Decode and encode structured data type-safely

Utilities to handle structured data like values, identifiers, entities, data objects and data arrays. With JSON serialization support.

✨ New (2023-08): The version 1.0.0 requiring Dart 3. When Dart 2 is required, please use the older version attributes 0.8.2 supporting Dart 2.

Key features:

• Type-safe and null-safe accessors for data and value properties
• JSON decoding and encoding for data objects and arrays
• Utility functions to convert dynamic values to null-safe primitive value types
• DataObject: represent data like JSON Objects or properties as a map
• DataArray: represent data like JSON Arrays or properties as a list
• Identifier: an identifier, represented as String, int or BigInt
• Entity: a dynamic data entity with optional id and required properties

Introduction

Let's first consider some alternatives for Dart and Flutter developers to parse JSON and map data fields to a domain model.

To do this we need a sample domain model:

class Address {
  final String street;
  final String city;

  const Address({required this.street, required this.city});
}

class Person {
  final String name;
  final int age;
  final double? length;
  final Address address;
  final DateTime updatedUTC;

  const Person(
      {required this.name,
      required this.age,
      this.length,
      required this.address,
      required this.updatedUTC});
}

class PersonCollection {
  final Iterable<Person> persons;

  const PersonCollection({required this.persons});
}

Serializing JSON traditionally without "attributes"

There are multiple offically supported solutions to handle JSON. Of these solutions we first review a basic tradional way of implementing fromJson factories and toJson methods.

A sample with a simple domain model and JSON serialization support:

class Address {
  // ... showing only decoding part ...
  factory Address.fromJson(Map<String, dynamic> json) => Address(
        street: json['street'] as String,
        city: json['city'] as String,
      );
}

class Person {
  // ... showing only decoding part ...
  factory Person.fromJson(Map<String, dynamic> json) => Person(
      name: json['name'] as String,
      age: json['age'] as int,
      length: json['length'] as double?,
      address: Address.fromJson(json['address'] as Map<String, dynamic>),
      updatedUTC: DateTime.parse(json['updated'] as String).toUtc());
}

class PersonCollection {
  // ... showing only decoding part ...
  static PersonCollection fromJson(Iterable<dynamic> json) => PersonCollection(
        persons: json
            .map<Person>((dynamic element) =>
                Person.fromJson(element as Map<String, dynamic>))
            .toList(growable: false),
      );
}

As you can see this approach may require a lot of type casts (at least when you have disabled implicit-casts and implicit-dynamic on your analysis-options.yaml as you maybe should), null checks, value conversions and validations that can be error-prone. Also this solution couples your domain model to JSON encoding (be it a good or bad feature).

You can find the full sample code of this sample of NOT using the attributes package.

Type-safe and null-safe data objects to help

The attributes package provides type-safe and null-safe accessors to consume structured data like JSON.

The following sample replaces the Map<String, dynamic> type on serialization code with the DataObject class and the Iterable<dynamic> type with the DataArray class.

This allows data fields accessed more type safely, for example using DataObject instances to access JSON Object data:

• required fields read like data.getString('street') returning non-null values
• optional fields read like data.tryDouble('length') returning nullable values

Sample code shows differences best between this solution and a traditional way:

import 'package:attributes/data.dart';

class Address {
  // ... showing only decoding part ...
  static Address fromData(DataObject data) => Address(
        street: data.getString('street'),
        city: data.getString('city'),
      );
}

class Person {
  // ... showing only decoding part ...
  static Person fromData(DataObject data) => Person(
      name: data.getString('name'),
      age: data.getInt('age'),
      length: data.tryDouble('length'),
      address: Address.fromData(data.object('address')),
      updatedUTC: data.getTimeUTC('updated'));
}

class PersonCollection {
  // ... showing only decoding part ...
  static PersonCollection fromData(DataArray data) =>
      PersonCollection(persons: data.objectsToList(Person.fromData));
}

Please see the full sample code describing both decoding and encoding parts.

