Functional programming in Dart and Flutter

All the main functional programming types and patterns fully documented, tested, and with examples

GitHub: SandroMaglione Twitter: SandroMaglione

Introduction

fpdart is fully documented. You do not need to have any previous experience with functional programming to start using fpdart. Give it a try!

fpdart is inspired by fp-ts, cats, and dartz.

Follow my Twitter for updates, or subscribe to the newsletter


📖 Learn functional programming and fpdart

Would you like to know more about functional programming, fpdart, and how to use the package?

📚 Collection of tutorials on fpdart

Are you new to fpdart and functional programming?

👨‍💻 Getting started with fpdart complete guide

Interested in what fpdart is and how it came to be?

🚶 Full history of fpdart and functional programming in dart

✍️ Blog posts and tutorials

🧑‍🏫 Getting started with functional programming

💻 Installation

# pubspec.yaml
dependencies:
  fpdart: ^1.1.0

✨ Examples

fpdart + riverpod

Step by step course on how to build a safe, maintainable, and testable Flutter app using fpdart and riverpod.

Pokeapi

Flutter app that lets you search and view your favorite Pokemon:

  • API request
  • Response validation
  • JSON conversion
  • State management (riverpod)

Open Meteo API

Re-implementation using fpdart and functional programming of the Open Meteo API from the flutter_weather app example in the bloc package.

A 2 parts series explains step by step the Open Meteo API code:

Read/Write local file

Example of how to read and write a local file using functional programming.

Manage imports

Using fpdart with other libraries and noticing naming conflicts? Learn how to rename the classes that conflict with other SDK or third-party packages.

Option

Used when a return value can be missing.

For example, when parsing a String to int, since not all String can be converted to int

/// Create an instance of [Some]
final option = Option.of(10);

/// Create an instance of [None]
final none = Option<int>.none();

/// Map [int] to [String]
final map = option.map((a) => '$a');

/// Extract the value from [Option]
final value = option.getOrElse(() => -1);

/// Pattern matching
final match = option.match(
  () => print('None'),
  (a) => print('Some($a)'),
);

/// or use Dart's pattern matching as well 🤝
final dartMatch = switch (option) {
  None() => 'None',
  Some(value: final a) => 'Some($a)',
};

/// Convert to [Either]
final either = option.toEither(() => 'missing');

/// Chain computations
final flatMap = option.flatMap((a) => Option.of(a + 10));

/// Return [None] if the function throws an error
final tryCatch = Option.tryCatch(() => int.parse('invalid'));

Either

Used to handle errors (instead of Exceptions).

Either<L, R>: L is the type of the error (for example a String explaining the problem), R is the return type when the computation is successful

/// Create an instance of [Right]
final right = Either<String, int>.of(10);

/// Create an instance of [Left]
final left = Either<String, int>.left('none');

/// Map the right value to a [String]
final mapRight = right.map((a) => '$a');

/// Map the left value to a [int]
final mapLeft = right.mapLeft((a) => a.length);

/// Return [Left] if the function throws an error.
/// Otherwise return [Right].
final tryCatch = Either.tryCatch(
  () => int.parse('invalid'),
  (e, s) => 'Error: $e',
);

/// Extract the value from [Either]
final value = right.getOrElse((l) => -1);

/// Chain computations
final flatMap = right.flatMap((a) => Either.of(a + 10));

/// Pattern matching
final match = right.match(
  (l) => print('Left($l)'),
  (r) => print('Right($r)'),
);

/// or use Dart's pattern matching as well 🤝
final dartMatch = switch (right) {
  Left(value: final l) => 'Left($l)',
  Right(value: final r) => 'Right($r)',
};

/// Convert to [Option]
final option = right.toOption();

IO

Wrapper around an sync function. Allows to compose synchronous functions that never fail.

/// Create instance of [IO] from a value
final IO<int> io = IO.of(10);

/// Create instance of [IO] from a sync function
final ioRun = IO(() => 10);

/// Map [int] to [String]
final IO<String> map = io.map((a) => '$a');

/// Extract the value inside [IO] by running its function
final int value = io.run();

/// Chain another [IO] based on the value of the current [IO]
final flatMap = io.flatMap((a) => IO.of(a + 10));

Task

Wrapper around an async function (Future). Allows to compose asynchronous functions that never fail.

