Built Values for Dart

Introduction

Built Values provides:

Immutable value types;
EnumClass, classes that behave like enums;
JSON serialization.

Immutable collections are from built_collection.

Articles

Tutorials

Tools

The Json to Dart built_value class converter creates classes for you that correspond to a snippet of JSON

Examples

For an end to end example see the chat example, which was demoed at the Dart Summit 2016. The data model, used both client and server side, uses value types, enums and serialization from built_value.

Simple examples are here.

Since v5.2.0 codegen is triggered by running pub run build_runner build to do a one-off build or pub run build_runner watch to continuously watch your source and update the generated output when it changes. Note that you need a dev dependency on built_value_generator and build_runner. See the example pubspec.yaml.

If using Flutter, the equivalent command is flutter packages pub run build_runner build. Alternatively, put your built_value classes in a separate Dart package with no dependency on Flutter. You can then use built_value as normal.

If using a version before v5.2.0, codegen is triggered via either a build.dart to do a one-off build or a watch.dart to continuously watch your source and update generated output.

Value Types

Value types are, for our purposes, classes that are considered interchangeable if their fields have the same values.

Common examples include Date, Money and Url. Most code introduces its own value types. For example, every web app probably has some version of Account and User.

Value types are very commonly sent by RPC and/or stored for later retrieval.

The problems that led to the creation of the Built Value library have been discussed at great length in the context of AutoValue for Java.

In short: creating and maintaining value types by hand requires a lot of boilerplate. It's boring to write, and if you make a mistake, you very likely create a bug that's hard to track down.

Any solution for value types needs to allow them to participate in object oriented design. Date, for example, is the right place for code that does simple date manipulation.

AutoValue solves the problem for Java with code generation, and Built Values does the same for Dart. The boilerplate is generated for you, leaving you to specify which fields you need and to add code for the behaviour of the class.

Generating boilerplate for Value Types

Value types require a bit of boilerplate in order to connect it to generated code. Luckily, even this bit of boilerplate can be got automated using code snippets support in your favourite text editor. For example, in IntelliJ you can use following live template:

abstract class $CLASS_NAME$ implements Built<$CLASS_NAME$, $CLASS_NAME$Builder> {
  $CLASS_NAME$._();
  factory $CLASS_NAME$([void Function($CLASS_NAME$Builder) updates]) = _$$$CLASS_NAME$;
}

Using this template you would only have to manually enter a name of your data class name which is something that can't be automated.

Enum Class

Enum Classes provide classes with enum features.

Enums are very helpful in modelling the real world: whenever there are a small fixed set of options, an enum is a natural choice. For an object oriented design, though, enums need to be classes. Dart falls short here, so Enum Classes provide what's missing!

Design:

Constants have name and toString, can be used in switch statements, and are real classes that can hold code and implement interfaces
Generated values method that returns all the enum values in a BuiltSet (immutable set)
Generated valueOf method that takes a String

Serialization

Built Values comes with JSON serialization support which allows you to serialize a complete data model of Built Values, Enum Classes and Built Collections. The chat example shows how easy this makes building a full application with Dart on the server and client.

Here are the major features of the serialization support:

It fully supports object oriented design: any object model that you can design can be serialized, including full use of generics and interfaces. Some other libraries require concrete types or do not fully support generics.

It allows different object oriented models over the same data. For example, in a client server application, it's likely that the client and server want different functionality from their data model. So, they are allowed to have different classes that map to the same data. Most other libraries enforce a 1:1 mapping between classes and types on the wire.

It requires well behaved types. They must be immutable, can use interface but not concrete inheritance, must have predictable nullability, hashCode, equals and toString. In fact, they must be Enum Classes, Built Collections or Built Values. Some other libraries allow badly behaved types to be serialized.

It supports changes to the data model. Optional fields can be added or removed, and fields can be switched from optional to required, allowing your data model to evolve without breaking compatbility. Some other libraries break compatability on any change to any serializable class.

It's modular. Each endpoint can choose which classes to know about; for example, you can have multiple clients that each know about only a subset of the classes the server knows. Most other libraries are monolithic, requiring all endpoints to know all types.

It's multi language. Support will be come first for Dart, Java and Java/GWT. Many other libraries support a single language only.

It has first class support for validation via Built Values. An important part of a powerful data model is ensuring it's valid, so classes can make guarantees about what they can do. Other libraries also support validation but usually in a less prominent way.

It's pluggable. You can add serializers for your own types, and you can add plugins which run before and after all serializers. This could be used to interoperate with other tools or to add hand coded high performance serializers for specific classes. Some other libraries are not so extensible.

Common Usage

While full, compiled examples are available in example/lib, a common usage example is shown here. This example assumes that you are writing a client for a JSON API representing a person that looks like the following:

{
  id: 12345,
  age: 35,
  first_name: 'Jimmy',
  hobbies: ['jumping', 'basketball']
}

The corresponding dart class employing built_value might look like this. Note that it is using the @nullable annotation to indicate that the field does not have to be present on the response, as well as the @BuiltValueField annotation to map between the property name on the response and the name of the member variable in the Person class.

import 'package:built_value/built_value.dart';
import 'package:built_value/serializer.dart';
import 'package:built_collection/built_collection.dart';

part 'person.g.dart';

abstract class Person implements Built<Person, PersonBuilder> {
  static Serializer<Person> get serializer => _$personSerializer;

  // Can never be null.
  int get id;

  @nullable
  int get age;

  @nullable
  @BuiltValueField(wireName: 'first_name')
  String get firstName;

  @nullable
  BuiltList<String> get hobbies;

  Person._();
  factory Person([void Function(PersonBuilder) updates]) = _$Person;
}

FAQ

Should I check in and/or publish in the generated `.g.dart` files?

See the build_runner docs. You usually should not check in generated files, but you do need to publish them.

Features and bugs

Please file feature requests and bugs at the issue tracker.

yfer/built_value.dart