[5.4 regression] Conditional type changes behavior with 5.4-beta

🔎 Search Terms

5.4 regression

🕗 Version & Regression Information

This changed between versions 5.3.3 and 5.4-beta

⏯ Playground Link

https://www.typescriptlang.org/play?ts=5.4.0-dev.20240129#code/KYDwDg9gTgLgBDAnmYcCiIZQIYGMYBKEArjMAArY4C2AzgDwAqcoZAdgCa1y1YCWbAOYA+OAF4e-IS0zBO3RgFgAUHDgB+OAWC5oHerygDBAGklGhwlWoBccZqzlc4AAwAkAbwEAzYFDgA+gDKMFQwAL42nj5+cJQ04QD00Wy+-tq84S7WGnAecADaANZwAnFU2NRwAD5wRcCIEN7omDj4RKQUFXT0GTDCALp2hsZw4Tl2DrLyrilpgSFhkXOx8ZVZOZr5xaVs5TRD5qPjqnB2HuEA3CoqHDoANlSo3sRs+HwQe-cCRUwy7M4RpYABQ5KiCWjnHJqMDYGAAC0m11O4TgADI4MCLv8nNwMFg8IQSGQ1j1GKItnBYTRaOpzqjUecqd1IS0Ce1iV0aUxROEAJQmFR8uwANwgfA4yJUiUSLCgUGguwQ3AArAA6ADMmrg2E4cHVABY1QAGAC0ACNgKEVN82EUscyEXYAOSJGzeCAQN3mqjOszUyqs7H85EyuUK-xlGCqzWasxsCDhxVRmNGs2W0JquBBeEke4cOCWnV7PwRtU2n4O2FOuCu92e71URJ+5k0pkeiAujvOsZjPnIoA

💻 Code

export type ExtractRouteParams<T extends string> = string extends T
  ? Record<string, string>
  : T extends `${infer _Start}:${infer Param}/${infer Rest}`
  ? { [k in Param | keyof ExtractRouteParams<Rest>]: string }
  : T extends `${infer _Start}:${infer Param}`
  ? { [k in Param]: string }
  : {};

declare function link<T extends string>(
  args: {
    path: T;
  } & ({} extends ExtractRouteParams<T> ? { params?: {} } : { params: ExtractRouteParams<T> }),
): void;

// error in ts 5.3.3 and 5.4.0-beta
link({ path: '/:foo/:bar', params: {} });
// error in ts 5.3.3, no error in ts 5.4.0-beta
link({ path: '/:foo/:bar/', params: { foo: 'foo' } });

🙁 Actual behavior

The last line is only an error in TS 5.3.3, not TS 5.4-beta.

🙂 Expected behavior

I'd expect it to be an error in both (foo isn't specified).

Additional information about the issue

Putting NoInfer in the conditional type makes the expected error reappear:

- } & ({} extends ExtractRouteParams<T> ? { params?: {} } : { params: ExtractRouteParams<T> }),
+ } & ({} extends ExtractRouteParams<NoInfer<T>> ? { params?: {} } : { params: ExtractRouteParams<NoInfer<T>> }),

So perhaps this is WAI? Just wanted to flag it as a change I noticed from 5.3.3 to 5.4-beta.

bisected to #56515

I can confirm that inferToMappedType is now able to infer more because the constraintType that it receives is a type parameter (T) and not an intersection like before (T & string). The mentioned intersection was reduced to just T as per #56515 so I think this is working as intended.

The NoInfer looks like a good workaround to this issue because that's precisely what it is useful for - it limits the inference sources and #56515 made the second argument a viable inference source here.

I feel like this is the same problem that breaks the inference in MikroORM for the populate hints (and many other things that work on the same basis), it starts to break with 5.4.0-dev.20231121, previously I was able to infer string literals and now I only get string which breaks other types. I'll try to work on a playground when I have some spare time and open a separate issue, but the types are pretty complex so it might take some time.

If this is considered WAI, I am honestly scared... 🤣 The use case on our end is more complex than just validation, we need inference so the types can be used in the return values.

Will work on the repro and open a new issue soon'ish, just wanted to bring this up, as I feel a storm is coming (already got a report about this some time ago).

edit: solved via mikro-orm/mikro-orm#5197

The difference here is in the type parameter inference. In 5.4 we get

"foo" | "/:foo/:bar/"

whereas in 5.3 it's only

"/:foo/:bar/"

Generic inference has always been willing, in other cases, to produce a string literal union:

declare function choose<T extends string>(x: T, y: T): T;
// p: "a" | "b"
const p = choose("a" as const, "b" as const);

so it's at least somewhat apparent that the 5.3 behavior is the incorrect (or at least inconsistent) one.

This issue has been marked as "Not a Defect" and has seen no recent activity. It has been automatically closed for house-keeping purposes.

Navigation Menu

Repository navigation

[5.4 regression] Conditional type changes behavior with 5.4-beta #57221

🔎 Search Terms

🕗 Version & Regression Information

⏯ Playground Link

💻 Code

🙁 Actual behavior

🙂 Expected behavior

Additional information about the issue

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[5.4 regression] Conditional type changes behavior with 5.4-beta #57221

Description

🔎 Search Terms

🕗 Version & Regression Information

⏯ Playground Link

💻 Code

🙁 Actual behavior

🙂 Expected behavior

Additional information about the issue

Activity

Andarist commented on Jan 29, 2024

Andarist commented on Jan 29, 2024

B4nan commented on Jan 31, 2024

RyanCavanaugh commented on Feb 1, 2024

typescript-bot commented on Feb 4, 2024

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions