Inferred never values don't properly type guard

[a-non-a-mouse]

🔎 Search Terms

never type guard

🕗 Version & Regression Information

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about type guarding

⏯ Playground Link

https://www.typescriptlang.org/play/?noUncheckedIndexedAccess=true&moduleResolution=99&target=99&module=199&exactOptionalPropertyTypes=true&ts=5.6.3#code/GYVwdgxgLglg9mABAExgUwBQEoBcixoBuaATogN4BQiNiJaUIJSYIANm4gIYDO+RpANyUAvpUqhIsBIhhhgpACLpsFarXqNmKFVmFjKyNBDZd6iCAh5READzzWScgOaIAPvnZthlGMEQYAIS2WGq0svJKuvritoKIAPQJdrJ81jAciAAOcDw8MABGbACenhzifgHBoVThqJh6orHxSSlgcADuFlxgAOQ2BWhlbEA

💻 Code

function die(): never {
    return null as never;
}

function inferDie() {
    return die();
}

declare const x: string | null;

if (!x) {
    inferDie();
}

x; // x is still possibly null

if (!x) {
    die();
}

x; // x now can't be null

🙁 Actual behavior

Two functions that have "identical" signatures, () => never, are not acting the same

🙂 Expected behavior

I'd expect both of these to act the same way since, for everything I can see, TS sees their signatures as being identical.

Additional information about the issue

No response

[Andarist]

as per #32695 :

• the call occurs as a top-level expression statement, and
• the call specifies a single identifier or a dotted sequence of identifiers for the function name, and
• each identifier in the function name references an entity with an explicit type, and
• the function name resolves to a function type with an asserts return type or an explicit never return type annotation.

Your inferDie has no explicit never return type so it's not treated as an assertion

[a-non-a-mouse]

Yes, I realize the why, but there is no way whatsoever to tell what a function is going to do in this case without literally looking at the source of it.

TS shows that both are () => never. So if you look at a tool tip or whatnot in any editor, you have no idea what the code is actually going to do.

[a-non-a-mouse]

I'd argue that this behavior is wrong, but alternatively tooling should be able to tell the difference. As far as I can tell, TS doesn't actually expose the difference.

[MartinJohns]

The behavior is intentional, not wrong, and was done for performance reasons.

But it would probably be good to add this to the FAQ. And to the handbook, which is covered by #59930.

[a-non-a-mouse]

Oh ffs, I'm done reporting issues to you guys. You make stupid ass decisions and then try to justify them. It's utterly pathetic.

[MartinJohns]

Neither me nor Andarist are part of the TypeScript team. But you can be assured that the team does not make these decisions willy-nilly or out of spite, there's reasons behind them. They're not "stupid ass decisions", they're technical decisions with benefits and drawbacks.

And you're advised to check out the Code of Conduct. :-)

[a-non-a-mouse]

Oh nooo, you're offended.

Like I give a single fuck