[Feature Request] The type checker should be able to infer string values of fields being switched on and match them to string literal types

[JonathanDCohen]

Bug Report

🔎 Search Terms

switch
"string literal"

🕗 Version & Regression Information

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about switch and "string literal"

⏯ Playground Link

Playground link with relevant code

💻 Code

interface I {
    field: string;
}

interface A extends I {
    field: "a";
}

function f(i: I) {
    switch(i.field) {
        case "a":
            /*
                Argument of type 'I' is not assignable to parameter of type 'A'.
                Types of property 'field' are incompatible.
                Type 'string' is not assignable to type '"a"'.
            */
            takesA(i);
    }
}

function takesA(a: A) {}

🙁 Actual behavior

Type checking fails to be able to infer the string value in the passed argument matches the string literal value in the type

🙂 Expected behavior

The type checker should know that inside the case statement, i.field === "a" and pass the type check

[MartinJohns]

The types are not sealed. Just because there's "a" in that field, doesn't mean you're having a A. Your suggestion would be unsound and lead to bugs.

[jcalz]

Duplicate of #16976, I think.

Narrowing the type of an object by checking a property only happens when the object is a discriminated union type and the property is its discriminant. But I is not a discriminated union (or any kind of union), so checking the field property only serves to narrow that property and not the whole object.

[MartinJohns]

@jcalz I don't see how 16976 is related.

[jcalz]

In the example, the only difference between A and I is the type of the field property; once you narrow i.field to "a", then i is definitely an A. Indeed, you can write takesA({ ...i, field: i.field }) and the compiler is happy about it. It only has a problem with takesA(i) because narrowing i.field does not serve to narrow i, which is #16976.

[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.