Reduce with initial keys not directly related to current value | TypeScript Community | Page 1

steady gazelleBOT Feb 10, 2023, 2:40 PM

#

Daniell#4062

Playground Link

Preview:```ts
import {} from "discord-api-types"
import {
ApplicationCommandOptionData,
CommandInteractionOption,
ApplicationCommandOptionType,
User,
Channel,
Role,
Attachment,
} from "discord.js"

export type ApplicationCommandOptionTypeValues<
T extends ApplicationCommandOptionData[]

...```

split nova Feb 10, 2023, 2:41 PM

#

Could use some help to solve the type error(s) in a clean way

uncut edge Feb 10, 2023, 2:56 PM

#

not sure there's a safe way to solve it

#

but here's what i think is the least bad way to solve it:

steady gazelleBOT Feb 10, 2023, 2:56 PM

#

n_n#2622

Playground Link

Preview:```ts
...
ApplicationCommandOptionType.Attachment
) {
previousValue[option.name] = option.attachment
}

return previousValue

}, {} as Partial<Record<string, ApplicationCommandOptionTypeValues<T>[ApplicationCommandOptionType]>>) as InferCommandOptions<T>
...```

uncut edge Feb 10, 2023, 2:57 PM

#

you widen it to have a string index signature, and let any value be assignable to it

#

which is... not quite ideal

#

but it's a lot better than nothing imo

#

also i changed the [].some() to option.type === A || option.type === B

#

since [].some() doesn't narrow option.type

#

not that it matters here, so it's more of a personal preference

split nova Feb 10, 2023, 3:04 PM

#

Thanks! Is the any 100% needed? No passing in generic hack or more accurate assert?

uncut edge Feb 10, 2023, 3:10 PM

#

you could replace it with as unknown as T[number];

#

but as unknown as T is just as unsafe (read: 100% unsafe)

#

the inferred type is InferCommandOptions<ApplicationCommandOptionData[]>...

#

which may be correct tbh

#

  [ApplicationCommandOptionType.Subcommand]: T[number];
  [ApplicationCommandOptionType.SubcommandGroup]: T[number];

#

this does look a bit suspicious

#

where does the T from T extends ApplicationCommandOptionData[] even come from btw

split nova Feb 10, 2023, 3:35 PM

#

uncut edge where does the `T` from `T extends ApplicationCommandOptionData[]` even come fro...

https://tsplay.dev/w6x6Em Added to your playground at the bottom, I wasn't sure how to type it and I guess it's magically inferred now? 😅

TS Playground - An online editor for exploring TypeScript and JavaS...

The Playground lets you write TypeScript or JavaScript online in a safe and sharable way.

#

[ApplicationCommandOptionType.Subcommand]: T; is probably wrong though and should be something like InferCommandOptions<T[number]> or something

#

Or well it should be whatever transform(option.options ? [...option.options] : []) returns, not sure if I was correct in that case

split nova Feb 11, 2023, 6:32 AM

#

uncut edge ```ts [ApplicationCommandOptionType.Subcommand]: T[number]; [ApplicationComm...

I think InferCommandOptions<T> is correct here, does that allow for a safe assert?

uncut edge Feb 11, 2023, 6:37 AM

#

split nova I think InferCommandOptions<T> is correct here, does that allow for a safe asser...

wdym safe assert?

#

uhm

#

ok so the issue is that Subcommand has a different list of options, right?

#

and so:

  [ApplicationCommandOptionType.Subcommand]: T[number];
  [ApplicationCommandOptionType.SubcommandGroup]: T[number];

#

i don't think this is correct

#

but i'm not sure you can even have a good, narrow type for these

split nova Feb 11, 2023, 6:42 AM

#

uncut edge ok so the issue is that `Subcommand` has a *different* list of options, right?

