Skip to content

Latest commit

 

History

History
287 lines (207 loc) · 8.48 KB

README.md

File metadata and controls

287 lines (207 loc) · 8.48 KB

Rationale

Rationale is inspired by RamdaJS. It is a collection of helper utility functions that are absent in the OCaml/ReasonML standard library.

Note that not all of Ramda was ported over, as many of Ramda's utilities are making up for deficits in Javascript, which Reason doesn't have. Furthermore, many of the functions that operate on objects, simply don't make sense in Reason.

Installation

Run npm install --save rationale and add rationale to bs-dependencies in bsconfig.json.

Features

Exception-free List operations

In the OCaml/ReasonML standard library, many of the common List operations throw exceptions if there's a problem. Rationale's utilities do not throw exceptions, and instead return options.

  • head
  • tail
  • init
  • last
  • nth
  • etc

Monadic Options and Belt.Results

Rationale includes monadic and functor operations ala Haskell for the option and Belt.Result types.

open Rationale.Option.Infix;
open Rationale.Function;

RList.init(a)
  >>= ((x) => RList.last(a) <$> f <$> flip(RList.append, x))
  <$> RList.concat(b)
  |> Option.default(xs);

Additional ADT-s

Reader

Writer

IO, KIO

The main idea of the IO monad is to isolate our effects as much possible. Some languages like Haskell don't even allow the user to manually start an effect, which is not the case for Reason, but with a little bit of self-discipline we can handle side effects in a monadic fashion as well.

KIO is a monadic structure with the type signature: io('a, 'env) = IO(Lazy.t('env => 'a)) which makes it posible to store some effect with its config environment. ( dependency injection )

We can use return or lift methods to wrap our unsafe mutations. The difference is that with return we throw away the environment, while with lift we are using it.

As you can see in our example we used return to wrap Js.log

let log = (x: string) =>
  KIO.return({
    Js.log(x)
  });

In the following saveFile example we use lift which makes it possible to use an injected env config ( the injection happens when we call KIO.runIO )

let saveFile = str => KIO.lift(env => {
  Node.Fs.writeFileSync(env.target, str, `ascii);
});

To run our effects we need to call: KIO.runIO(main, env); Ideally in your program this method will be called once on the bottom of your index file. Example:

  exception ReadError(string);
  type envT = {
    path: string,
    dir: string,
    target: string
  };
  let env = {path: "./input.txt", dir: "/", target: "./out.txt"};

  let readFile = KIO.lift(env => {
      try (Node.Fs.readFileSync(env.path, `ascii)) {
      | ReadError(msg) => raise @@ ReadError("File read failed: " ++ msg)
      };
    });

  let saveFile = str => KIO.lift(env => {
    Node.Fs.writeFileSync(env.target, str, `ascii);
  });

  let log = (x: string) =>
    KIO.return({
      Js.log(x)
    });

  let parseFile = input => {
    let l  = Js.String.split("\n", input);
    Array.map(x => x ++ "100", l);
  } 

  let joinArray = (xs: array(string)) => Js.Array.join(xs)

  let main = KIO.(
    readFile 
      <$> parseFile
      <$> joinArray
      >>= saveFile
  )

  KIO.runIO(main, env);

Support for Point-free style

Rationale has compose and pipe functions, as well as supporting infix operators: <|| and ||> respectively.

Infix Lens composition

Rationale also allows for fluid lens composition via infix operators: -<< and >>-.

Lens.view(aLens >>- bLens >>- optional(0), { a: { b: Some(3) } });

Function signatures for composition

Like in Ramda, functions always keep their data at the end making piping and composing a breeze:

list
  |> take(9)
  |> drop(3)
  |> splitAt(4);

Usage

Using Optional Returns in RList and Dict

Returning option('a) from functions is generally preferred to throwing an exception. It protects you from runtime errors, and forces you to deal with potential errors at compile time. However, if you're not used to doing it, things can get a little confusing.