If you look closely, it's the same as IO but for async functions 💡

/// Create instance of [Task] from a value
final Task<int> task = Task.of(10);

/// Create instance of [Task] from an async function
final taskRun1 = Task(() async => 10);
final taskRun2 = Task(() => Future.value(10));

/// Map [int] to [String]
final Task<String> map = task.map((a) => '$a');

/// Extract the value inside [Task] by running its async function
final int value = await task.run();

/// Chain another [Task] based on the value of the current [Task]
final flatMap = task.flatMap((a) => Task.of(a + 10));

Utility types

These types compose together the 4 above (Option, Either, IO, Task) to join together their functionalities:

  • IOOption: sync function (IO) that may may miss the return value (Option)
  • IOEither: sync function (IO) that may fail (Either)
  • TaskOption: async function (Task) that may miss the return value (Option)
  • TaskEither: async function (Task) that may fail (Either)

Reader

Read values from a context without explicitly passing the dependency between multiple nested function calls. View the example folder for an explained usecase example.

ReaderTask

Combine the Reader type (dependecy) with Task (asynchronous).

ReaderTaskEither

Combine the Reader type (dependecy) with Task (asynchronous) and Either (error handling).

State

Used to store, update, and extract state in a functional way. View the example folder for an explained usecase example.

🔗 Do notation

Version v0.6.0 introduced the Do notation in fpdart. Using the Do notation makes chaining functions easier.

For example, a typical chain of methods in fpdart looks as follows:

/// Without the Do notation
String goShopping() => goToShoppingCenter()
    .alt(goToLocalMarket)
    .flatMap(
      (market) => market.buyBanana().flatMap(
            (banana) => market.buyApple().flatMap(
                  (apple) => market.buyPear().flatMap(
                        (pear) => Option.of('Shopping: $banana, $apple, $pear'),
                      ),
                ),
          ),
    )
    .getOrElse(
      () => 'I did not find 🍌 or 🍎 or 🍐, so I did not buy anything 🤷‍♂️',
    );

Notice how you need to call flatMap multiple times to collect multiple variables and use them together (market, banana, apple, pear).

Everything looks more linear and simple by using the Do notation:

/// Using the Do notation
String goShoppingDo() => Option.Do(
      (_) {
        final market = _(goToShoppingCenter().alt(goToLocalMarket));
        final amount = _(market.buyAmount());

        final banana = _(market.buyBanana());
        final apple = _(market.buyApple());
        final pear = _(market.buyPear());

        return 'Shopping: $banana, $apple, $pear';
      },
    ).getOrElse(
      () => 'I did not find 🍌 or 🍎 or 🍐, so I did not buy anything 🤷‍♂️',
    );

You initialize the Do notation using the Do() constructor.

You have access to a _ function, that you can use to extract and use the value inside each Option, without using flatMap.

Note: We recommend using the Do notation whenever possible to improve the legibility of your code 🤝

📦 Immutable Collections

If you are going to use fpdart in your project, make sure to use immutable collections as well ☝️

Immutability is at the core of functional programming (alongside pure functions 🤝).

fpdart does not provide immutable collections. Instead, we strongly suggest to use the fast_immutable_collections package.

fast_immutable_collections provides all the necessary immutable collections (List, Set, Map and more) with an extensive API and top-class performance.

fpdart instead provides some extension methods on Iterable, List, and Map. These are designed to extend the native Dart API with immutable methods, as well as providing many handy additional functions:

/// Dart: `1`
[1, 2, 3, 4].first;

/// fpdart: `Some(1)`
[1, 2, 3, 4].head;

/// Dart: Throws a [StateError] ⚠️
[].first;

/// fpdart: `None()`
[].head;

final map = {'a': 1, 'b': 2};

/// Dart: mutable ⚠️
map.updateAll((key, value) => value + 10);

/// fpdart: immutable equivalent 🤝
final newMap = map.mapValue((value) => value + 10);

You can use these extension methods on any native Dart collection and also in combination with fast_immutable_collections immutable collections.

More

Many more examples are coming soon. Check out my website and my Twitter for daily updates.