I got the inspiration of inferring options from here https://github.com/Naval-Base/yuudachi/blob/main/packages/framework/src/types/ArgumentsOf.ts#L67 but I don't like how overcomplicated the types are written

potent zealotBOT Feb 11, 2023, 6:42 AM

#

: T extends ApplicationCommandOptionType.SubcommandGroup

uncut edge Feb 11, 2023, 6:42 AM

#

i think this is the correct solution:

steady gazelleBOT Feb 11, 2023, 6:43 AM

#

n_n#2622

Playground Link

Preview:```ts
...
[ApplicationCommandOptionType.Subcommand]: InferCommandOptions<
ApplicationCommandOptionData[]

ApplicationCommandOptionData[]

[
...```

uncut edge Feb 11, 2023, 6:43 AM

#

the error does go away, but that's because the types are wider

#

but, i don't think it's possible to have narrower types?

split nova Feb 11, 2023, 6:44 AM

#

I think that's fine! In the end it's not possible to narrow anyway because the function should work with all commands and their different options

#

And the option object is typed anyway where it matters

#

Thanks

uncut edge Feb 11, 2023, 6:47 AM

#

oh

#

i just saw

#

const transform = <T extends ApplicationCommandOptionData[]>(
  options: ReadonlyArray<CommandInteractionOption<"cached">>
)

#

there is nowhere to infer T from

#

and so the generic does nothing here

split nova Feb 11, 2023, 6:52 AM

#

uncut edge there is nowhere to infer `T` from

But when you hover the call to transform you see that the generic is filled in, how is that possible?

uncut edge Feb 11, 2023, 6:54 AM

#

split nova But when you hover the call to transform you see that the generic is filled in, ...

target typing (aka return type inference)... highly unsafe

split nova Feb 11, 2023, 6:55 AM

#

Ohh okay I'll get rid of it then

uncut edge Feb 11, 2023, 6:55 AM

#

yeah idk if there's a nice solution

#

because interaction.options isn't generic

split nova Feb 11, 2023, 6:56 AM

#

Yup

#

What's up with the bigint error though 🤔

uncut edge Feb 11, 2023, 6:56 AM

#

wait where

split nova Feb 11, 2023, 6:56 AM

#

In transform

uncut edge Feb 11, 2023, 6:57 AM

#

there's an error?

split nova Feb 11, 2023, 6:57 AM

#

Oh not anymore, saw a red underline on mobile

uncut edge Feb 11, 2023, 6:57 AM

#

yeah weird stuff happens when it's still loading types

#

anyway fwiw i just have a massive nested switch-case for commands

#

and just do unsafe casts

split nova Feb 11, 2023, 6:58 AM

#

Can I see your repo

uncut edge Feb 11, 2023, 6:59 AM

#

it's not public tho

split nova Feb 11, 2023, 6:59 AM

#

Ah

uncut edge Feb 11, 2023, 6:59 AM

#

i use discordeno, not djs

#

but the idea is the same

split nova Feb 11, 2023, 6:59 AM

#

Don't you prefer something like above?

#

Overall it work pretty good

uncut edge Feb 11, 2023, 7:00 AM

#

it does, but it's too complicated for me

#

interface FooData {
    /** unix epoch */
    timestamp: number;
    foo: number;
    bar: number;
}

#

function getOpts<T>(obj: { options?: unknown[]; }): T {
    // deno-lint-ignore no-explicit-any
    return Object.fromEntries((obj.options ?? []).map((opt: any) => [opt.name, opt.value])) as T;
}

#

this is so horribly unsafe

#

but hey, i get nice types and that's all i care about

#

you do have to make sure it stays in sync with your options

#

but realistically after the first few mistakes you start to remember to update things everywhere else

#

plus you'll catch it when testing anyway

split nova Feb 11, 2023, 7:02 AM

#

I see, honestly I mainly made this to test my limits so I'm trying to avoid any and type all those complex patterns

uncut edge Feb 11, 2023, 7:02 AM

#

uncut edge ```ts function getOpts<T>(obj: { options?: unknown[]; }): T { // deno-lint-i...

and finally, usage:

getOpts<FooData>(subcommandGroup /* interaction.options */ as never);

split nova Feb 11, 2023, 7:03 AM

#

Yea, that would have been easier

#

But now I have 0 any which is kind of satisfying

uncut edge Feb 11, 2023, 7:04 AM

#

i guess

#

but do remember, just because you avoid any it doesn't mean it's any less unsafe

#

sometimes it's better to use any to make it obvious where you're abusing the type system

split nova Feb 11, 2023, 7:05 AM

#

Yup I think I have a good feel where it should be used or there's no other option

uncut edge Feb 11, 2023, 7:05 AM

#

also fwiw i think decorators are great for ux

uncut edge Feb 11, 2023, 7:05 AM

#

uncut edge also fwiw i think decorators are great for ux

which i guess is the reason why a bunch of discord bot frameworks use them...

split nova Feb 11, 2023, 7:06 AM

#

I wish decorators would work outside classes :/

#Reduce with initial keys not directly related to current value