Skip to content

sammyrulez/typed-monads

Repository files navigation

Type-safe Monads

Build PyPI version codecov Code style: black

This is an experiment in building monads in Python supported by strict type annotations. The goal is to be able to compose monads with the type checker ensuring their correctness.

Motivation

I'm a fan of monads, but believe they work best with the support of a strong type system. I've attempted to use libraries like PyMonad, but been frustrated by a lack of type constraints preventing incorrect usage. I could've attempted to add type annotations to one of those libraries, but building my own is more fun.

This is a fork of the original work by Correl Roush

I added some utility methods to make it easier to use in my day to day code and better interate with the pythonic style ( ie List Comprehension )

Installation

$ pip install typed-monads 

Documentation

You can find all the docs here

Curring

Mixing Higher order functions ( functions that return a function ) with moand is a very common programming style other functional programming languages. With curry decorator you can transform a function in a curried function: just apss some positional parameters and get back a function with the remaining ones.

@curry
def power(exp: int, base: int ) -> int:
    return math.pow(base, exp)

square_fn = power(2) # a function that returns the square of the parameter

Base Classes

Functor

map (*)

Applies a function to the contents of a functor, transforming it from one thing to another.

The * operator implements map on functors, and is both left and right associative:

def wordcount(s: str):
    return len(s.split())


f.map(wordcount) == wordcount * f == f * wordcount

Applicative

Extends Functor.

pure

Wraps a value in an applicative functor.

e.g.:

Maybe.pure("abc") == Just("abc")
Result.pure(123) == Ok(123)

apply (&)

Transforms the value contained in the instance's functor with a function wrapped in the same type of functor.

The & operator implements apply on applicatives, and is right-associative.

e.g.:

increment = lambda x: x + 1

Just(3).apply(Just(increment)) == Just(increment) & Just(3) == Just(4)

This can be very handily combined with map to apply curried functions to multiple arguments:

subtract = lambda x: lambda y: x - y

subtract * Just(10) & Just(4) == Just(6)

Monad

Extends Applicative.

bind (>>)

Passes the value within the monad through an operation returning the same type of monad, allowing multiple operations to be chained.

The >> operator implements bind on monads, and is left-associative.

@curry
def lookup(key: str, dictionary: Dict[str, str]) -> Maybe[str]:
    try:
        return Just(dictionary[key])
    except KeyError:
        return Nothing()


result = Just({"hello": "world"}).bind(lookup("hello")).bind(lambda s: s.upper())
result = (
    Just({"hello": "world"})
    >> lookup("hello")
    >> (lambda s: s.upper())
)

Monoid

mappend (+)

Describes an associative binary operation for a type.

mzero

Provides an identity value for the mappend operation.

mconcat

Accumulates a list of values using mappend. Returns the mzero value if the list is empty.

Monads

Wrapped values should be immutable: they are protected from accidental direct writing with Final type and the pythonic naming convention.

Maybe[T]

Represents optional data. A Maybe instance of a certain type T will either be a Just object wrapping a value of that type, or Nothing.

  • Mapping a function over Nothing will return Nothing without calling the function.
  • Binding an operation with a Nothing will return Nothing without attempting the operation.

Result[T, E]

Represents a state of success or failure, declaring a type for each. A Result instance will either be an Ok object wrapping a value of the success type T, or an Err object wrapping a value of the failure type E.

  • Mapping a function over an Err will return the Err unchanged without calling the function.
  • Binding an operation with an Err will return the Err unchanged without attempting the operation.

List[T]

Represents a ordered sequence of items.

  • Also implements Monoid.

Set[T]

Represents a unordered sequence of unique items.

  • Also implements Monoid.

Future[T]

Represents an asynchronous action.

  • Also implements Awaitable.

Reader[T]

Represents the application of a function to it's argument.

Monads as iterable

It is handy to iterate over some monad contents. List is obliviously the first candidate:

m_list: List[int] = List([1, 2, 4, 9])
for i in m_list:
    ...

#Or filter with a generator

evens: List[int] = [k for k in m_list if k % 2 == 0 ]

If you want to something to happen just if a Maybe monad is defined

for n in Just("one"):
  ...

The same apply for Results

About

Easy functional monads for your python code

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages