@momo-kits/camerakit 0.152.4 → 0.152.5

package/android/build.gradle

@@ -68,7 +68,7 @@ dependencies {
 
     implementation 'com.google.mlkit:barcode-scanning:17.3.0'
 
-    implementation 'jp.co.cyberagent.android:gpuimage:2.1.0'
+    implementation fileTree(dir: "libs", include: ["*.aar"])
     implementation 'com.google.zxing:core:3.5.1'
 }
 repositories {

package/ios/ReactNativeCameraKit/RealCamera.swift

@@ -10,6 +10,9 @@ import UIKit
 import CoreMotion
 import Vision
 
+// Global OCR queue (so OCR never blocks sessionQueue)
+let globalOCRQueue = DispatchQueue(label: "com.tesla.react-native-camera-kit.ocr", qos: .userInitiated)
+
 /*
  * Real camera implementation that uses AVFoundation
  */
@@ -21,16 +24,16 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
     private let session = AVCaptureSession()
     // Communicate with the session and other session objects on this queue.
     private let sessionQueue = DispatchQueue(label: "com.tesla.react-native-camera-kit")
-
+    
     // utilities
     private var setupResult: SetupResult = .notStarted
     private var isSessionRunning: Bool = false
     private var backgroundRecordingId: UIBackgroundTaskIdentifier = .invalid
-
+    
     private var videoDeviceInput: AVCaptureDeviceInput?
     private let photoOutput = AVCapturePhotoOutput()
     private let metadataOutput = AVCaptureMetadataOutput()
-
+    
     private var resizeMode: ResizeMode = .cover
     private var flashMode: FlashMode = .auto
     private var torchMode: TorchMode = .off
@@ -45,23 +48,25 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
     private var lastOnZoom: Double?
     private var zoom: Double?
     private var maxZoom: Double?
-
+    
     private var deviceOrientation = UIDeviceOrientation.unknown
     private var motionManager: CMMotionManager?
-
+    
     // KVO observation
     private var adjustingFocusObservation: NSKeyValueObservation?
 
     // Keep delegate objects in memory to avoid collecting them before photo capturing finishes
     private var inProgressPhotoCaptureDelegates = [Int64: PhotoCaptureDelegate]()
-
+    
     private var onTextRead: ((_ text: String) -> Void)?
     private let videoDataOutput = AVCaptureVideoDataOutput()
     private var textRequest: VNRecognizeTextRequest?
     private var textDetectionEnabled = false
     private var lastTextProcess = Date.distantPast
     private let textThrottle: TimeInterval = 0.35 // seconds
-
+    
+    private var zoomStartedAt: Double = 1.0
+    
     // MARK: - Lifecycle
 
     func cameraRemovedFromSuperview() {
@@ -87,15 +92,15 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
         UIDevice.current.endGeneratingDeviceOrientationNotifications()
 #endif
     }
-
+    
     deinit {
         removeObservers()
     }
-
+    
     // MARK: - Public
 
     func setup(cameraType: CameraType, supportedBarcodeType: [CodeFormat]) {
-
+        
         // Setup the capture session with priority on basic video preview first
         sessionQueue.async {
             self.setupResult = self.setupBasicVideoInput(cameraType: cameraType)
@@ -117,7 +122,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                 self.setVideoOrientationToInterfaceOrientation()
             }
         }
-
+        
         DispatchQueue.global(qos: .utility).async {
             self.initializeMotionManager()
         }
@@ -125,37 +130,36 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
     }
 
     // MARK: - Private optimization methods
-
+    
     private func setupBasicVideoInput(cameraType: CameraType) -> SetupResult {
         guard let videoDevice = self.getBestDevice(for: cameraType),
               let videoDeviceInput = try? AVCaptureDeviceInput(device: videoDevice) else {
             return .sessionConfigurationFailed
         }
-
+        
         session.beginConfiguration()
         defer { session.commitConfiguration() }
-
+        
         if session.canAddInput(videoDeviceInput) {
             session.addInput(videoDeviceInput)
             self.videoDeviceInput = videoDeviceInput
             self.resetZoom(forDevice: videoDevice)
             return .success
-        } else {
-            return .sessionConfigurationFailed
         }
+        return .sessionConfigurationFailed
     }
-
+    
     private func setupAdditionalOutputs(supportedBarcodeType: [CodeFormat]) {
         session.beginConfiguration()
         defer { session.commitConfiguration() }
-
+        
         // Add photo output
         if #available(iOS 13.0, *) {
-            if let maxPhotoQualityPrioritization {
+            if let maxPhotoQualityPrioritization = maxPhotoQualityPrioritization {
                 photoOutput.maxPhotoQualityPrioritization = maxPhotoQualityPrioritization.avQualityPrioritization
             }
         }
-
+        
         if session.canAddOutput(photoOutput) {
             session.addOutput(photoOutput)
 
