react-native-sherpa-onnx 0.3.6 → 0.3.8

package/android/src/main/java/com/sherpaonnx/SherpaOnnxArchiveHelper.kt

@@ -4,6 +4,7 @@ import android.content.Context
 import android.util.Log
 import com.facebook.react.bridge.Arguments
 import com.facebook.react.bridge.Promise
+import java.util.concurrent.ConcurrentHashMap
 import java.util.concurrent.ExecutorService
 import java.util.concurrent.Executors
 import java.util.concurrent.atomic.AtomicBoolean
@@ -13,11 +14,12 @@ import java.util.concurrent.atomic.AtomicBoolean
  * This class delegates to C++ native implementation via JNI.
  */
 class SherpaOnnxArchiveHelper {
-  private val cancelRequested = AtomicBoolean(false)
-
   companion object {
-    /** Single-thread executor so extractions run off the React Native bridge thread and do not block listDownloadedModelsByCategory / RNFS. */
-    private val extractExecutor: ExecutorService = Executors.newSingleThreadExecutor()
+    /** Thread pool for extractions – allows up to 2 concurrent extractions while keeping them off the React Native bridge thread. */
+    private val extractExecutor: ExecutorService = Executors.newFixedThreadPool(2)
+
+    /** Per-source-path cancellation flags. Key = absolute source archive path. */
+    private val cancelFlags = ConcurrentHashMap<String, AtomicBoolean>()
 
     init {
       try {
@@ -29,15 +31,24 @@ class SherpaOnnxArchiveHelper {
   }
 
   fun cancelExtractTarBz2() {
-    cancelRequested.set(true)
+    // Cancel ALL ongoing extractions (legacy global cancel)
+    for (flag in cancelFlags.values) flag.set(true)
     nativeCancelExtract()
   }
 
   fun cancelExtractTarZst() {
-    cancelRequested.set(true)
+    // Cancel ALL ongoing extractions (legacy global cancel)
+    for (flag in cancelFlags.values) flag.set(true)
     nativeCancelExtract()
   }
 
+  /** Cancel a specific extraction identified by its source archive path. */
+  fun cancelExtractBySourcePath(sourcePath: String) {
+    // Only set the per-path flag; do not call nativeCancelExtract() since that is
+    // a global cancel that would also interrupt unrelated concurrent extractions.
+    cancelFlags[sourcePath]?.set(true)
+  }
+
   fun extractTarBz2(
     sourcePath: String,
     targetPath: String,
@@ -55,7 +66,9 @@ class SherpaOnnxArchiveHelper {
     }
 
     try {
-      cancelRequested.set(false)
+      // Register per-path cancel flag
+      val cancelFlag = AtomicBoolean(false)
+      cancelFlags[sourcePath] = cancelFlag
 
       // Create a progress callback object that JNI can call
       val progressCallback = object : Any() {
@@ -65,15 +78,23 @@ class SherpaOnnxArchiveHelper {
       }
 
       // Run extraction on a background thread so the React Native bridge thread is not blocked.
-      // Otherwise listDownloadedModelsByCategory (RNFS) and other native calls would wait until extraction finishes.
+      // The thread pool allows multiple extractions in parallel.
       extractExecutor.execute {
         try {
+          // Check per-path cancel flag before starting the native extraction.
+          if (cancelFlag.get()) {
+            resolveOnce(false, "Cancelled")
+            return@execute
+          }
           nativeExtractTarBz2(sourcePath, targetPath, force, progressCallback, promise)
         } catch (e: Exception) {
           resolveOnce(false, "Archive extraction error: ${e.message}")
+        } finally {
+          cancelFlags.remove(sourcePath)
         }
       }
     } catch (e: Exception) {
+      cancelFlags.remove(sourcePath)
       resolveOnce(false, "Archive extraction error: ${e.message}")
     }
   }
@@ -95,7 +116,9 @@ class SherpaOnnxArchiveHelper {
     }
 
