguage: null,
       label: null,
       type: ContentType.IMAGE,
       primary: false,
@@ -5433,6 +5436,7 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
             // Add an empty segmentIndex, for the benefit of the period combiner
             // in our builtin DASH parser.
             const segmentIndex = new shaka.media.MetaSegmentIndex();
+            const language = video.closedCaptions.get(id);
             const textStream = {
               id: this.nextExternalStreamId_++,  // A globally unique ID.
               originalId: id, // The CC ID string, like 'CC1', 'CC3', etc.
@@ -5444,7 +5448,8 @@ shaka.Player = class extends shaka.util.FakeEventTarget {
               encrypted: false,
               drmInfos: [],
               keyIds: new Set(),
-              language: video.closedCaptions.get(id),
+              language,
+              originalLanguage: language,
               label: null,
               type: ContentType.TEXT,
               primary: false,