Polyfill using webkitAudioContext (#3576)

* Polyfill using webkitAudioContext * Update entry * Apply suggestions from code review Co-authored-by: Corina <14900841+corinagum@users.noreply.github.com> * Rephrase HACK to WORKAROUND Co-authored-by: Corina <14900841+corinagum@users.noreply.github.com>
microsoft · Nov 3, 2020 · 88e3a8f · 88e3a8f
1 parent efaf3db
commit 88e3a8f
Show file tree

Hide file tree

Showing 5 changed files with 68 additions and 8 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -44,6 +44,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 -  Fixes [#3514](https://github.com/microsoft/BotFramework-WebChat/issues/3514). Fix PoliCheck language errors, by [@corinagum](https://github.com/corinagum) in PR [#3545](https://github.com/microsoft/BotFramework-WebChat/pull/3545)
 -  Fixes [#3537](https://github.com/microsoft/BotFramework-WebChat/issues/3537). [Accessibility]: Ensure `aria-roledescription` is only used on elements with implicit/explicit role based off of [WAI ARIA role attributes](https://www.w3.org/WAI/PF/aria/roles), by [@corinagum](https://github.com/corinagum) in PR [#3551](https://github.com/microsoft/BotFramework-WebChat/pull/3551)
 -  Fixes [#3431](https://github.com/microsoft/BotFramework-WebChat/issues/3431). Activities should not be delayed due to missing activity of type "typing", by [@compulim](https://github.com/compulim) in PR [#3554](https://github.com/microsoft/BotFramework-WebChat/pull/3554)
+-  Fixes [#3574](https://github.com/microsoft/BotFramework-WebChat/issues/3574). Creates workaround for Cognitive Services Speech SDK 1.13.1 regarding removed support of macOS/iOS, by [@compulim](https://github.com/compulim) in PR [#3576](https://github.com/microsoft/BotFramework-WebChat/pull/3576)
 
 ### Changed
 

diff --git a/packages/bundle/src/createCognitiveServicesSpeechServicesPonyfillFactory.js b/packages/bundle/src/createCognitiveServicesSpeechServicesPonyfillFactory.js
@@ -1,4 +1,5 @@
 import { AudioConfig } from 'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig';
+import { MicAudioSource } from 'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource';
 import createPonyfill from 'web-speech-cognitive-services/lib/SpeechServices';
 
 function resolveFunction(fnOrValue) {
@@ -53,16 +54,38 @@ export default function createCognitiveServicesSpeechServicesPonyfillFactory({
     );
   }
 
-  // HACK: We should prevent AudioContext object from being recreated because they may be blessed and UX-wise expensive to recreate.
-  //       In Cognitive Services SDK, if they detect the "end" function is falsy, they will not call "end" but "suspend" instead.
-  //       And on next recognition, they will re-use the AudioContext object.
+  // WORKAROUND: We should prevent AudioContext object from being recreated because they may be blessed and UX-wise expensive to recreate.
+  //             In Cognitive Services SDK, if they detect the "end" function is falsy, they will not call "end" but "suspend" instead.
+  //             And on next recognition, they will re-use the AudioContext object.
   if (!audioConfig) {
     audioConfig = audioInputDeviceId
       ? AudioConfig.fromMicrophoneInput(audioInputDeviceId)
       : AudioConfig.fromDefaultMicrophoneInput();
 
     const source = audioConfig.privSource;
 
+    // WORKAROUND: In Speech SDK 1.12.0-1.13.1, it dropped support of macOS/iOS Safari.
+    //             This code is adopted from microsoft-cognitiveservices-speech-sdk/src/common.browser/MicAudioSource.ts.
+    //             We will not need this code when using Speech SDK 1.14.0 or up.
+    // TODO: [P1] #3575 Remove the following lines when bumping to Speech SDK 1.14.0 or higher
+    source.createAudioContext = () => {
+      if (!!source.privContext) {
+        return;
+      }
+
+      const AudioContext = window.AudioContext || window.webkitAudioContext;
+
+      if (typeof AudioContext === 'undefined') {
+        throw new Error('Browser does not support Web Audio API (AudioContext/webkitAudioContext is not available).');
+      }
+
+      if (navigator.mediaDevices.getSupportedConstraints().sampleRate) {
+        source.privContext = new AudioContext({ sampleRate: MicAudioSource.AUDIOFORMAT.samplesPerSec });
+      } else {
+        source.privContext = new AudioContext();
+      }
+    };
+
     // This piece of code is adopted from microsoft-cognitiveservices-speech-sdk/common.browser/MicAudioSource.ts.
     // Instead of closing the AudioContext, it will just suspend it. And the next time it is needed, it will be resumed (by the original code).
     source.destroyAudioContext = () => {