@@ -165,12 +169,12 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                 }
             }
         }
-
+        
         // Add metadata output for barcode scanning
         if self.session.canAddOutput(metadataOutput) {
             self.session.addOutput(metadataOutput)
             metadataOutput.setMetadataObjectsDelegate(self, queue: DispatchQueue.main)
-
+            
             let availableTypes = self.metadataOutput.availableMetadataObjectTypes
             let filteredTypes = supportedBarcodeType
                 .map { $0.toAVMetadataObjectType() }
@@ -178,21 +182,39 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
 
             metadataOutput.metadataObjectTypes = filteredTypes
         }
-
+        
         // add for text detections
         if (textRequest != nil && self.session.canAddOutput(self.videoDataOutput)) {
             self.session.addOutput(self.videoDataOutput)
         }
     }
-
-    private var zoomStartedAt: Double = 1.0
+    
+    // MARK: - Pause / Resume non-essential outputs
+    private func pauseNonEssentialOutputs() {
+        videoDataOutput.setSampleBufferDelegate(nil, queue: nil)
+        metadataOutput.rectOfInterest = CGRect(x: 0, y: 0, width: 0, height: 0)
+    }
+    
+    private func resumeNonEssentialOutputs() {
+        if textDetectionEnabled {
+            videoDataOutput.setSampleBufferDelegate(self, queue: globalOCRQueue)
+        }
+        // Restore real rect of interest
+        if let scanner = scannerFrameSize, scanner != .zero {
+            metadataOutput.rectOfInterest =
+                cameraPreview.previewLayer.metadataOutputRectConverted(fromLayerRect: scanner)
+        } else {
+            metadataOutput.rectOfInterest = CGRect(x: 0, y: 0, width: 1, height: 1)
+        }
+    }
+    
     func zoomPinchStart() {
         sessionQueue.async {
             guard let videoDevice = self.videoDeviceInput?.device else { return }
             self.zoomStartedAt = videoDevice.videoZoomFactor
         }
     }
-
+    
     func zoomPinchChange(pinchScale: CGFloat) {
         guard !pinchScale.isNaN else { return }
 
@@ -201,7 +223,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
 
             let desiredZoomFactor = (self.zoomStartedAt / self.defaultZoomFactor(for: videoDevice)) * pinchScale
             let zoomForDevice = self.getValidZoom(forDevice: videoDevice, zoom: desiredZoomFactor)
-
+            
             if zoomForDevice != self.normalizedZoom(for: videoDevice) {
                 // Only trigger zoom changes if it's an uncontrolled component (zoom isn't manually set)
                 // otherwise it's likely to cause issues inf. loops
@@ -212,14 +234,14 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
     func update(maxZoom: Double?) {
         self.maxZoom = maxZoom
 
         // Re-update zoom value in case the max was increased
         self.update(zoom: self.zoom)
     }
-
+    
     func update(zoom: Double?) {
         sessionQueue.async {
             self.zoom = zoom
@@ -230,7 +252,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             self.setZoomFor(videoDevice, to: zoomForDevice)
         }
     }
-
+    
     /**
      `desiredZoom` can be nil when we want to notify what the zoom factor really is
      */
@@ -251,11 +273,11 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
         lastOnZoom = desiredOrCameraZoom
         self.onZoomCallback?(["zoom": desiredOrCameraZoom])
     }
-
+    
     func update(onZoom: RCTDirectEventBlock?) {
         self.onZoomCallback = onZoom
     }
-
+    
     func focus(at touchPoint: CGPoint, focusBehavior: FocusBehavior) {
         DispatchQueue.main.async {
             let devicePoint = self.cameraPreview.previewLayer.captureDevicePointConverted(fromLayerPoint: touchPoint)
@@ -270,7 +292,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                     self.resetFocus = nil
                     self.focusFinished = nil
                 }
-
+                
                 do {
                     try videoDevice.lockForConfiguration()
 