     try {
-      cancelRequested.set(false)
+      val cancelFlag = AtomicBoolean(false)
+      cancelFlags[sourcePath] = cancelFlag
+
       val progressCallback = object : Any() {
         fun invoke(bytesExtracted: Long, totalBytes: Long, percent: Double) {
           onProgress(bytesExtracted, totalBytes, percent)
@@ -103,12 +126,20 @@ class SherpaOnnxArchiveHelper {
       }
       extractExecutor.execute {
         try {
+          // Check per-path cancel flag before starting the native extraction.
+          if (cancelFlag.get()) {
+            resolveOnce(false, "Cancelled")
+            return@execute
+          }
           nativeExtractTarZst(sourcePath, targetPath, force, progressCallback, promise)
         } catch (e: Exception) {
           resolveOnce(false, "Archive extraction error: ${e.message}")
+        } finally {
+          cancelFlags.remove(sourcePath)
         }
       }
     } catch (e: Exception) {
+      cancelFlags.remove(sourcePath)
       resolveOnce(false, "Archive extraction error: ${e.message}")
     }
   }
@@ -124,7 +155,6 @@ class SherpaOnnxArchiveHelper {
     if (BuildConfig.DEBUG) {
       Log.i("SherpaOnnx", "extractTarZstFromAsset assetPath=$assetPath targetPath=$targetPath")
     }
-    cancelRequested.set(false)
     val progressCallback = object : Any() {
       fun invoke(bytesExtracted: Long, totalBytes: Long, percent: Double) {
         onProgress(bytesExtracted, totalBytes, percent)

package/android/src/main/java/com/sherpaonnx/SherpaOnnxModule.kt

@@ -8,6 +8,7 @@ import com.facebook.react.bridge.ReadableMap
 import com.facebook.react.bridge.Arguments
 import com.facebook.react.module.annotations.ReactModule
 import com.facebook.react.modules.core.DeviceEventManagerModule
+import com.k2fsa.sherpa.onnx.WaveReader
 
 @ReactModule(name = SherpaOnnxModule.NAME)
 class SherpaOnnxModule(reactContext: ReactApplicationContext) :
@@ -314,6 +315,11 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
     promise.resolve(null)
   }
 
+  override fun cancelExtractBySourcePath(sourcePath: String, promise: Promise) {
+    archiveHelper.cancelExtractBySourcePath(sourcePath)
+    promise.resolve(null)
+  }
+
   override fun computeFileSha256(filePath: String, promise: Promise) {
     archiveHelper.computeFileSha256(filePath, promise)
   }
@@ -452,6 +458,7 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
     val provider = if (options.hasKey("provider")) options.getString("provider") else null
     val ruleFsts = if (options.hasKey("ruleFsts")) options.getString("ruleFsts") else null
     val ruleFars = if (options.hasKey("ruleFars")) options.getString("ruleFars") else null
+    val dither = if (options.hasKey("dither")) options.getDouble("dither") else null
     val blankPenalty = if (options.hasKey("blankPenalty")) options.getDouble("blankPenalty") else null
     val debug = if (options.hasKey("debug")) options.getBoolean("debug") else null
     val rule1MustContainNonSilence = if (options.hasKey("rule1MustContainNonSilence")) options.getBoolean("rule1MustContainNonSilence") else null
@@ -476,6 +483,7 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
       provider,
       ruleFsts,
       ruleFars,
+      dither,
       blankPenalty,
       debug,
       rule1MustContainNonSilence,
@@ -712,6 +720,72 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
     }
   }
 
+  /**
+   * Decode audio to mono float samples (approx. [-1, 1]) and effective sample rate.
+   * Same path/URI handling as [convertAudioToFormat]. WAV may use [WaveReader] when no resample is requested.
+   */
+  override fun decodeAudioFileToFloatSamples(inputPath: String, targetSampleRateHz: Double?, promise: Promise) {
+    var tmpFile: java.io.File? = null
+    try {
+      val targetHz = (targetSampleRateHz ?: 0.0).toInt()
+      if (targetHz < 0) {
+        promise.reject("DECODE_ERROR", "targetSampleRateHz must be >= 0")
+        return
+      }
+      val (pathToUse, tmp) = resolveInputForConvert(inputPath)
+      tmpFile = tmp
+
+      if (pathToUse.endsWith(".wav", ignoreCase = true)) {
+        try {
+          val wave = WaveReader.readWave(pathToUse)
+          val s = wave.samples
+          if (s != null && s.isNotEmpty() && wave.sampleRate > 0 && (targetHz == 0 || targetHz == wave.sampleRate)) {
+            val map = Arguments.createMap()
+            val arr = Arguments.createArray()
+            for (i in s.indices) {
+              arr.pushDouble(s[i].toDouble())
+            }
+            map.putArray("samples", arr)
+            map.putInt("sampleRate", wave.sampleRate)
+            promise.resolve(map)
+            return
+          }
+        } catch (_: Throwable) {
+          // Fall through to FFmpeg/native path (e.g. odd WAV or resample requested).
+        }
+      }
+
+      val result = Companion.nativeDecodeAudioFileToFloatSamples(pathToUse, targetHz)
+      if (result.size == 1 && result[0] is String) {
+        promise.reject("DECODE_ERROR", result[0] as String)
+        return
+      }
+      if (result.size != 2 || result[0] !is FloatArray) {
+        promise.reject("DECODE_ERROR", "Unexpected native decode result")
+        return
+      }
+      val floats = result[0] as FloatArray
+      val rateObj = result.getOrNull(1) as? Number ?: run {
+        promise.reject("DECODE_ERROR", "Unexpected sample rate in native decode result")
+        return
+      }
+      val sr = rateObj.toInt()
+      val map = Arguments.createMap()
+      val arr = Arguments.createArray()
+      for (i in floats.indices) {
+        arr.pushDouble(floats[i].toDouble())
+      }
+      map.putArray("samples", arr)
+      map.putInt("sampleRate", sr)
+      promise.resolve(map)
+    } catch (e: Exception) {
+      android.util.Log.e(NAME, "DECODE_EXCEPTION: ${e.message}", e)
+      promise.reject("DECODE_EXCEPTION", e.message ?: "Failed to decode audio", e)
+    } finally {
+      tmpFile?.delete()
+    }
+  }
+
   // ==================== TTS Methods ====================
 
