JNI Basics - 1

Introduction

In this tutorial, I will try to explain how to call native functions which are written in C or C++ from Java, using JNI.

What is JNI

JNI is an acronym of Java Native Interface. Using JNI, we can call functions which are written in other languages from Java. Advantages and disadvantages of this method are as follows:

Advantages:

1. You can use your existing library which was previously written in another language.
2. You can call Windows API functions.
3. For the sake of execution speed.
4. You can call API functions of some server product which is in C or C++, from Java client.

Disadvantages:

1. You can’t say "write once run anywhere".
2. Difficult to debug runtime error in native code.
3. Potential security risk.
4. You can’t call it from Applets.

Here we will try to call the MFC function AfxMessageBox() from Java. Our function in C++ will take no arguments and will not return any value.

//Test1.java

class Test1
{
    static
    {
        System.loadLibrary("TestDll");
    }
    public static void main(String ar[])
    {
        System.out.println("Hello world from Java");
        Test1 t=new Test1();
        t.inDll();
    }
    public native void inDll();
}

In static block, we are loading a DLL named TestDll.dll which should be placed in <WINDIR>/SYSTEM32 directory. I am not saying you cannot put this DLL in other directories, you can, but then you will have to register that path in JVM. Here you can call another method as well, instead of loadLibrary().

System.load(“c://winnt//system32//TestDll.dll”);

I think load() method is better as far as usage is concerned because you can easily place your DLL in any directory.

public native void inDll();

As you can guess, this native keyword specifies that inDll() is a native function which is implemented in the library. Also noticeable, there is no implementation. When we call this function in our code, JVM will look for the function in the loaded DLL (due to static block, DLL will be loaded immediately after the loading of the class) and then call that function. Otherwise it will throw java.lang.UnsatisfiedLinkError exception.

Compile this class normally. Now we will need to generate header file of this class.

You can use javah.exe utility which is included with JDK.

Javah -jni Test1

Header file is generated in current directory with the name Test1.h. Keep this header file safe as we will use this soon.

Now we are going to write a DLL which will contain the implementation of our native function. I have chosen MFC AppWizard (DLL) in project and named my project: TestDll.

Include your class header file which was generated previously using javah.exe in TestDll.h.

Copy following line from Test1.h and paste it at the end of TestDll.h.

JNIEXPORT void JNICALL Java_Test1_inDll(JNIEnv *, jobject);

Change this line to:

JNIEXPORT void JNICALL Java_Test1_inDll(JNIEnv *env, jobject obj)
{
    AfxMessageBox("Hello World from dll");
}

Compile this DLL, you will get an error from compiler because compiler was unable to find jni.h file which is included in Test1.h. So include the following two directories in IDE settings.

• <Drive>:\<JDK Directory>\include
• <Drive>:\<JDK Directory>\include\win32

Go to Project/Settings, on C/C++ tab, go to Project option, insert following lines at the end:

/I "<Drive>:\<JDK Directory>\include "
/I "<Drive>:\<JDK Directory>\include\win32"

Now compile and generate your DLL, put newly created DLL in system32 directory and then run your class file. It should work properly. If you receive an error like UnsatisfiedLinkError, then recheck your procedure. If you are using any IDE, then try using command prompt. Hope it will work.

In this tutorial, I have tried to explain the basics of JNI. I will write a second tutorial of the series soon in which I will explain how to pass parameters to native functions and return something, along with a practical example.

Waiting for your comments and questions.