@@ -293,11 +315,11 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
     func update(onOrientationChange: RCTDirectEventBlock?) {
         self.onOrientationChange = onOrientationChange
     }
-
+    
     func update(torchMode: TorchMode) {
         sessionQueue.async {
             self.torchMode = torchMode
@@ -314,11 +336,11 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
     func update(flashMode: FlashMode) {
         self.flashMode = flashMode
     }
-
+    
     func update(maxPhotoQualityPrioritization: MaxPhotoQualityPrioritization?) {
         guard #available(iOS 13.0, *) else { return }
         guard maxPhotoQualityPrioritization != self.maxPhotoQualityPrioritization else { return }
@@ -329,7 +351,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             self.photoOutput.maxPhotoQualityPrioritization = maxPhotoQualityPrioritization?.avQualityPrioritization ?? .balanced
         }
     }
-
+    
     func update(cameraType: CameraType) {
         sessionQueue.async {
             if self.videoDeviceInput?.device.position == cameraType.avPosition {
@@ -343,14 +365,14 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                   let videoDeviceInput = try? AVCaptureDeviceInput(device: videoDevice) else {
                 return
             }
-
+            
             self.removeObservers()
             self.session.beginConfiguration()
             defer { self.session.commitConfiguration() }
 
             // Remove the existing device input first, since using the front and back camera simultaneously is not supported.
             self.session.removeInput(currentViewDeviceInput)
-
+            
             if self.session.canAddInput(videoDeviceInput) {
                 self.session.addInput(videoDeviceInput)
                 self.resetZoom(forDevice: videoDevice)
@@ -359,14 +381,14 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                 // If it fails, put back current camera
                 self.session.addInput(currentViewDeviceInput)
             }
-
+            
             self.addObservers()
 
