Microsoft/TypeScript
View on GitHubRecursive tuple cannot be computed through generic, but valid in a declarative form
Open
#62704 opened on Nov 1, 2025
BugDomain: check: Type CircularityHelp Wanted
Description
🔎 Search Terms
ts2589 "Type instantiation is excessively deep and possibly infinite"
🕗 Version & Regression Information
- This is the behavior in every version I tried, and I reviewed the FAQ for entries about ts2589
- I was unable to test this on prior versions before v4.7.0 because
infer T extends Rsyntax not introduced
⏯ Playground Link
💻 Code
type Foo = {
leading: 'a',
rest: [Foo],
};
type Expand<T> = T extends {
leading: infer L;
rest: infer R extends any[]
} ? [L, ...ExpandMany<R>] : never;
type ExpandMany<T extends any[]> =
T extends [infer First, ...infer Rest] ? [Expand<First>, ...ExpandMany<Rest>] : [];
type TestA = Expand<Foo>;
// ^~ Type instantiation is excessively deep and possibly infinite.(2589)
type TestB = ['a', TestB];
🙁 Actual behavior
TestA meant to be:
TestA
-> Expand<Foo>
-> ['a', ...ExpandMany<[Foo]>]
-> ['a', ...[Expand<Foo>]]
-> ['a', Expand<Foo>]
-> ['a', TestA]
which is identical to TestB but TS complains the instantiation cannot stop.
🙂 Expected behavior
Stop expansion at first reference at appeared structure Expand<Foo>, that is:
type TestA = Expand<Foo>;
// ^?~ type TestA = ['a', Expand<Foo>]
// or more intelligent `type TestA = ['a', TestA]`
Additional information about the issue
This code is useful when Foo is as const inferred from a runtime constant, and we want a tuple-like typing calculated from Foo. Since TypeScript support self-referential tuple by declarative just like TestB, it would be better for support that in a computed way like Expand<Foo>.