From 0e4c873e52ca576d32ec803e3c547db5219a2a52 Mon Sep 17 00:00:00 2001
From: Maximilian Krause <hello@max-krause.com>
Date: Thu, 21 Nov 2024 16:12:33 +0100
Subject: [PATCH] feat: some stability & alignment improvements

---
 .../src/main/java/com/voicekit/VoiceError.kt  |  4 +++
 .../main/java/com/voicekit/VoiceKitService.kt | 16 ++++++++--
 example/src/App.tsx                           |  3 +-
 ios/Core/VoiceKitService.swift                | 31 +++++++++++++++----
 ios/Utils/VoiceError.swift                    |  4 +++
 src/types/native.ts                           |  7 +++++
 6 files changed, 56 insertions(+), 9 deletions(-)

diff --git a/android/src/main/java/com/voicekit/VoiceError.kt b/android/src/main/java/com/voicekit/VoiceError.kt
index 82bd4e9..a1827f7 100644
--- a/android/src/main/java/com/voicekit/VoiceError.kt
+++ b/android/src/main/java/com/voicekit/VoiceError.kt
@@ -28,6 +28,10 @@ sealed class VoiceError(
     "ERR_PERMISSION_NOT_DETERMINED",
     "Speech recognition permission was not yet determined"
   )
+  object InvalidState : VoiceError(
+    "ERR_INVALID_STATE",
+    "Invalid state, cannot perform action"
+  )
   class Unknown(message: String = "An unknown error occurred") : VoiceError(
     "ERR_UNKNOWN",
     message
diff --git a/android/src/main/java/com/voicekit/VoiceKitService.kt b/android/src/main/java/com/voicekit/VoiceKitService.kt
index 61ca1f6..767d11b 100644
--- a/android/src/main/java/com/voicekit/VoiceKitService.kt
+++ b/android/src/main/java/com/voicekit/VoiceKitService.kt
@@ -59,13 +59,19 @@ class VoiceKitService(private val context: ReactApplicationContext) {
     audioManager?.setStreamVolume(AudioManager.STREAM_NOTIFICATION, previousNotificationVolume, AudioManager.FLAG_ALLOW_RINGER_MODES)
   }
 
-  fun startListening(options: ReadableMap, skipMuteBeep: Boolean = false): Boolean {
+  fun startListening(options: ReadableMap, skipMuteBeep: Boolean = false) {
     val currentActivity = context.currentActivity
     if (currentActivity == null) {
       Log.e(TAG, "Activity is null")
       throw VoiceError.Unknown("Activity is null")
     }
 
+    if (isListening) {
+      Log.w(TAG, "Already listening, aborting startListening")
+      sendEvent("RNVoiceKit.error", VoiceError.InvalidState)
+      return
+    }
+
     this.options = options
 
     // TODO: We currently don't wait for the permission to be granted, but we should
@@ -114,10 +120,16 @@ class VoiceKitService(private val context: ReactApplicationContext) {
 
     isListening = true
 
-    return true
+    return
   }
 
   fun stopListening() {
+    if (!isListening) {
+      Log.w(TAG, "Not listening, aborting stopListening")
+      sendEvent("RNVoiceKit.error", VoiceError.InvalidState)
+      return
+    }
+
     lastResultTimer?.removeCallbacksAndMessages(null)
     lastResultTimer = null
     lastTranscription = null
diff --git a/example/src/App.tsx b/example/src/App.tsx
index 992d347..fdf393f 100644
--- a/example/src/App.tsx
+++ b/example/src/App.tsx
@@ -19,8 +19,9 @@ export default function App() {
             console.error('Error starting listening', error, error instanceof VoiceError ? error.details : null);
           });
         }}
+        disabled={!available || listening}
       />
-      <Button title="Stop Listening" onPress={() => stopListening()} />
+      <Button title="Stop Listening" onPress={() => stopListening()} disabled={!listening} />
       <Button title="Reset Transcript" onPress={() => resetTranscript()} />
       <Text>Transcript: {transcript}</Text>
     </View>
diff --git a/ios/Core/VoiceKitService.swift b/ios/Core/VoiceKitService.swift
index 9b0f9bc..5370bcc 100644
--- a/ios/Core/VoiceKitService.swift
+++ b/ios/Core/VoiceKitService.swift
@@ -42,6 +42,12 @@ class VoiceKitService: NSObject, SFSpeechRecognizerDelegate {
     Logger.log(level: .info, message: "Starting recording")
     Logger.log(level: .debug, message: "Options: \(options)")
 
