Merge pull request #442 from toasted-nutbread/audio-system-refactoring

Audio system refactoring

3 files changed, 78 insertions, 41 deletions

diff --git a/ext/mixed/js/api.js b/ext/mixed/js/api.js
index 7080d93a..c97dc687 100644
--- a/ext/mixed/js/api.js
+++ b/ext/mixed/js/api.js
@@ -64,8 +64,8 @@ function apiTemplateRender(template, data) {
     return _apiInvoke('templateRender', {data, template});
 }
 
-function apiAudioGetUri(definition, source, optionsContext) {
-    return _apiInvoke('audioGetUri', {definition, source, optionsContext});
+function apiAudioGetUri(definition, source, details) {
+    return _apiInvoke('audioGetUri', {definition, source, details});
 }
 
 function apiCommandExec(command, params) {
diff --git a/ext/mixed/js/audio-system.js b/ext/mixed/js/audio-system.js
index 45b733fc..3273f982 100644
--- a/ext/mixed/js/audio-system.js
+++ b/ext/mixed/js/audio-system.js
@@ -40,7 +40,7 @@ class TextToSpeechAudio {
         }
     }
 
-    play() {
+    async play() {
         try {
             if (this._utterance === null) {
                 this._utterance = new SpeechSynthesisUtterance(this.text || '');
@@ -66,10 +66,10 @@ class TextToSpeechAudio {
 }
 
 class AudioSystem {
-    constructor({getAudioUri}) {
-        this._cache = new Map();
+    constructor({audioUriBuilder, useCache}) {
+        this._cache = useCache ? new Map() : null;
         this._cacheSizeMaximum = 32;
-        this._getAudioUri = getAudioUri;
+        this._audioUriBuilder = audioUriBuilder;
 
         if (typeof speechSynthesis !== 'undefined') {
             // speechSynthesis.getVoices() will not be populated unless some API call is made.
@@ -79,21 +79,31 @@ class AudioSystem {
 
     async getDefinitionAudio(definition, sources, details) {
         const key = `${definition.expression}:${definition.reading}`;
-        const cacheValue = this._cache.get(definition);
-        if (typeof cacheValue !== 'undefined') {
-            const {audio, uri, source} = cacheValue;
-            return {audio, uri, source};
+        const hasCache = (this._cache !== null);
+
+        if (hasCache) {
+            const cacheValue = this._cache.get(key);
+            if (typeof cacheValue !== 'undefined') {
+                const {audio, uri, source} = cacheValue;
+                const index = sources.indexOf(source);
+                if (index >= 0) {
+                    return {audio, uri, index};
+                }
+            }
         }
 
-        for (const source of sources) {
+        for (let i = 0, ii = sources.length; i < ii; ++i) {
+            const source = sources[i];
             const uri = await this._getAudioUri(definition, source, details);
             if (uri === null) { continue; }
 
             try {
-                const audio = await this._createAudio(uri, details);
-                this._cacheCheck();
-                this._cache.set(key, {audio, uri, source});
-                return {audio, uri, source};
+                const audio = await this._createAudio(uri);
+                if (hasCache) {
+                    this._cacheCheck();
+                    this._cache.set(key, {audio, uri, source});
+                }
+                return {audio, uri, index: i};
             } catch (e) {
                 // NOP
             }
@@ -102,7 +112,7 @@ class AudioSystem {
         throw new Error('Could not create audio');
     }
 
-    createTextToSpeechAudio({text, voiceUri}) {
+    createTextToSpeechAudio(text, voiceUri) {
         const voice = this._getTextToSpeechVoiceFromVoiceUri(voiceUri);
         if (voice === null) {
             throw new Error('Invalid text-to-speech voice');
@@ -114,20 +124,24 @@ class AudioSystem {
         // NOP
     }
 
