#Generics: How to allow either T or T[] but disallow T|T[] in a generic data structure?

30 messages · Page 1 of 1 (latest)

fierce aspen
#

seems like the reverse of this is what you're looking for

#

so overloads could help with this

#

although of course T could be an array already unless you prevent it with T extends number | string or whatever

#

in what situation is this relevant/necessary to separate T from T[] for stringifications though?

#

i feel like you aren't actually giving the full situation, are you recieving multiple parameters?

safe sundialBOT
#
RickWong#0936
Playground Link

Preview:```ts
class ToString<T> {
run(val: any) {
return val.toString()
}
}

const sys = new ToString<number | number[]>() // i want this to complain

let a = sys.run(1)
let b = sys.run([1])
console.log(a, b, a === b) // i dont want this```

You can choose specific lines to embed by selecting them before copying the link.
fierce aspen
#

that's not really how it works

#

that's also kinda not a great pattern

#

what is this supposed to be for

# 
const a = new ToString<number>().run(1);
// a = "1"
const b = new ToString<number[]>().run([1]);
// b = "1"
#

i really have no idea what you're trying to achieve here

#

there is, but that's not what generics are for

#

im confused, Sets already support arbitrary types

#

uh it does support objects though

#

they're compared by reference as they should

#

this is js lol not python

#

this just isn't a pattern used in js

#

i feel like you're missing the point of what im saying

#

js doesn't use objects by value, ts isn't designed with that in mind

#

and js ecosystem, weird types, yada yada, you've probably heard that before
ts allows types that don't work in other languages becuase of the js it's built on

#

when did i say anything about that

meager elm
#

I get the feeling you're trying to be helpful, so thanks for that

fierce aspen
#

on mobile now so can't test, but if you insist;

a conditional type could probably work, something along the lines of T has U[] ? T has U ? never : T : T
the 2 checks would be with [T] to avoid distributive conditionals, extends, infer
i honestly have no idea how it'd actually look though

#

i mean, see how convoluted it already is lol

meager elm
#

It's ok if you don't have the answer

fierce aspen
#

although, you'd of course need a ton more checks if you actually want it to work as a set

#

actually it'd be way more convoluted than i said, since number|number[][] would also have to fail
or string|number
or string with anything
or objects with a toString with anything

#

strings maybe wouldn't be so bad with [string] extends [T]