🎯 Types

  • x Option
  • x Either
  • x Unit
  • x IO
  • x IORef
  • x IOOption
  • x IOEither
  • x Task
  • x TaskOption
  • x TaskEither
  • x Reader
  • x ReaderTask
  • x ReaderTaskEither
  • x State
  • x StateAsync
  • x Iterable (List) extension
  • x Map extension
  • x Predicate
  • ReaderTaskOption
  • ReaderIO
  • ReaderIOEither
  • ReaderIOOption
  • ReaderEither
  • ReaderOption
  • StateReaderTaskEither
  • Lens
  • Writer

💡 Motivation

📚 Read the full story of the origin and motivation for fpdart

Functional programming is becoming more and more popular, and for good reasons.

Many non-functional languages are slowly adopting patterns from functional languages, dart included. Dart already supports higher-order functions, generic types, type inference. Since Dart 3, the language supports also pattern matching, destructuring, multiple return values (Read more about these new features here).

Other functional programming features are coming to the language, like higher-order types.

Many packages are bringing functional patterns to dart, like the amazing freezed for unions/pattern matching.

fpdart aims to provide all the main types found in functional languages to dart. Types like Option (handle missing values without null), Either (handle errors and error messages), Task (composable async computations), and more.

Goal

Differently from many other functional programming packages, fpdart aims to introduce functional programming to every developer. For this reason, every type and method is commented and documented directly in the code.

You do not need to have any previous experience with functional programming to start using fpdart.

fpdart also provides real-world examples of why a type is useful and how it can be used in your application. Check out my website for blog posts and articles.

Comparison with dartz

One of the major pain points of dartz has always been is lack of documentation. This is a huge issue for people new to functional programming to attempt using the package.

dartz was released in 2016, initially targeting Dart 1.

dartz is also missing some features and types (Reader, TaskEither, and others).

fpdart is a rewrite based on fp-ts and cats. The main differences are:

  • fpdart is fully documented.
  • fpdart implements higher-kinded types using defunctionalization.
  • fpdart is based on Dart 3.
  • fpdart is completely null-safe from the beginning.
  • fpdart has a richer API.
  • fpdart implements some missing types in dartz.
  • fpdart does not provide implementation for immutable collections (ISet, IMap, IHashMap, AVLTree).

🤔 Roadmap

Being documentation and stability important goals of the package, every type will go through an implementation-documentation-testing cycle before being considered as 'stable'.

The current work on fpdart aims to:

  • Expand the API to include more methods to work and convert between types easier (send an issue on the repository if you have a suggestion)
  • Simplify the current API to make it easier to use and learn
  • Expand the documentation with more articles and documentation comments

Note: There is also an experimental research in progress to implement ZIO in fpdart, stay tuned 🔜


The long-term goal is to provide all the main types and typeclasses available in other functional programming languages and packages. All the types should be completely documented and fully tested.

A well explained documentation is the key for the long-term success of the project. Any article, blog post, or contribution is welcome.

In general, any contribution or feedback is welcome (and encouraged!).

📃 Versioning

  • v1.1.0 - 13 August 2023
  • v1.0.0 - 26 July 2023

  • v0.6.0 - 6 May 2023
  • v0.5.0 - 4 March 2023
  • v0.4.1 - 25 February 2023
  • v0.4.0 - 16 December 2022
  • v0.3.0 - 11 October 2022
  • v0.2.0 - 16 July 2022
  • v0.1.0 - 17 June 2022
  • v0.0.14 - 31 January 2022
  • v0.0.13 - 26 January 2022
  • v0.0.12 - 24 October 2021
  • v0.0.11 - 22 September 2021
  • v0.0.10 - 13 August 2021
  • v0.0.9 - 3 August 2021
  • v0.0.8 - 13 July 2021
  • v0.0.7 - 6 July 2021
  • v0.0.6 - 29 June 2021
  • v0.0.5 - 20 June 2021
  • v0.0.4 - 15 June 2021
  • v0.0.3 - 13 June 2021
  • v0.0.2 - 13 June 2021
  • v0.0.1 - 28 May 2021

😀 Support

If you are interested in my work you can subscribe to my newsletter.

I share tutorials, guides, and code snippets about fpdart and functional programming: Subscribe to the Newsletter here 📧

For more frequent updates you can also follow me on my Twitter.

👀 License

MIT License, see the LICENSE.md file for details.

Libraries

fpdart