Dart Annotation Code Generation in Flutter: Concepts and Implementation

This article introduces the annotation‑based code generation technique used in Dart/Flutter, comparing it with Java APT + JavaPoet.

Background : Flutter disables dart:mirrors, so reflection cannot be used. Developers must obtain class information at compile time and generate code based on annotations.

Simple Example : A basic Dart annotation is declared with const, making annotation definition simpler than Java. The example shows how to create an annotation, a generator that extends GeneratorForAnnotation, and a builder that triggers the generator.

build.yaml Configuration : The build.yaml file tells the build system which builders to run. Key fields include enable, generate_for (scans specific files or directories), options (passes custom key‑value data), run_before (controls execution order), and auto_apply (determines package scope). Images illustrate the YAML structure and the effect of different auto_apply settings.

source_gen Overview : The source_gen package wraps the build library and provides generators that process Dart AST elements. Important classes are Builder, SharedPartBuilder, LibraryBuilder, and PartBuilder. Generators receive a LibraryElement and a BuildStep to access source files and generate output.

Core Generator Method : generateForAnnotatedElement receives three parameters – the annotated Element, a ConstantReader for reading annotation values, and the BuildStep. The method can inspect class, function, or top‑level elements and emit code.

Code Generation Techniques : Three approaches are described – plain string concatenation, Mustache templates, and the code_builder library (similar to JavaPoet) which can construct expressions, statements, classes, etc., programmatically.

Comparison with Java : While Java APT uses runtime reflection and service registration, Dart relies on compile‑time builders and explicit build.yaml configuration. The article highlights the extra steps required in Flutter but also the flexibility of using templates and builders.

Conclusion : Dart annotation code generation offers a powerful alternative to Java APT, enabling generation of routing tables, template code, and dynamic proxies. It requires understanding of builders, YAML configuration, and libraries like source_gen, mustache, and code_builder to be effective.