Implementing Full Data Collection on Android Using Gradle Plugin, Transform API, and ASM

This article provides a comprehensive guide to implementing full data collection (full‑tracking) on Android using a custom Gradle plugin, the Transform API, and ASM bytecode instrumentation.

Full‑tracking concept : Full‑tracking (also called no‑code or automatic tracking) captures almost all user interactions without requiring developers to write explicit event‑logging code, though it generates large volumes of data.

Tracking classification : Three main approaches are discussed – code‑based tracking, visual tracking, and full‑tracking – with full‑tracking positioned between the other two in terms of complexity and coverage.

Implementation scheme : The solution adopts a static approach (Transform + ASM) because it operates during compilation, offering higher efficiency and better compatibility than dynamic methods.

Gradle plugin creation :

Create a new module named plugin and keep only the build.gradle and src directories.

Add required dependencies in build.gradle (Groovy DSL).

Write the plugin class AHAnalyticsPlugin.groovy under com.autohome.analytics.android.plugin .

Expose an extension autohomeAnalytics for configuration in the host app.

Define the plugin name in resources/META-INF/gradle-plugins/com.autohome.analytics.android.properties .

Publish the plugin to a remote or local Maven repository.

Using the plugin : Add the plugin dependency in the host project, then apply it in the app module’s build.gradle with apply plugin: "com.autohome.analytics.android" . Configuration options such as debug and disablePlug control its behavior.

Transform overview :

Register a custom AHAnalyticsTransform via registerTransform() on the Android Gradle AppExtension .

Implement required abstract methods: getName() , getInputTypes() , getScopes() , isIncremental() , and the core transform() method.

The transform processes both JarInput and DirectoryInput , handling added, changed, and removed files according to incremental settings.

ASM bytecode instrumentation :

ASM (a low‑level Java bytecode manipulation library) provides ClassReader , ClassVisitor , and ClassWriter for reading, modifying, and writing class files.

Core API (high performance) is used for the full‑tracking implementation.

Key classes: ClassReader reads the original bytecode, ClassVisitor traverses and modifies structures, and ClassWriter outputs the transformed bytecode.

Method-level injection is performed via a custom MethodVisitor (often extending AdviceAdapter ) that overrides onMethodExit() to insert logging statements before method return.

Example : The article shows how to inject a log statement into the onClick(View v) method of any class implementing View.OnClickListener using a custom AsmDemoClassVisitor and AsmDemoMethodVisitor . The injected code is Log.e("AsmDemo", var1.toString()); , demonstrating successful bytecode modification.

Summary : By combining a Gradle plugin, Transform API, and ASM, developers can achieve automatic, comprehensive data collection on Android without manual instrumentation, reducing development effort while providing rich analytics data.