Types are missing and causing errors

[shamilovtim]

Using react-navigation v4

Hook:
const navigation = useNavigation<any>();

Use hook:
navigation.popToTop();

Typescript error:
Property 'popToTop' does not exist on type 'NavigationScreenProp<any, NavigationParams>'.ts(2339)

[satya164]

IMO useNavigation should take the type of navigation prop as the generic instead of S (?):

const navigation = useNavigation<NavigationStackProp>();

cc @slorber

[slorber]

yeah that's probably right, will try to see how to expose custom nav actions more easily

[Grohden]

Edit: that was a suggestion? I thought that you said it could be used that way in the current version XD

@satya164 the return type of the hook is NavigationScreenProp<S & NavigationRoute>

and the type is defined as (in react-navigation package):

  export interface NavigationScreenProp<S, P = NavigationParams> {
    state: S & { params?: P };
    dispatch: NavigationDispatch;
    goBack: (routeKey?: string | null) => boolean;
    dismiss: () => boolean;
    navigate(options: {
      routeName:
        | string
        | {
            routeName: string;
            params?: NavigationParams;
            action?: NavigationNavigateAction;
            key?: string;
          };
      params?: NavigationParams;
      action?: NavigationAction;
      key?: string;
    }): boolean;
    navigate(
      routeNameOrOptions: string,
      params?: NavigationParams,
      action?: NavigationAction
    ): boolean;
    openDrawer: () => any;
    closeDrawer: () => any;
    toggleDrawer: () => any;
    getParam<T extends keyof P>(
      param: T,
      fallback: NonNullable<P[T]>
    ): NonNullable<P[T]>;
    getParam<T extends keyof P>(param: T): P[T];
    setParams: (newParams: Partial<P>) => boolean;
    emit: (eventName: 'refocus') => void;
    addListener: (
      eventName: string,
      callback: NavigationEventCallback
    ) => NavigationEventSubscription;
    isFocused: () => boolean;
    isFirstRouteInParent: () => boolean;
    router?: NavigationRouter;
    dangerouslyGetParent: () => NavigationScreenProp<S> | undefined;
  }

The S is required, but it's used for state, so using:
const navigation = useNavigation<NavigationStackProp>(); doesn't solve for me :/

Is there anything new to release on other packages?
I'm using these versions:

 "react-navigation": "4.0.9",
 "react-navigation-hooks": "1.0.3",
 "react-navigation-stack": "1.9.0"

Btw I think that this could be solved using:

export function useNavigation<NavigatorProps = any, S = any>(): NavigationScreenProp<
    S & NavigationRoute
> & NavigatorProps

[satya164]

Just a suggestion sorry 😛

[Wintus]

I agree with the idea that useNavigation should take a prop in addition. #42 (comment)

At this time to avoid the error, I figured out no workaround on this problem but a (poor) downcasting as below:

import {useNavigation} from 'react-navigation-hooks';
import {NavigationStackProp} from 'react-navigation-stack';

// FIXME: workaround
export const useStackNavigation = () => useNavigation() as NavigationStackProp;

And this works fine so far.

Here is a more sophisticated downcast.
https://gist.github.com/Wintus/c7fcfda1a735e6f7f89ce4c54c1d5d8e#file-usestacknavigation-ts

[slorber]

Hey, just thinking out loud here.

For you what's the difference between useNavigation<NavigationStackProp>() and useNavigation() as NavigationStackProp and useStackNavigation?

For me all those are different syntaxes to downcast to a stack navigation object. Is one really better than the other conceptually if we don't take into account the verboseness?

If you call useStackNavigation and it returns a switch navigation, that could be misleading as well. Maybe we should have a custom hook for each type of navigator to encapsulate the casting + a runtime assertion to ensure we don't return a switch navigator when user call useStackNavigation?

But what about custom navigators/routers (should we ask the user to create his own hook and do the casting/runtime check on his own?)

@satya164 how do you plan to solve this in v5?

[satya164]

how do you plan to solve this in v5?

In v5, useNavigation accepts an optional type parameter:

const navigation = useNavigation<StackNavigationProp<RootParamList>>()

For me all those are different syntaxes to downcast to a stack navigation object

In this case the end result is same, but type casting and generics have very different semantics.

With type casting, you tell the compiler that the value is a certain type and it won't complain even if it's not (as long as the interface matches vaguely). Typecasting should be avoided unless necessary.

A generic is like a parameter, TypeScript will show type hints that the hook accepts this type parameter, it maybe required or optional, the hook can ensure that supplied type parameter matches the desired constraint.

Consider the following 2 snippets:

useNavigation() as { setParams: Function }; // No error

useNavigation<{ setParams: Function }>(); // Shows error since the supplied type doesn't match constraint

If you call useStackNavigation and it returns a switch navigation, that could be misleading as well. Maybe we should have a custom hook for each type of navigator to encapsulate the casting + a runtime assertion to ensure we don't return a switch navigator when user call useStackNavigation?

The problem here is that we grab the value automatically from the context. There's no way to statically verify that the type/function being used is correct so we need to entirely trust user on this.

A navigation object can be a combination of multiple navigation objects as well (e.g. have both push and openDrawer methods) depending on nesting, so we can't restrict the type to only one navigator either.

A runtime assertion will provide minimal value here imo, at the cost of each navigator needing to export their own hook.

[slorber]

But in such case, if we want to support custom actions/navigators, we can't do any constraint on the generic params if we want to allow const navigation = useNavigation<MyCustomNavigator<MyParams>>().

useNavigation<T> = () => useUnsafeNavigation() as T is not really safer, it's just encapsulating the downcast so just feels a bit nicer to use for users.
Do we have any constraint here to apply on T that make sense in all cases (including custom navs?)

[satya164]

See the linked code for useNavigation. The code I posted gives error because there's a constraint.

[slorber]

Thanks, will try to see what I can take from v5

Seems like you have built a typesafe route/params system, looks nice. Wonder if this can be ported to v4 without too much work but maybe i'd rather do something more simple

[satya164]

It can be ported, but it'll break all existing types of v4 so probably not worth it.

[luisnaranjo733]

So no support for this until react-navigation v5?

[slorber]

@luisnaranjo733 if you find a good solution for v4, don't hesitate to send a PR