When comparing to the traditional way, we still need almost as much lines to be coded but reading data is much more safe when considering types and nullability.

This code is a bit cleaner too as a bonus!

Usage

Data objects

As already introduced dynamic property maps or JSON Objects are often represented as Map<String, dynamic> objects. Accessing dynamic data from such data structures a need for many checks or type conversions if you cannot be 100% sure that dynamic data is exactly what you are expecting.

However, for use cases when you just need to access dynamic data from some decoded JSON content without code generated classes or even specific model classes, then DataObject helps you on type and null safe access to property values.

Imports for examples below:

import 'package:attributes/attributes.dart';

At first, to create a data object, you can simply decode JSON data:

  // sample JSON data
  const sample = '''
    {
      "name": "Dash",
      "type": "mascot",
      "introduced": 2018,
      "fainted": "2021-03-03",
      "language": {
        "name": "Dart",
        "isNullSafe": true,
        "nullProperty": null,
        "nonNullProperty": "nonNull"
      },
      "toolkit": {
        "name": "Flutter",
        "fps": 60.0,
        "platforms": [ "iOS", "Android", "Windows", "macOS", "Linux", "Web"]
      }
    }
  ''';

  // Decode JSON data as a data object.
  final props = DataObject.decodeJson(sample);

DataObject has two main type of accessors for primitive values like:

/// Returns a non-null `String` value or throws `FormatException` if data is 
/// unavailable or cannot be converted to a `String`. 
String getString(String key);

/// Returns a nullable `String` value (null when data is unavailable or cannot
/// be converted to a `String`).
String? tryString(String key);

Similar accessors are available also for int, BigInt, num, double, bool, DateTime and Identifier values.

Some examples to access primitive values from a data object:

  // Access required null-safe properties using type-safe getXXX accessors.
  // These calls throw if a property is missing or does not convert to a type.
  final name = props.getString('name');
  final introduced = props.getInt('introduced');
  final fainted = props.getTimeUTC('fainted');

  // Access optional nullable properties using type-safe tryXXX accessors.
  // These calls never throw but return null if a property is missing or does
  // not convert to a type. An optional default value cab be given after
  // null-aware operator `??`.
  final web = props.tryString('web') ?? 'https://flutter.dev/dash';

  /// It's easy to check nullable values from accessors of optional properties.
  final users = props.tryBigInt('knownUsers');
  if (users != null) {
    print('The number of users ($users) is now known and it is huge!');
  } else {
    print('Data for known users not yet collected.');
  }

  // Hierarchical data is represented by sub data objects (JSON Objects) or
  // sub data arrays (JSON Arrays). For example data objects can be
  // accessed by "object" (required data) or "tryObject" (optional) accessors.
  final toolkit = props.object('toolkit');

  // Numeric values can be clamped to a range if value validation is needed.
  // Min and max limits are optional parameters when accessing num, int, double,
  // or BigInt.
  final fps = toolkit.getDouble('fps', min: 60.0, max: 120.0);

  // As already described dynamic data like JSON may also contain nulls or an
  // element for a certain key might not exist at all. Sometimes it's reasonable
  // just to check whether an value exists without trying to access it.
  final lang = props.object('language');
  if (lang.exists('nullProperty')) {
    // executes when exists and a value is either null or non-null
  }
  if (lang.existsNull('nullProperty')) {
    // executes when exists and a value is null
  }
  if (lang.existsNonNull('nonNullProperty')) {
    // executes when exists and a value is NOT null
  }

  // This is not a check but accesses a required boolean value.
  if (lang.getBool('isNullSafe')) {
    print('Dart is null-safe!');
  }

Data arrays

Dynamic property lists or JSON Arrays are often represented as List<dynamic> objects, at least when handling decoded JSON data.

Just like for data objects, it's possible to decode a data array from JSON using DataArray.decodeJson factory constructor.