   /**
@@ -1057,6 +1131,25 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
     }
   }
 
+  override fun readAssetFileAsUtf8(assetPath: String, promise: Promise) {
+    // Validate assetPath to prevent path traversal: reject paths containing
+    // "..", starting with "/" or "\", or containing backslashes.
+    if (assetPath.contains("..") ||
+        assetPath.startsWith("/") ||
+        assetPath.startsWith("\\") ||
+        assetPath.contains("\\")) {
+      promise.reject("ASSET_READ_ERROR", "Invalid asset path: $assetPath")
+      return
+    }
+    try {
+      val content = reactApplicationContext.assets.open(assetPath).bufferedReader().use { it.readText() }
+      promise.resolve(content)
+    } catch (e: Exception) {
+      android.util.Log.e(NAME, "Failed to read asset $assetPath: ${e.message}", e)
+      promise.reject("ASSET_READ_ERROR", "Failed to read asset $assetPath: ${e.message}", e)
+    }
+  }
+
   companion object {
     const val NAME = "SherpaOnnx"
 
@@ -1099,5 +1192,11 @@ class SherpaOnnxModule(reactContext: ReactApplicationContext) :
     /** Convert any supported audio file to WAV 16 kHz mono 16-bit PCM. Returns empty string on success, error message otherwise. Requires FFmpeg prebuilts. */
     @JvmStatic
     private external fun nativeConvertAudioToWav16k(inputPath: String, outputPath: String): String
+
+    /**
+     * On success: [FloatArray samples, Integer sampleRate]. On error: [String message].
+     */
+    @JvmStatic
+    private external fun nativeDecodeAudioFileToFloatSamples(inputPath: String, targetSampleRateHz: Int): Array<Any>
   }
 }

package/android/src/main/java/com/sherpaonnx/SherpaOnnxOnlineSttHelper.kt

@@ -132,6 +132,7 @@ internal class SherpaOnnxOnlineSttHelper(
     provider: String?,
     ruleFsts: String?,
     ruleFars: String?,
+    dither: Float?,
     blankPenalty: Float?,
     debug: Boolean?,
     rule1MustContainNonSilence: Boolean?,
@@ -233,7 +234,7 @@ internal class SherpaOnnxOnlineSttHelper(
     }
 
     return OnlineRecognizerConfig(
-      featConfig = FeatureConfig(sampleRate = 16000, featureDim = 80, dither = 0f),
+      featConfig = FeatureConfig(sampleRate = 16000, featureDim = 80, dither = dither ?: 0f),
       modelConfig = modelConfig,
       endpointConfig = endpointConfig,
       enableEndpoint = enableEndpoint,
@@ -260,6 +261,7 @@ internal class SherpaOnnxOnlineSttHelper(
     provider: String?,
     ruleFsts: String?,
     ruleFars: String?,
+    dither: Double?,
     blankPenalty: Double?,
     debug: Boolean?,
     rule1MustContainNonSilence: Boolean?,
@@ -286,6 +288,7 @@ internal class SherpaOnnxOnlineSttHelper(
         provider = provider,
         ruleFsts = ruleFsts,
         ruleFars = ruleFars,
+        dither = dither?.toFloat(),
         blankPenalty = blankPenalty?.toFloat(),
         debug = debug,
         rule1MustContainNonSilence = rule1MustContainNonSilence,