Apply TS 2.1 `keyof` & mapped types about npm-ramda HOT 5 CLOSED

Whoa, that's awesome, thank you! 😄 It evidently did better than my threads on path there...
I think I can make versions out of that for different path lengths; that should help.
As it stands it appears unable to deal with navigation of arrays though, as [K1 in A] notation apparently demands strings. I may want to ask them for ideas on that...

from npm-ramda.

Thanks! I've now tried to use these in assoc, dissoc, map, mapObjIndexed, pick, pluck, project, prop, propIs, propEq, propOr, propSatisfies, and a failed/WIP attempt at path. I've been talking to donnut and blakeembrey, and hopefully we'll manage an NPM release soon.

Now lenses are a bit tough. Since you brought them up, I tried to expand on typings there, but I'm not sure yet how I can use keyof there, and I'm open to input there. The issue pertains order of generics/information, and this is one of the great general challenges in typescript-ramda. Let's say we wanted to type a lens such as to make use of this:

    interface KeyLens<K extends keyof T> {
        <T extends Struct<any>>(obj: T): T[K]; // get
        <T extends Struct<any>>set(v: T[K], obj: T): T;
        // <T extends Struct<any>>map(fn: (v: T[K]) => T[K], obj: T): T
    }

The main issue here is of course, such a lens is actually generic, in the sense it's agnostic toward the actual data structure it might be applied to. We'd like to be able to define such a lens just by the key it navigates to, but if we wish to apply keyof there, we run into a problem: the object structure T is not defined yet at this point.

Alternatively, we can add this T in the interface generics already, before K, and hope TS will take our word for the fact this will hold for any given T.
I made a PoC for this, but unfortunately, TS doesn't appear to like this reasoning. I tries to evaluate things instead of lazily keeping our options open, and as a result, ends up discarding this option entirely.

There might be a way to improve on that, but I'm not sure if it could work in the current state of TS.
I'll close for the moment as I think I've handled anything actionable as of now. This doesn't mean I'm not looking for further suggestions to improve here though -- to the contrary!

from npm-ramda.

@tycho01
This is how path can be typed: microsoft/TypeScript#12290

from npm-ramda.

Hello :)

I'm clearly struggling understanding this new syntax...

Example:

import R from "ramda";

interface Car {
  wheels: number;
  doors: number;
};

type carKeys = keyof Car;

interface FakeCar {
  wheels: number;
};

type fakeCarKeys = keyof FakeCar;

const car: Car = { wheels: 4, doors: 5 };

const fakeCar: FakeCar = R.pick<Car, fakeCarKeys>(["wheels"], car);

console.log(fakeCar);

This get's me:

error TS2536: Type '0' cannot be used to index type 'T'

What should I use as the second parameter of <,> (if type inference is failing me)?

EDIT: nevermind, the problem is not on my code, but here. A bug?

from npm-ramda.

Hey, this appears to be the regression described in #129. I haven't had the time to test their fix though... if the new nightly works we can just update it.

from npm-ramda.