However, below is a example to access an optional data object from a data object (toolkit) of the previous example:

  // Access an optional data array as a nullable variable.
  final platforms = toolkit.tryArray('platforms');

  // Trying to get an item by index in a data array (here nullable). Returns
  // null if not available, but in this example should return a String.
  final android = platforms?.tryString(1);

Data arrays also have similar type-safe accessors for nullable and non-null properties as data objects.

Identifiers

Identifiers could be represented as String or integer values. For dynamic data it's possible that primitive data types for identifiers are not known by code consuming such data.

The Identifier class allows creating an instance from a primitive value that could be either String or integer. Then a client can dynamically check a type, or convert an identifier to String, int or BigInt representation.

This is demonstrated below:

  // Identifiers can be based on String, int or BigInt values, here a String id.
  const dashId = Identifier.fromString('dash-2018');
  if(dashId.isInt) {
    final intId = dashId.asInt();
  } else {
    final stringId = dashId.asString();
  }

Identifier type can be checked using isString, isInteger, isInt and isBigInt properties. When expecting a specific type, asString, asInteger, asInt and asBigInt accessors are used. If not sure about a type, but you want to avoid format exceptions, then tryString, tryInteger, tryInt and tryBigInt returns a value or a null if not available as a requested type.

Entities

In the context of this package, an Entity represents a structured data entity that has an optional id as an Identifier object and contains associated property values in a DataObject instance.

An example how to create an entity (with dashId and props refering to variables from previous examples):

  // An entity contains required properties and an optional id.
  final dash = Entity.of(
    id: dashId,
    properties: props);

Value conversions

Conversions from JSON elements or other dynamic data structures can be converted to primitive values using utility functions provided by the package. These functions are also used by DataObject or DataArray when accessing primitive property values with methods described earlier like getString, tryString, getInt, etc.

For example there are converter functions:

• String toStringValue(Object? data);
  • Converts data to String or throws FormatException if cannot convert.
• int toIntValue(Object? data, {int? min, int? max})
  • Converts data to int or throws FormatException if cannot convert.
  • If provided min and max are used to clamp the returned value.

Similar functions are available for BigInt, double, num, bool, DateTime and Identifier.

These conversion functions try their best to convert to a desired type, not just type casting. For example the implementation for toDoubleValue explains this:

double toDoubleValue(Object? data, {double? min, double? max}) {
  if (data == null) throw NullValueException();
  double result;
  if (data is num) {
    result = data.toDouble();
  } else if (data is BigInt) {
    result = data.toDouble();
  } else if (data is String) {
    result = double.parse(data);
  } else if (data is bool) {
    result = data ? 1.0 : 0.0;
  } else {
    throw ConversionException(target: double, data: data);
  }
  if (min != null && result < min) {
    result = min;
  }
  if (max != null && result > max) {
    result = max;
  }
  return result;
}

Installing

The package requires at least Dart 3 from the stable channel.

In the pubspec.yaml of your project add the dependency:

dependencies:
  attributes: ^1.0.0

If you are not ready to use Dart 3, then you should check the older version attributes 0.8.2 supporting Dart 2.

Libraries

The package contains following mini-libraries:

Library	Description
collection	Base classes for collection implementations. Currently only the Counted interface.
data	Data objects and arrays representing generic data and with JSON integration.
data_ext	Same as data but contains also base implementation classes.
entity	Data entities consisting of a data object (properties) and an identifier.
exceptions	Exceptions specializing the standard FormatException.
values	Value accessors, conversions (dynamic objects to typed values) and helpers.

For example to access a mini library you should use an import like:

import 'package:attributes/data.dart';

To use all libraries of the package:

import 'package:attributes/attributes.dart';

Authors

This project is authored by Navibyte.

More information and other links are available at the dataflow repository from GitHub.

License

This project is licensed under the "BSD-3-Clause"-style license.

Please see the LICENSE.