fix: better getWordCount logic

plateaukao · Sep 22, 2024 · eb9de3b · eb9de3b
1 parent 5a5f9d2
commit eb9de3b
Show file tree

Hide file tree

Showing 4 changed files with 56 additions and 21 deletions.
diff --git a/app/src/main/java/info/plateaukao/einkbro/tts/ETts.kt b/app/src/main/java/info/plateaukao/einkbro/tts/ETts.kt
@@ -1,10 +1,12 @@
 package info.plateaukao.einkbro.tts
 
+import android.util.Log
 import info.plateaukao.einkbro.EinkBroApplication
 import info.plateaukao.einkbro.tts.entity.VoiceItem
 import okhttp3.Headers.Companion.toHeaders
 import okhttp3.OkHttpClient
 import okhttp3.Request
+import okhttp3.Response
 import okhttp3.WebSocket
 import okhttp3.WebSocketListener
 import okio.ByteString
@@ -154,12 +156,22 @@ class ETts {
 private open class TTSWebSocketListener : WebSocketListener() {
     var byteArray: ByteArray = ByteArray(0)
 
+    override fun onClosing(webSocket: WebSocket, code: Int, reason: String) {
+        super.onClosing(webSocket, code, reason)
+        webSocket.close(1000, null)
+    }
     override fun onMessage(webSocket: WebSocket, text: String) {
         if (text.contains("Path:turn.end")) {
             webSocket.close(1000, null)
         }
     }
 
+    override fun onFailure(webSocket: WebSocket, t: Throwable, response: Response?) {
+        super.onFailure(webSocket, t, response)
+        response?.close()
+        Log.d("TTSWebSocketListener", "onFailure: ${t.message}")
+    }
+
     override fun onMessage(webSocket: WebSocket, bytes: ByteString) =
         fixHeadHook(bytes.toByteArray())
 

diff --git a/app/src/main/java/info/plateaukao/einkbro/unit/HelperUnit.kt b/app/src/main/java/info/plateaukao/einkbro/unit/HelperUnit.kt
@@ -702,9 +702,18 @@ fun String.getWordCount(): Int {
     val trimmedInput = trim()
     if (trimmedInput.isEmpty()) return 0
 
+    // CJ
+    if (endsWith("。") || endsWith("？") || endsWith("！")) {
+        return trimmedInput.length
+    }
+    // korean
+    val hangulRegex = "[가-힣]+".toRegex() // Matches any Hangul syllable
+    // Find all matches and return the count
+    val hangulCount = hangulRegex.findAll(trimmedInput).sumOf { it.value.length }
+    if (hangulCount > 3) return hangulCount
+
     // Use regex to match words based on Unicode word boundaries
     val wordRegex = "\\p{L}+".toRegex()
-
     // Find all matches and return the count
     return wordRegex.findAll(trimmedInput).count()
 }
diff --git a/app/src/main/java/info/plateaukao/einkbro/view/dialog/compose/TtsSettingDialogFragment.kt b/app/src/main/java/info/plateaukao/einkbro/view/dialog/compose/TtsSettingDialogFragment.kt
@@ -59,6 +59,7 @@ class TtsSettingDialogFragment : ComposeDialogFragment() {
                 val ettsVoice = remember { mutableStateOf(config.ettsVoice) }
                 val gptVoice = remember { mutableStateOf(config.gptVoiceOption) }
                 val readProgress = ttsViewModel.readProgress.collectAsState()
+                val readingState = ttsViewModel.isReading.collectAsState()
 
                 Column(
                     modifier = Modifier
@@ -87,8 +88,8 @@ class TtsSettingDialogFragment : ComposeDialogFragment() {
                         onVoiceSelected = { config.ettsVoice = it; dismiss() },
                     )
                     TtsDialogButtonBar(
-                        isPlaying = ttsViewModel.isReading(),
-                        isVoiceReading = ttsViewModel.isVoicePlaying(),
+                        readingState = readingState.value,
+                        isVoicePlaying = ttsViewModel.isVoicePlaying(),
                         showSystemSetting = ttsType.value == TtsType.SYSTEM,
                         readProgress = readProgress.value,
                         gotoSettingAction = { IntentUnit.gotoSystemTtsSettings(requireActivity()) },
@@ -277,8 +278,8 @@ private fun MainTtsSettingDialog(
 
 @Composable
 fun TtsDialogButtonBar(
-    isPlaying: Boolean,
-    isVoiceReading: Boolean,
+    readingState: Boolean,
+    isVoicePlaying: Boolean,
     showSystemSetting: Boolean,
     readProgress: String,
     stopAction: () -> Unit,
@@ -307,7 +308,7 @@ fun TtsDialogButtonBar(
                     )
                 }
             } else {
-                if (isPlaying) {
+                if (readingState) {
                     Text(
                         readProgress,
                         modifier = Modifier.padding(horizontal = 8.dp),
@@ -329,7 +330,7 @@ fun TtsDialogButtonBar(
                         modifier = Modifier.wrapContentWidth()
                     ) {
                         Icon(
-                            if (isVoiceReading) Icons.Default.Pause else Icons.Default.PlayArrow,
+                            if (isVoicePlaying) Icons.Default.Pause else Icons.Default.PlayArrow,
                             "pause or resume",
                             tint = MaterialTheme.colors.onBackground
                         )

diff --git a/app/src/main/java/info/plateaukao/einkbro/viewmodel/TtsViewModel.kt b/app/src/main/java/info/plateaukao/einkbro/viewmodel/TtsViewModel.kt
@@ -37,9 +37,13 @@ class TtsViewModel : ViewModel(), KoinComponent {
     private val mediaPlayer by lazy { MediaPlayer() }
     private var byteArrayChannel: Channel<ByteArray>? = null
 
