infer R extends in template literal does not narrow correctly when extending union of literals

[ehoogeveen-medweb]

Bug Report

🔎 Search Terms

template literal infer extends union narrowing

🕗 Version & Regression Information

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about infer extends in template literals.

⏯ Playground Link

Playground link with relevant code

💻 Code

// Works as expected.
type GetGenericPrefix<Str extends string> = Str extends `${infer pre extends string}${'!' | '?'}` ? pre : '';
type someString = GetGenericPrefix<'foo?'>; // The expected literal type 'foo'.

// Compact, but doesn't work.
type GetKnownSuffix1<Str extends string> = Str extends `${string}${infer suf extends '!' | '?'}` ? suf : '';
type knownSuffix1 = GetKnownSuffix1<'foo?'>; // The union '!' | '?' instead of the expected literal type '?'.

// Works, but not scalable.
type GetKnownSuffix2<Str extends string> =
  Str extends `${string}${infer suf extends '!'}`
  ? suf
  : Str extends `${string}${infer suf extends '?'}`
  ? suf
  : '';
type knownSuffix2 = GetKnownSuffix2<'foo?'>; // The expected literal type '?'.

🙁 Actual behavior

Type GetKnownSuffix1 returns the entire extended union ('!' | '?') instead of the matching literal ('?').

🙂 Expected behavior

Type GetKnownSuffix1 should return just the matching literal ('?'), the same way infer pre extends string in GetGenericPrefix returns the literal type 'foo' and not the extended type string.

[whzx5byb]

Related: #49839 (comment)

[whzx5byb]

Workaround: use distribution behavior to simplify the "not scalable" code

type GetKnownSuffix<Str extends string, Suffix extends string> = 
  (Suffix extends infer S extends string
    ? Str extends `${string}${infer R extends S}`
      ? R
      : never
    : never
  ) extends infer T ? [T] extends [never] ? '' : T : never;
    
type knownSuffix = GetKnownSuffix<'foo?', '!' | '?'>
//   ^?  "?"
type knownSuffix2 = GetKnownSuffix<'foo', '!' | '?'>
//   ^?  ""

Playground

[RyanCavanaugh]

This seems to be a manifestation of the same problem described at #49839 (comment)

[typescript-bot]

This issue has been marked as a 'Duplicate' and has seen no recent activity. It has been automatically closed for house-keeping purposes.