Type Instantiation is Excessively Deep Error Regression in #37348

🔎 Search Terms

type instantiation is excessively deep, unused type parameter, unused tuple element, #37348

🕗 Version & Regression Information

This changed in PR #37348

⏯ Playground Link

https://www.typescriptlang.org/play/?#code/CYUwxgNghgTiAEkoGdnwPoEkB2AXEM2UEAKgJ4AOIAIlLlALID2oE8A3gFDw-wVwVYNAgEsAbiGC16ALngBtKNjIAaeErIBdANycAvrs6gkcRNFTxyVaYxYgIAHm68A9C-iYAZvFwALEWi4lAiCMFAAtiD4MPAB8KxRkj6+CGAiMGAArtAwIkHwAOZMIGhQAO5QZAB0zjwAqtiZyJIqtfDUohJSdFCtAHzwIAAe+NjAaFh4BESkwTbMrBxt-CChwrldNnJ+ccOj45ZzPQv2DhpqItieBO2dkjZ9bbzwAPzwbpb+aM1gTGODEBAkTwsWwyQQuEyFEBsVKEGQTHg2HAJWQsDI8E8TBiCMi8DgKD+NWeJN48h2yDUGk0T14cg0uj0QA

💻 Code

declare class _InternalTypeDataModel {
    prepareDerivedData: [any, any];
};

declare class TypeDataModel<
    // If this type parameter is deleted the circularity goes away.
    Unused,
    DerivedData,
> extends _InternalTypeDataModel {
    prepareDerivedData: this extends TypeDataModel<any, infer DerivedData>
        ? // This second element in the tuple is also necessary for some reason.
            [this, any]
        : any;
}

🙁 Actual behavior

Type instantiation is excessively deep and possibly infinite.

🙂 Expected behavior

No error.

Additional information about the issue

I understand the weirdness in this minimized code but the actual code I was reducing actually does useful things and doesn't really seem so contrived in context. In actual code all type parameters and tuple elements actually do useful things but this reproduction shows it doesn't matter if they're used, just that they exist, for some reason.

Start with the provided Playground reproduction and compare its behavior with the change introduced by PR #37348. Trace the type-instantiation path responsible for the excessive-depth diagnostic, then verify that this example and the reported regression no longer produce an error.