-    async _createAudio(uri, details) {
+    async _createAudio(uri) {
         const ttsParameters = this._getTextToSpeechParameters(uri);
         if (ttsParameters !== null) {
-            if (typeof details === 'object' && details !== null) {
-                if (details.tts === false) {
-                    throw new Error('Text-to-speech not permitted');
-                }
-            }
-            return this.createTextToSpeechAudio(ttsParameters);
+            const {text, voiceUri} = ttsParameters;
+            return this.createTextToSpeechAudio(text, voiceUri);
         }
 
         return await this._createAudioFromUrl(uri);
     }
 
+    _getAudioUri(definition, source, details) {
+        return (
+            this._audioUriBuilder !== null ?
+            this._audioUriBuilder.getUri(definition, source, details) :
+            null
+        );
+    }
+
     _createAudioFromUrl(url) {
         return new Promise((resolve, reject) => {
             const audio = new Audio(url);
diff --git a/ext/mixed/js/display.js b/ext/mixed/js/display.js
index 63687dc2..c2284ffe 100644
--- a/ext/mixed/js/display.js
+++ b/ext/mixed/js/display.js
@@ -45,7 +45,14 @@ class Display {
         this.index = 0;
         this.audioPlaying = null;
         this.audioFallback = null;
-        this.audioSystem = new AudioSystem({getAudioUri: this._getAudioUri.bind(this)});
+        this.audioSystem = new AudioSystem({
+            audioUriBuilder: {
+                getUri: async (definition, source, details) => {
+                    return await apiAudioGetUri(definition, source, details);
+                }
+            },
+            useCache: true
+        });
         this.styleNode = null;
 
         this.eventListeners = new EventListenerCollection();
@@ -784,16 +791,14 @@ class Display {
 
             const expression = expressionIndex === -1 ? definition : definition.expressions[expressionIndex];
 
-            if (this.audioPlaying !== null) {
-                this.audioPlaying.pause();
-                this.audioPlaying = null;
-            }
+            this._stopPlayingAudio();
 
-            const sources = this.options.audio.sources;
-            let audio, source, info;
+            let audio, info;
             try {
-                ({audio, source} = await this.audioSystem.getDefinitionAudio(expression, sources));
-                info = `From source ${1 + sources.indexOf(source)}: ${source}`;
+                const {sources, textToSpeechVoice, customSourceUrl} = this.options.audio;
+                let index;
+                ({audio, index} = await this.audioSystem.getDefinitionAudio(expression, sources, {textToSpeechVoice, customSourceUrl}));
+                info = `From source ${1 + index}: ${sources[index]}`;
             } catch (e) {
                 if (this.audioFallback === null) {
                     this.audioFallback = new Audio('/mixed/mp3/button.mp3');
@@ -802,7 +807,7 @@ class Display {
                 info = 'Could not find audio';
             }
 
-            const button = this.audioButtonFindImage(entryIndex);
+            const button = this.audioButtonFindImage(entryIndex, expressionIndex);
             if (button !== null) {
                 let titleDefault = button.dataset.titleDefault;
                 if (!titleDefault) {
@@ -812,10 +817,19 @@ class Display {
                 button.title = `${titleDefault}\n${info}`;
             }
 
+            this._stopPlayingAudio();
+
             this.audioPlaying = audio;
             audio.currentTime = 0;
             audio.volume = this.options.audio.volume / 100.0;
-            audio.play();
+            const playPromise = audio.play();
+            if (typeof playPromise !== 'undefined') {
+                try {
+                    await playPromise;
+                } catch (e2) {
+                    // NOP
+                }
+            }
         } catch (e) {
             this.onError(e);
         } finally {
@@ -823,6 +837,13 @@ class Display {
         }
     }
 
+    _stopPlayingAudio() {
+        if (this.audioPlaying !== null) {
+            this.audioPlaying.pause();
+            this.audioPlaying = null;
+        }
+    }
+
     noteUsesScreenshot(mode) {
         const optionsAnki = this.options.anki;
         const fields = (mode === 'kanji' ? optionsAnki.kanji : optionsAnki.terms).fields;
@@ -901,9 +922,16 @@ class Display {
         viewerButton.dataset.noteId = noteId;
     }
 
-    audioButtonFindImage(index) {
+    audioButtonFindImage(index, expressionIndex) {
         const entry = this.getEntry(index);
-        return entry !== null ? entry.querySelector('.action-play-audio>img') : null;
+        if (entry === null) { return null; }
+
+        const container = (
+            expressionIndex >= 0 ?
+            entry.querySelector(`.term-expression:nth-of-type(${expressionIndex + 1})`) :
+            entry
+        );
+        return container !== null ? container.querySelector('.action-play-audio>img') : null;
     }
 
     async getDefinitionsAddable(definitions, modes) {
@@ -947,9 +975,4 @@ class Display {
             }
         };
     }
-
-    async _getAudioUri(definition, source) {
-        const optionsContext = this.getOptionsContext();
-        return await apiAudioGetUri(definition, source, optionsContext);
-    }
 }