Hybrid polylingual object model: an efficient and seamless integration of Java and native components on the Dalvik...

Area of Science:

• Computer Science
• Software Engineering

Background:

• Java Native Interface (JNI) in Android suffers from low efficiency and high complexity.
• Existing research has not simultaneously addressed both efficiency and complexity issues of JNI on Android.

Purpose of the Study:

• To propose a Hybrid Polylingual Object (HPO) model as an efficient and seamless alternative to JNI for Android development.
• To enable CAR objects to be accessed as Java objects within the Dalvik virtual machine.

Main Methods:

• Developed a Hybrid Polylingual Object (HPO) model for seamless integration of CAR-compliant components.
• Implemented a metadata injection mechanism for automatic mapping and reflection between CAR and Java objects.
• Extended the Dalvik virtual machine to create HPO-Dalvik, supporting HPO model features like lifespan management and garbage collection.

Main Results:

• The HPO model demonstrated lower overhead on the native side compared to standard JNI.
• Achieved better execution performance without the need for JNI bridging code.
• Provided automatic handling of lifespan management, garbage collection, and data type transformation for HPO objects.

Conclusions:

• The proposed HPO model is a viable and superior substitute for JNI in the Android platform.
• HPO model effectively reduces coding complexity and enhances execution efficiency for reusing native components in Android applications.