Boost Android Local Search Speed with JavaScriptCore and JNI

Introduction

The article shows how to use JavaScriptCore (JSC) together with the Java Native Interface (JNI) to accelerate an Android local‑search feature that previously took about 7 seconds to process 8,000 songs.

Background

The existing implementation runs dynamic JavaScript code on the Java side, and the overall latency is split roughly 50 % in database/data processing and 50 % in JavaScript engine data transfer. The focus of the solution is to reduce the JavaScript‑engine overhead.

Solution Overview

The proposed architecture keeps the search algorithm in JavaScript (so it can be updated dynamically) while executing the script in a native JSC instance accessed from Java through JNI. The data flow is:

Load the JSC shared library (reusing the React‑Native .so).

Implement core logic in C++ that calls JSC and exposes a native .so for the Java layer.

The Java layer invokes the native library to run the script and retrieve results.

A flow diagram (see image) illustrates the interaction between Java, JNI, and JSC.