<== Chapter 5 -- Chapter 7 ==>

Chapter 6 - The JNI Header

We are now heading over to our jni directory. You have ~~two~~ three options for this section:

1. Copy and paste an example and tweak it
2. Have javah generate it all for you
3. Write it by hand

What is the JNI

• As already mentioned, the Java Native Interface is a way to convert your Java function calls to C++.
• The JNI file will need to make sure it imports the JNI files with #include <jni.h>.
• This file is not a class, but rather a way to be the bridge between the native C and Java code.
• We will want to make a static instance of our main class from the native code.

JNI Data Types

Because C++ and Java have different data types, we need to use JNI data types to bridge them together

Java primitive type	C type with JNI	Bytes
boolean	jboolean	1
byte	jbyte	1
char	jchar	2 (unsigned)
short	jshort	2 (signed)
int	jint	4
long	jlong	8
float	jfloat	4
double	jdouble	8

Java Reference Types	C type with JNI
java.lang.Class	jclass
java.lang.String	jstring
java.lang.Throwable	jthrowable
java.util.Array	jarray

• Note: there is a JNI array type for each of the different primitive types as well.

Bridging the function calls

• For this example image we called public static native void onGlSurfaceChanged(int width, int height); from the Java thread and will need to create a JNI function call to handle it.
• JNIEXPORT void JNICALL will
  • JNIEXPORT contains any compiler directives required to ensure that the given function is exported properly. On Android (and other linux-based systems), that'll be empty.
  • JNICALL contains any compiler directives required to ensure that the given function is treated with the proper calling convention. Probably empty on Android as well (it's __stdcall on w32).
  • void is the return type and this can be anything such as a jint, jdouble, etc.
• The JNI expects a certain function name.
  • Java_com_demo_tutorial_tango_tango_1ndk_1tutorial_TangoJNINative_onGlSurfaceChanged
  • This a combination of 4 parts
  • "Java"
  • Package Name (com.demo.tutorial.tango.tango_ndk_tutorial)
  • Java Class Name (TangoJNINative)
  • Function Name (onGLSurfaceChanged)
  • We need to now combine this all into the function name above by concatenating it with underscores
  • Note: if the package name contains an underscore we need to represent it as _1
• Lastly we now have to pass in the parameters (JNIEnv* env, jobject clazz, jint width, jint height)
  • JNIEvn* reference to JNI environment, which lets you access all the JNI functions.
  • jobject reference to "this" Java object.
    • jclass can be used if referencing a class public function
  • jint width, jint height these are the arguments passed in from the Java call.
• Now we can pass the values to the C++ function we have created with all the logic
  • app.OnSurfaceChanged(width, height);
  • We could do all the logic in here, but best practice to keep the JNI code from your actual business logic separate.

All together we have a bridging function call from the Java source to the native code

JNIEXPORT void JNICALL
Java_com_demo_tutorial_tango_tango_1ndk_1tutorial_TangoJNINative_onGlSurfaceChanged(
    JNIEnv*, jobject, jint width, jint height) {
  app.OnSurfaceChanged(width, height);
}

That wasn't so bad! It might be a lot of copy and pasting, but after a while you will get the flow of the JNI world.

<== Chapter 5 -- Chapter 7 ==>