+    if isListening {
+      Logger.log(level: .warning, message: "Already listening, aborting startRecording")
+      delegate?.onError(.invalidState)
+      return
+    }
+
     // Cancel any ongoing tasks
     recognitionTask?.cancel()
     recognitionTask = nil
@@ -64,12 +70,22 @@ class VoiceKitService: NSObject, SFSpeechRecognizerDelegate {
     recognitionTask = speechRecognizer?.recognitionTask(with: recognitionRequest!) { [weak self] result, error in
       guard let self else { return }
 
-      if let result {
-        Logger.log(level: .debug, message: "SpeechRecognizerResult received: \(result)")
+      if error != nil {
+        handleError(error)
+      }
 
+      if let result {
         // Store the latest transcription
         lastTranscription = result.bestTranscription.formattedString
 
+        // if trimmed lastTranscription is empty, stop and dont do anything
+        if lastTranscription?.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty ?? true {
+          Logger.log(level: .debug, message: "Last transcription is empty, stopping")
+          return
+        }
+
+        Logger.log(level: .debug, message: "SpeechRecognizerResult received: \(lastTranscription)")
+
         // Emit partial results
         delegate?.onPartialResult(lastTranscription ?? "")
 
@@ -88,10 +104,6 @@ class VoiceKitService: NSObject, SFSpeechRecognizerDelegate {
           }
         }
       }
-
-      if error != nil {
-        handleError(error)
-      }
     }
 
     // Configure audio engine
@@ -128,6 +140,13 @@ class VoiceKitService: NSObject, SFSpeechRecognizerDelegate {
 
   func stopRecording() {
     Logger.log(level: .info, message: "Stopping recording")
+
+    if !isListening {
+      Logger.log(level: .warning, message: "Not listening, aborting stopRecording")
+      delegate?.onError(.invalidState)
+      return
+    }
+
     audioEngine.stop()
     audioEngine.inputNode.removeTap(onBus: 0)
     recognitionRequest?.endAudio()
diff --git a/ios/Utils/VoiceError.swift b/ios/Utils/VoiceError.swift
index 75dd47d..0a93272 100644
--- a/ios/Utils/VoiceError.swift
+++ b/ios/Utils/VoiceError.swift
@@ -15,6 +15,7 @@ enum VoiceError: Error {
   case permissionDenied
   case permissionRestricted
   case permissionNotDetermined
+  case invalidState
   case unknown(message: String = "An unknown error occurred")
 
   var code: String {
@@ -25,6 +26,7 @@ enum VoiceError: Error {
     case .permissionDenied: "ERR_PERMISSION_DENIED"
     case .permissionRestricted: "ERR_PERMISSION_RESTRICTED"
     case .permissionNotDetermined: "ERR_PERMISSION_NOT_DETERMINED"
+    case .invalidState: "ERR_INVALID_STATE"
     case .unknown: "ERR_UNKNOWN"
     }
   }
@@ -43,6 +45,8 @@ enum VoiceError: Error {
       "Speech recognition is restricted on this device"
     case .permissionNotDetermined:
       "Speech recognition permission was not yet determined"
+    case .invalidState:
+      "Invalid state, cannot perform action"
     case let .unknown(message):
       message
     }
diff --git a/src/types/native.ts b/src/types/native.ts
index e2dd03a..efa1c63 100644
--- a/src/types/native.ts
+++ b/src/types/native.ts
@@ -1,6 +1,13 @@
 import type { VoiceStartListeningOptions } from '.';
 
 export enum VoiceErrorCode {
+  SPEECH_RECOGNIZER_NOT_AVAILABLE = 'ERR_SPEECH_RECOGNIZER_NOT_AVAILABLE',
+  RECORDING_START_FAILED = 'ERR_RECORDING_START_FAILED',
+  RECOGNITION_FAILED = 'ERR_RECOGNITION_FAILED',
+  PERMISSION_DENIED = 'ERR_PERMISSION_DENIED',
+  PERMISSION_RESTRICTED = 'ERR_PERMISSION_RESTRICTED',
+  PERMISSION_NOT_DETERMINED = 'ERR_PERMISSION_NOT_DETERMINED',
+  INVALID_STATE = 'ERR_INVALID_STATE',
   UNKNOWN = 'ERR_UNKNOWN',
 }