Pipe as lazy function

[darky]

Suggestion

⭐ Suggestion

pipe in FxTS works in eager fashion.

pipe(1, (n) => n + 1) // receiving of 2 immediately

Would be great to have something like pipeLazy, which works in lazy fashion.

const fn = pipeLazy((n: number) => n, n => n + 1) // Function<(n: number) => number>
fn(1) // 2

💻 Use Cases

• Huge amount of FP libraries, that I know, using lazy pipe, not eager
  • lodash.flow
  • ramda.pipe
  • rxjs.pipe
  • rocket.pipes
    ...
• Lazy pipes more JIT friendly and faster. They can be initialized once on application bootstrap and will works with maximum speed. Eager pipes will be generated every time and spend additional CPU resources.
• Eager pipes increasing GC work much more than Lazy pipes, because lazy pipes can be initialized on application bootstrap and don't recreate pipe functions every time
• Lazy pipes can be easily composed with each other.
• Lazy pipes reduces pipe hell. Lazy pipes propagates to creating many little pipes. Eager pipes propagates to creating one big pipe.

[ppeeou]

Good idea. I already checked your library rocket.pipes 👍

Could you please create a PR??

[darky]

Good idea. I already checked your library rocket.pipes 👍

Could you please create a PR??

Thanks, yep 👌
But need to wait some time, when I will get free time 🕐

[ppeeou]

@darky There is something I would like to discuss.

Currently, pipeLazy supports code type 1, but code type 2 is not supported. Existing functions used in FxTS are designed to receive and determine whether it is Iterable or AsyncIterable. In the case of shape 2, it is not included in the specified signature because we do not know what type will come from the head function.

1. pass

const res1 = pipeLazy(
    (a: string) => head(a),
);

2. fail

const res2 = pipeLazy(
   head,
);

So the initial design was intended for users to type something like the following to use the pipeLazy function.

const res3 = (a:string) => pipe(a, head)

Is there any way to support type 2?🧐

[darky]

@ppeeou

Sorry, but I don't think that case 2 is possible.
Due to my experience with rocket-pipes, ramda pipe, lodash flow first function in pipe should be typed in explicit way usually
Seems it's limitation of TypeScript type system
You desire more fits with ML based languages like Ocaml, ReScript, Flow where type of function parameters can inferred from caller side

P. S. Maybe modern TypeScript allows to implement case 2 via some secret trick, but I don't know it. Will glad to accept this insight. Suggestions welcome!