Pattern matching for errors all the time would be extremely cumbersome. Fortunately, we provide a host of useful methods to working with optional returns. Hopefully, this doc will show you that you don't need to use excessive pattern matching to work with optional returns.

Default

The most straight forward way to get out of an option is by calling default.

Option.default(0, Some(1)); /* 1 */

Option.default(0, None); /* 0 */

Monads

Calling default will definitely get you out of the option, but what if you want to do some things to it first? What if you need other funtions that also return option?

the Option module includes monadic operations for option, so you can take a railway oriented approach to working with them.

First, let's check if the last item of a list is equal to a certain value:

let lastEquals = (a, xs) => Option.fmap(Util.eq(a), RList.last(xs)) |> Option.default(false);

lastEquals(3, [1,2,3]); /* true */
lastEquals(3, [4,5,6]); /* false */
lastEquals(3, []); /* false */

/* Or, with infix operators */
open Option.Infix;

let lastEquals = (a, xs) => RList.last(xs) <$> Util.eq(a) |> Option.default(false);

Here we used fmap and its infix variation <$> to apply a function to the value inside the option. Note that, Util.eq returns a bool not an option. So what if the next function also returns an option? Well you'd get nested options, which doesn't really help anyone. So, instead, we would use bind.

Now let's replace the last item of one list with the last item of another. Note that both last and init return option:

let swapLast = (xs, ys) =>
  Option.(bind(RList.last(xs), ((x) => fmap(RList.append(x), RList.init(ys)))) |> default(ys));

swapLast([1,2,3], [4,5,6]); /* [4,5,3] */
swapLast([], [4,5,6]); /* [4,5,6] */

/* Or, with infix operators */
open Option.Infix;

let swapLast = (xs, ys) =>
  RList.last(xs) >>= ((x) => RList.init(ys) <$> RList.append(x)) |> Option.default(ys);

Here we used bind and its infix variation >>= to apply a function that also returned an option.

Applicatives

Let's try checking if the last elements of two lists are equal. We could accomplish this using bind, but that can be a little awkward.

let lastEqual = (xs, ys) =>
  Option.(apply(apply(Some(Util.eq), RList.last(xs), RList.last(ys))) |> default(false));

lastEqual([1,2,3], [4,5,3]); /* true */
lastEqual([1,2,3], [4,5,6]); /* false */
lastEqual([], [4,5,6]); /* false */
lastEqual([1,2,3], []); /* false */

/* Or, with infix operators */
open Option.Infix;

let lastEqual = (xs, ys) =>
  Some(Util.eq) <*> RList.last(xs) <*> RList.last(ys) |> Option.default(false);

Alternative

With alternative you can implement simple but powerful fallback mechanism your ADT-s;

Example:

  open Option;
  let someData = some("Hello");
  let guard = fun
  | true => pure() 
  | _ => empty();


  let startWith: (string, string) => option(unit) = (str, char) => guard(Js.String.startsWith(char, str))
  let dataWithFallback =
    someData 
      >>= val_ => startWith(val_, "T")
      >>= (_ => {Js.log(val_); some(Js.String.toUpperCase(val_))})
      <|> some("Not started with T")

Translating JS Idioms

Or chains

Take the following example in Javascript:

let x = a || b || c || d;

We can't translate that directly to Reason, because there is no null or undefined in Reason. The closest approximation would be option, in which we can string together Some and None to get the first one that is Some.

There is a helper function called firstSome and its infix variation |? that do exactly this.

/* a, b, and c are all options, but d is not */
let x = a |? b |? c |> default(d);

Reference

Infix Operators

  • >>=: Monadic Bind
  • <$>: Functor Fmap
  • <*>: Applicative Apply
  • <||: Point-free Function Compose
  • ||>: Point-free Function Pipe
  • -<<: Lens Compose
  • >>-: Lens Pipe
  • |?: Optional Or