             // We need to reapply the configuration after reloading the camera
             self.update(torchMode: self.torchMode)
         }
     }
-
+    
     func update(resizeMode: ResizeMode) {
         DispatchQueue.main.async {
             switch resizeMode {
@@ -377,24 +399,28 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
     func capturePicture(onWillCapture: @escaping () -> Void,
-                        onSuccess: @escaping (_ imageData: Data, _ thumbnailData: Data?, _ dimensions: CMVideoDimensions) -> Void,
-                        onError: @escaping (_ message: String) -> Void) {
+        onSuccess: @escaping (_ imageData: Data, _ thumbnailData: Data?, _ dimensions: CMVideoDimensions) -> Void,
+        onError: @escaping (_ message: String) -> Void) {
         /*
          Retrieve the video preview layer's video orientation on the main queue before
          entering the session queue. Do this to ensure that UI elements are accessed on
          the main thread and session configuration is done on the session queue.
          */
+        
+        // Pause OCR + barcode before capturing
+        self.pauseNonEssentialOutputs()
+        
         DispatchQueue.main.async { [weak self] in
             guard let self = self else {
                 onError("Camera was deallocated")
                 return
             }
-
+            
             let videoPreviewLayerOrientation =
             self.videoOrientation(from: self.deviceOrientation) ?? self.cameraPreview.previewLayer.connection?.videoOrientation
-
+            
             self.sessionQueue.async { [weak self] in
                 guard let self = self else {
                     onError("Camera was deallocated")
@@ -407,14 +433,14 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                     onError("Camera session is not running")
                     return
                 }
-
+                
                 // Ensure photo output has an active video connection
                 guard let photoOutputConnection = self.photoOutput.connection(with: .video) else {
                     print("Cannot capture photo: no video connection available")
                     onError("Camera connection is not available")
                     return
                 }
-
+                
                 // Verify the connection is active and enabled
                 guard photoOutputConnection.isActive && photoOutputConnection.isEnabled else {
                     print("Cannot capture photo: video connection is not active or enabled")
@@ -443,30 +469,33 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                         // Use weak self to prevent crash if camera is deallocated during capture
                         self?.inProgressPhotoCaptureDelegates[uniqueID] = nil
                         onSuccess(imageData, thumbnailData, dimensions)
+                        self?.resumeNonEssentialOutputs()
                     },
                     onCaptureError: { [weak self] uniqueID, errorMessage in
                         // Use weak self to prevent crash if camera is deallocated during capture
                         self?.inProgressPhotoCaptureDelegates[uniqueID] = nil
                         onError(errorMessage)
+                        self?.resumeNonEssentialOutputs()
                     }
                 )
-
+                
                 self.inProgressPhotoCaptureDelegates[photoCaptureDelegate.requestedPhotoSettings.uniqueID] = photoCaptureDelegate
                 self.photoOutput.capturePhoto(with: settings, delegate: photoCaptureDelegate)
             }
         }
     }
-
+    
+    // MARK: - Barcode scanning
     func isBarcodeScannerEnabled(_ isEnabled: Bool,
                                  supportedBarcodeTypes supportedBarcodeType: [CodeFormat],
                                  onBarcodeRead: ((_ barcode: String,_ codeFormat:CodeFormat) -> Void)?) {
         sessionQueue.async {
             self.onBarcodeRead = onBarcodeRead
+            
+            let availableTypes = self.metadataOutput.availableMetadataObjectTypes
             let newTypes: [AVMetadataObject.ObjectType]
             if isEnabled && onBarcodeRead != nil {
-                let availableTypes = self.metadataOutput.availableMetadataObjectTypes
-                newTypes = supportedBarcodeType.map { $0.toAVMetadataObjectType() }
-                    .filter { availableTypes.contains($0) }
+                newTypes = supportedBarcodeType.map { $0.toAVMetadataObjectType() }.filter { availableTypes.contains($0) }
             } else {
                 newTypes = []
             }
@@ -479,11 +508,11 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
     func update(barcodeFrameSize: CGSize?) {
         self.barcodeFrameSize = barcodeFrameSize
     }
-
+    
     func update(scannerFrameSize: CGRect?) {
         guard self.scannerFrameSize != scannerFrameSize else { return }
         self.sessionQueue.async {
@@ -491,7 +520,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             if !self.session.isRunning {
                 return
             }
-
+            
             DispatchQueue.main.async {
                 var visibleRect: CGRect?
                 if scannerFrameSize != nil && scannerFrameSize != .zero {
@@ -510,7 +539,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
             }
         }
     }
-
+    
 
     func isTextDetectionEnabled(_ isEnabled: Bool, onTextRead: ((String) -> Void)?) {
         sessionQueue.async {
@@ -525,37 +554,39 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
                 self.textRequest?.recognitionLanguages = ["en", "fr", "de", "es", "vi"]
                 self.textRequest?.recognitionLevel = .accurate
                 self.textRequest?.usesLanguageCorrection = false
-
+                
                 self.videoDataOutput.alwaysDiscardsLateVideoFrames = true
-                self.videoDataOutput.setSampleBufferDelegate(self, queue: self.sessionQueue)
+                self.videoDataOutput.setSampleBufferDelegate(self, queue: globalOCRQueue)
                 self.videoDataOutput.videoSettings = [
                     kCVPixelBufferPixelFormatTypeKey as String: kCVPixelFormatType_420YpCbCr8BiPlanarFullRange
                 ]
-
+                
             } else {
                 self.textRequest = nil
             }
         }
     }
-
+    
     // AVCaptureVideoDataOutputSampleBufferDelegate
     func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) {
         guard textDetectionEnabled, let request = textRequest else { return }
         let now = Date()
         if now.timeIntervalSince(lastTextProcess) < textThrottle { return }
         lastTextProcess = now
-
-        guard let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else { return }
-        var requestOptions: [VNImageOption: Any] = [:]
-
-        let handler = VNImageRequestHandler(cvPixelBuffer: pixelBuffer, orientation: .up, options: requestOptions)
-        do {
-            try handler.perform([request])
-        } catch {
-            // ignore OCR errors; don't crash the pipeline
+        
+        globalOCRQueue.async {
+            guard let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer) else { return }
+            var requestOptions: [VNImageOption: Any] = [:]
+
+            let handler = VNImageRequestHandler(cvPixelBuffer: pixelBuffer, orientation: .up, options: requestOptions)
+            do {
+                try handler.perform([request])
+            } catch {
+                // ignore OCR errors; don't crash the pipeline
+            }
         }
     }
-
+    
     // MARK: - AVCaptureMetadataOutputObjectsDelegate
 