diff --git a/packages/bundle/webpack.config.js b/packages/bundle/webpack.config.js
@@ -39,6 +39,8 @@ let config = {
   resolve: {
     alias: {
       'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig': resolve(__dirname, 'node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig.js'),
+      // TODO: [P1] #3575 Remove the following line when bumping to Speech SDK 1.14.0 or higher
+      'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource': resolve(__dirname, 'node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource.js'),
       'microsoft-cognitiveservices-speech-sdk/distrib/lib/microsoft.cognitiveservices.speech.sdk': resolve(__dirname, 'node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/microsoft.cognitiveservices.speech.sdk.js'),
       'microsoft-cognitiveservices-speech-sdk': resolve(__dirname, 'node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/microsoft.cognitiveservices.speech.sdk.js'),
       react: resolve(__dirname, 'node_modules/isomorphic-react/dist/react.js'),

diff --git a/packages/directlinespeech/src/createAdapters.js b/packages/directlinespeech/src/createAdapters.js
@@ -2,6 +2,7 @@
 
 import { AudioConfig } from 'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig';
 import { BotFrameworkConfig, DialogServiceConnector, PropertyId } from 'microsoft-cognitiveservices-speech-sdk';
+import { MicAudioSource } from 'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource';
 
 import createWebSpeechPonyfillFactory from './createWebSpeechPonyfillFactory';
 import DirectLineSpeech from './DirectLineSpeech';
@@ -70,6 +71,30 @@ export default async function create({
     } else {
       audioConfig = AudioConfig.fromDefaultMicrophoneInput();
     }
+
+    // WORKAROUND: In Speech SDK 1.12.0-1.13.1, it dropped support of macOS/iOS Safari.
+    //             This code is adopted from microsoft-cognitiveservices-speech-sdk/src/common.browser/MicAudioSource.ts.
+    //             We will not need this code when using Speech SDK 1.14.0 or up.
+    // TODO: [P1] #3575 Remove the following lines when bumping to Speech SDK 1.14.0 or higher
+    const { privSource: source } = audioConfig;
+
+    source.createAudioContext = () => {
+      if (!!source.privContext) {
+        return;
+      }
+
+      const AudioContext = window.AudioContext || window.webkitAudioContext;
+
+      if (typeof AudioContext === 'undefined') {
+        throw new Error('Browser does not support Web Audio API (AudioContext/webkitAudioContext is not available).');
+      }
+
+      if (navigator.mediaDevices.getSupportedConstraints().sampleRate) {
+        source.privContext = new AudioContext({ sampleRate: MicAudioSource.AUDIOFORMAT.samplesPerSec });
+      } else {
+        source.privContext = new AudioContext();
+      }
+    };
   }
 
   if (speechRecognitionEndpointId) {
@@ -180,18 +205,18 @@ export default async function create({
       if (dialogServiceConnector.privIsDisposed) {
         clearInterval(interval);
       }
-      
+
       const refreshedDirectLineToken = await refreshDirectLineToken(directLineToken);
-      
+
       if (!refreshedDirectLineToken) {
         return console.warn(
           'botframework-directlinespeech-sdk: Renew token failed because call to refresh token Direct Line API did not return a new token.'
         );
       }
 
       config.setProperty(PropertyId.Conversation_ApplicationId, refreshedDirectLineToken);
-      
-      dialogServiceConnector.properties.setProperty(PropertyId.Conversation_ApplicationId, refreshedDirectLineToken)
+
+      dialogServiceConnector.properties.setProperty(PropertyId.Conversation_ApplicationId, refreshedDirectLineToken);
       dialogServiceConnector.connect();
     }, DIRECT_LINE_TOKEN_RENEWAL_INTERVAL);
   }

diff --git a/packages/directlinespeech/webpack.config.js b/packages/directlinespeech/webpack.config.js
@@ -15,7 +15,16 @@ let config = {
       filename: 'stats.json',
       transform: (_, opts) => JSON.stringify(opts.compiler.getStats().toJson({ chunkModules: true }), null, 2)
     })
-  ]
+  ],
+  resolve: {
+    alias: {
+      // TODO: [P1] #3575 Remove the following line when bumping to Speech SDK 1.14.0 or higher
+      'microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource': resolve(
+        __dirname,
+        'node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.browser/MicAudioSource.js'
+      )
+    }
+  }
 };
 
 // VSTS always emits uppercase environment variables.