feat: support for android (#12)

* feat: support for android

* chore: revert unnecessary changes

android/CMakeLists.txt

@@ -8,13 +8,13 @@ set (CMAKE_CXX_STANDARD 17)
 find_package(ReactAndroid REQUIRED CONFIG)
 find_package(fbjni REQUIRED CONFIG)
 
-# TODO: Fix linking TFLite
-# find_library(
-#         TFLITE
-#         tensorflowlite_jni
-#         PATHS "./src/main/cpp/lib/tensorflow/jni/${ANDROID_ABI}"
-#         NO_DEFAULT_PATH
-#         NO_CMAKE_FIND_ROOT_PATH)
+find_library(
+        TFLITE
+        tensorflowlite_jni
+        PATHS "./src/main/cpp/lib/tensorflow/jni/${ANDROID_ABI}"
+        NO_DEFAULT_PATH
+        NO_CMAKE_FIND_ROOT_PATH
+)
 
 string(APPEND CMAKE_CXX_FLAGS " -DANDROID")
 
@@ -23,9 +23,8 @@ add_library(
   SHARED
   ../cpp/jsi/Promise.cpp
   ../cpp/jsi/TypedArray.cpp
-  # TODO: Uncomment this when tensorflow-lite C/C++ API can be successfully built/linked here
-  #../cpp/TensorflowPlugin.cpp
-  #../cpp/TensorHelpers.cpp
+  ../cpp/TensorflowPlugin.cpp
+  ../cpp/TensorHelpers.cpp
   src/main/cpp/Tflite.cpp
 )
 
@@ -35,7 +34,7 @@ target_include_directories(
   PRIVATE
   "../cpp"
   "src/main/cpp"
-  # "src/main/cpp/lib/tensorflow/headers/"
+  "src/main/cpp/lib/tensorflow/headers/"
   "${NODE_MODULES_DIR}/react-native/ReactCommon"
   "${NODE_MODULES_DIR}/react-native/ReactCommon/callinvoker"
   "${NODE_MODULES_DIR}/react-native/ReactAndroid/src/main/jni/react/turbomodule" # <-- CallInvokerHolder JNI wrapper
@@ -49,5 +48,5 @@ target_link_libraries(
   ReactAndroid::jsi               # <-- jsi.h
   ReactAndroid::reactnativejni    # <-- CallInvokerImpl
   fbjni::fbjni                    # <-- fbjni.h
-  # ${TFLITE}
+  ${TFLITE}
 )

android/build.gradle

@@ -123,9 +123,9 @@ dependencies {
   implementation "com.facebook.react:react-native:+"
 
   // Tensorflow Lite .aar (includes C API via prefabs)
-  implementation "org.tensorflow:tensorflow-lite:2.13.0"
-  extractHeaders("org.tensorflow:tensorflow-lite:2.13.0")
-  extractSO("org.tensorflow:tensorflow-lite:2.13.0")
+  implementation "org.tensorflow:tensorflow-lite:2.12.0"
+  extractHeaders("org.tensorflow:tensorflow-lite:2.12.0")
+  extractSO("org.tensorflow:tensorflow-lite:2.12.0")
 }
 
 task extractAARHeaders {

android/src/main/cpp/Tflite.cpp

@@ -4,17 +4,18 @@
 #include <jsi/jsi.h>
 #include <memory>
 
-// TODO: Uncomment this when tensorflow-lite C/C++ API can be successfully built/linked here
-// #include "TensorflowPlugin.h"
+#include "TensorflowPlugin.h"
 #include <ReactCommon/CallInvoker.h>
 #include <ReactCommon/CallInvokerHolder.h>
 
 namespace mrousavy {
 
+JavaVM *java_machine;
+
 using namespace facebook;
 using namespace facebook::jni;
 