     func metadataOutput(_ output: AVCaptureMetadataOutput, didOutput metadataObjects: [AVMetadataObject], from connection: AVCaptureConnection) {
@@ -569,7 +600,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
 
         onBarcodeRead?(codeStringValue,barcodeType)
     }
-
+    
     // MARK: - Private
 
     private func videoOrientation(from deviceOrientation: UIDeviceOrientation) -> AVCaptureVideoOrientation? {
@@ -588,7 +619,7 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
         @unknown default: return nil
         }
     }
-
+    
     private func videoOrientation(from interfaceOrientation: UIInterfaceOrientation) -> AVCaptureVideoOrientation {
         switch interfaceOrientation {
         case .portrait:
@@ -603,14 +634,14 @@ class RealCamera: NSObject, CameraProtocol, AVCaptureMetadataOutputObjectsDelega
         @unknown default: return .portrait
         }
     }
-
+    
     private func getBestDevice(for cameraType: CameraType) -> AVCaptureDevice? {
         if let device = AVCaptureDevice.default(.builtInWideAngleCamera, for: .video, position: cameraType.avPosition) {
             return device // single-lens/physical device
         }
         return nil
     }
-
+    
     private func defaultZoomFactor(for videoDevice: AVCaptureDevice) -> CGFloat {
         let fallback = 1.0
         guard #available(iOS 13.0, *) else { return fallback }

package/package.json

@@ -1,41 +1,41 @@
 {
-    "name": "@momo-kits/camerakit",
-    "version": "0.152.4",
-    "repository": {
-        "type": "git",
-        "url": "https://github.com/teslamotors/react-native-camera-kit.git"
-    },
-    "publishConfig": {
-        "registry": "https://registry.npmjs.org/"
-    },
-    "description": "A high performance, fully featured, rock solid camera library for React Native applications",
-    "nativePackage": true,
-    "scripts": {
-        "build": "echo",
-        "test": "jest",
-        "lint": "yarn eslint -c .eslintrc.js"
-    },
-    "main": "./src/index.ts",
-    "dependencies": {},
-    "license": "MIT",
-    "devDependencies": {
-        "react": "19.0.0",
-        "react-native": "0.80.1"
-    },
-    "peerDependencies": {
-        "react": "*",
-        "react-native": "*"
-    },
-    "engines": {
-        "node": ">=18"
-    },
-    "codegenConfig": {
-        "name": "rncamerakit_specs",
-        "type": "all",
-        "jsSrcsDir": "src/specs",
-        "android": {
-            "javaPackageName": "com.rncamerakit"
-        }
-    },
-    "packageManager": "yarn@1.22.22"
-}
+  "name": "@momo-kits/camerakit",
+  "version": "0.152.5",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/teslamotors/react-native-camera-kit.git"
+  },
+  "publishConfig": {
+    "registry": "https://registry.npmjs.org/"
+  },
+  "description": "A high performance, fully featured, rock solid camera library for React Native applications",
+  "nativePackage": true,
+  "scripts": {
+    "build": "echo",
+    "test": "jest",
+    "lint": "yarn eslint -c .eslintrc.js"
+  },
+  "main": "./src/index.ts",
+  "dependencies": {},
+  "license": "MIT",
+  "devDependencies": {
+    "react": "19.0.0",
+    "react-native": "0.80.1"
+  },
+  "peerDependencies": {
+    "react": "*",
+    "react-native": "*"
+  },
+  "engines": {
+    "node": ">=18"
+  },
+  "codegenConfig": {
+    "name": "rncamerakit_specs",
+    "type": "all",
+    "jsSrcsDir": "src/specs",
+    "android": {
+      "javaPackageName": "com.rncamerakit"
+    }
+  },
+  "packageManager": "yarn@1.22.22"
+}