+    // for showing play controls state
     private val _speakingState = MutableStateFlow(false)
     val speakingState: StateFlow<Boolean> = _speakingState.asStateFlow()
 
+    private val _isReading = MutableStateFlow(false)
+    val isReading: StateFlow<Boolean> = _isReading.asStateFlow()
+
     private val _readProgress = MutableStateFlow("")
     val readProgress: StateFlow<String> = _readProgress.asStateFlow()
 
@@ -53,6 +57,8 @@ class TtsViewModel : ViewModel(), KoinComponent {
     private val articlesToBeRead: MutableList<String> = mutableListOf()
 
     fun readArticle(text: String) {
+        _isReading.value = true
+
         articlesToBeRead.add(text)
         if (isReading()) {
             return
@@ -94,25 +100,19 @@ class TtsViewModel : ViewModel(), KoinComponent {
             delay(2000)
         }
         _speakingState.value = false
+        _isReading.value = false
     }
 
     private suspend fun readByEngine(ttsType: TtsType, text: String) {
         byteArrayChannel = Channel(1)
+        val chunks = processedTextToChunks(text)
 
         viewModelScope.launch(Dispatchers.IO) {
             _speakingState.value = true
 
-            val chunks = processedTextToChunks(text)
             chunks.forEachIndexed { index, chunk ->
                 if (byteArrayChannel == null) return@launch
 
-                _readProgress.value =
-                    "${index + 1}/${chunks.size} " + if (articlesToBeRead.isNotEmpty()) {
-                        "(${articlesToBeRead.size})"
-                    } else {
-                        ""
-                    }
-
                 fetchSemaphore.withPermit {
                     Log.d("TtsViewModel", "tts sentence fetch: $chunk")
                     val byteArray = if (ttsType == TtsType.ETTS) {
@@ -122,9 +122,9 @@ class TtsViewModel : ViewModel(), KoinComponent {
                     }
 
                     if (byteArray != null) {
-                        Log.d("TtsViewModel", "tts sentence send: $chunk")
+                        Log.d("TtsViewModel", "tts sentence send ($index) : $chunk")
                         byteArrayChannel?.send(byteArray)
-                        Log.d("TtsViewModel", "tts sentence sent: $chunk")
+                        Log.d("TtsViewModel", "tts sentence sent ($index) : $chunk")
                     }
                 }
             }
@@ -134,22 +134,34 @@ class TtsViewModel : ViewModel(), KoinComponent {
         var index = 0
         for (byteArray in byteArrayChannel!!) {
             Log.d("TtsViewModel", "play audio $index")
+            _readProgress.value =
+                "${index + 1}/${chunks.size} " +
+                        if (articlesToBeRead.isNotEmpty()) {
+                            "(${articlesToBeRead.size})"
+                        } else {
+                            ""
+                        }
+
             playAudioByteArray(byteArray)
             index++
-            if (byteArrayChannel?.isClosedForSend == true && byteArrayChannel?.isEmpty == true
-            ) break
+            if (byteArrayChannel?.isClosedForSend == true &&
+                byteArrayChannel?.isEmpty == true) break
         }
 
         _speakingState.value = false
         byteArrayChannel = null
+
+        if (articlesToBeRead.isEmpty()) {
+            _isReading.value = false
+        }
     }
 
     private fun processedTextToChunks(text: String): MutableList<String> {
-        val processedText = text.replace("\\n", " ").replace("\\\"", "").replace("\\t", "")
+        val processedText = text.replace("\\n", " ").replace("\\\"", "").replace("\\t", "").replace("\\", "")
         val sentences = processedText.split("(?<=\\.)(?!\\d)|(?<=。)|(?<=？)|(?<=\\?)".toRegex())
         val chunks = sentences.fold(mutableListOf<String>()) { acc, sentence ->
             Log.d("TtsViewModel", "sentence: $sentence")
-            if (acc.isEmpty() || (acc.last() + sentence).getWordCount() > 80) {
+            if (acc.isEmpty() || (acc.last() + sentence).getWordCount() > 60) {
                 acc.add(sentence.trim())
             } else {
                 val last = acc.last()
@@ -195,6 +207,7 @@ class TtsViewModel : ViewModel(), KoinComponent {
         articlesToBeRead.clear()
 
         _speakingState.value = false
+        _isReading.value = false
     }
 
     fun isReading(): Boolean {