-// Java Insaller
+// Java Installer
 struct TfliteModule : public jni::JavaClass<TfliteModule> {
 public:
   static constexpr auto kJavaDescriptor = "Lcom/tflite/TfliteModule;";
@@ -29,37 +30,44 @@ struct TfliteModule : public jni::JavaClass<TfliteModule> {
     }
     auto jsCallInvoker = jsCallInvokerHolder->cthis()->getCallInvoker();
 
-    // TODO: Uncomment this when tensorflow-lite C/C++ API can be successfully built/linked here
-    /*auto fetchByteDataFromUrl = [](std::string url) {
+    auto fetchByteDataFromUrl = [](std::string url) {
+
+      // Attaching Current Thread to JVM 
+      JNIEnv* env = nullptr;
+      int getEnvStat = java_machine->GetEnv((void**)&env, JNI_VERSION_1_6);
+      if (getEnvStat == JNI_EDETACHED) {
+          if (java_machine->AttachCurrentThread(&env, nullptr) != 0) {
+              throw std::runtime_error("Failed to attach thread to JVM");
+          }
+      }
+
       static const auto cls = javaClassStatic();
       static const auto method =
-    cls->getStaticMethod<jbyteArray(std::string)>("fetchByteDataFromUrl");
+        cls->getStaticMethod<jbyteArray(std::string)>("fetchByteDataFromUrl");
 
       auto byteData = method(cls, url);
+
+      // TODO: to review by someone experienced much more in C++ 
+      // Detaching current thread causes app crash with exception: 
+      // "Unable to retrieve jni environment. is the thread attached?"
+      // anyway, there is still a risk of memory leakage without calling the function below:
+
+      // java_machine->DetachCurrentThread();
+
       auto size = byteData->size();
       auto bytes = byteData->getRegion(0, size);
       void* data = malloc(size);
       memcpy(data, bytes.get(), size);
+
       return Buffer {
-          .data = data,
-          .size = size
+        .data = data,
+        .size = size
       };
+
     };
-     */
 
     try {
-      // TODO: Uncomment this when tensorflow-lite C/C++ API can be successfully built/linked here
-      // TensorflowPlugin::installToRuntime(*runtime, jsCallInvoker, fetchByteDataFromUrl);
-
-      // TODO: Remove this when tensorflow-lite C/C++ API can be successfully built/linked here
-      auto func = jsi::Function::createFromHostFunction(
-          *runtime, jsi::PropNameID::forAscii(*runtime, "__loadTensorflowModel"), 1,
-          [=](jsi::Runtime& runtime, const jsi::Value& thisValue, const jsi::Value* arguments,
-              size_t count) -> jsi::Value {
-            throw jsi::JSError(runtime, "react-native-fast-tflite is not yet supported on Android! "
-                                        "I couldn't manage to get TFLite to build for NDK/C++ :/");
-          });
-      runtime->global().setProperty(*runtime, "__loadTensorflowModel", func);
+      TensorflowPlugin::installToRuntime(*runtime, jsCallInvoker, fetchByteDataFromUrl);
     } catch (std::exception& exc) {
       return false;
     }
@@ -77,5 +85,6 @@ struct TfliteModule : public jni::JavaClass<TfliteModule> {
 } // namespace mrousavy
 
 JNIEXPORT jint JNICALL JNI_OnLoad(JavaVM* vm, void*) {
+  java_machine = vm;
   return facebook::jni::initialize(vm, [] { mrousavy::TfliteModule::registerNatives(); });
 }

example/src/App.tsx

@@ -1,8 +1,7 @@
 import * as React from 'react'
 
-import { StyleSheet, View, Text } from 'react-native'
+import { StyleSheet, View, Text, Platform } from 'react-native'
 import {
-  loadTensorflowModel,
   useTensorflowModel,
 } from 'react-native-fast-tflite'
 
@@ -11,7 +10,7 @@ export default function App() {
 
   const model = useTensorflowModel(
     require('../assets/object_detection_mobile_object_localizer_v1_1_default_1.tflite'),
-    'core-ml'
+    Platform.OS === 'ios' ? 'core-ml' : 'default'
   )
 
   React.useEffect(() => {