TS2416 nonsensical fix suggestion (and possible false positive?) in a generic class implementing an interface verbatim
#59.377 aperta il 21 lug 2024
Metriche repository
- Star
- (48.455 star)
- Metriche merge PR
- (Merge medio 6g 17h) (9 PR mergiate in 30 g)
Descrizione
🔎 Search Terms
TS2416
🕗 Version & Regression Information
- This is the behavior in every version I tried, and I reviewed the FAQ for entries about covariance in method params
⏯ Playground Link
💻 Code
interface U<T1, T2> {
p2: T2; // removing this or removing T1 param from the interface fixes the error
}
interface IFoo<T> {
test<IT2, P extends U<T, IT2>>(t2: IT2, p: P): void;
// ^ compiler suggests to use `IT2 extends IT2` here, which makes no sense
}
class Foo<T> implements IFoo<T> {
test<T2, P extends U<T, T2>>(t2: T2, p: P): void {
// ^ `any` here fixes the error
// ^ Property 'test' in type 'Foo<T>' is not assignable to the same property in base type 'IFoo<T>'.
// Type '<T2, P extends U<T, T2>>(t2: T2, p: P) => void' is not assignable to type '<IT2, P extends U<T, IT2>>(t2: IT2, p: P) => void'.
// Types of parameters 'p' and 'p' are incompatible.
// Type 'P' is not assignable to type 'U<T, IT2>'.
// Type 'U<T, IT2>' is not assignable to type 'U<T, IT2>'. Two different types with this name exist, but they are unrelated.
// Type 'IT2' is not assignable to type 'IT2'. Two different types with this name exist, but they are unrelated.
}
}
🙁 Actual behavior
Compiler suggests to modify the interface to extend it's IT2 generic parameter (IT2 extends IT2), which makes no sense (and is an invalid ts code), which suggests, that compiler itself is confused about what exactly is wrong with this implementation.
The interface is implemented verbatim.
The error message shows, that verbatim copies of inferred types are not the same type (mind, i changed the generic param names to check, if i assumed something wrongly about inferred generics in specific places, yet it's Type 'IT2' is not assignable to type 'IT2'. Two different types with this name exist, but they are unrelated.)
Something is wrong somewhere but error message only confuses you more.
🙂 Expected behavior
If it is a bug, it should not error on verbatim implementation of an interface.
If it isn't, it should print an intelligible error message, instead of "hey, 'warm' !== 'warm', 'cold' !== 'cold' and 'soft' !== 'soft', din't you know?", and it should not suggest nonsensical fixes, like "replace generic T with T extends T"
Additional information about the issue
It is possibly related to issue #30071 which have been fixed in v4.3.5, apparently (playground)
Notice the similarities (bug is present in 4.2.3), removing then generic param T from interface or using : any for the parameter using T2 makes the error go away (but there is no nonsensical fix suggestion in this case)