Unary negation on an object should be a type error

[danvk]

🔎 Search Terms

• unary negation object

🕗 Version & Regression Information

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

This is the behavior for TS 3.3, TS 5.3 and presumably all other versions.

⏯ Playground Link

https://www.typescriptlang.org/play?ts=3.3.3#code/MYewdgzgLgBCBGArGBeGBvAHgLhgRgBoYBPXAJgF8BuAKFElgAcAbAVwlRgGoFFb7oMALYBLMO04BaXv3AQQzAKYA6ZiADmAChbsio8RACUVGDAD0ZmADkAhlaK2rNIA

💻 Code

const obj = {x: 1, y: 2};
const plus = +obj;  // ok!?
const minus = -obj;  // ok!?
console.log(plus, minus);  // NaN, NaN

🙁 Actual behavior

No type errors, logs NaN, NaN.

🙂 Expected behavior

I'd expect both +obj and -obj to be type errors. Applying unary negation or unary plus to an object is likely an error and produces a useless result.

Additional information about the issue

Thanks to JavaScript's coercive behavior around operators, I believe this is interpreted the same as -Number(obj) and Number(obj) is NaN. Maybe that should be a type error, too, but the unary operators seem more clear-cut.

Using binary operators (obj + obj) is already a type error, like you'd expect.

I ran into this in real code today when I refactored a type from:

- type Value = number;
+ type Value = number | { currency: '$' | '€'; value: number; };

I had code like -value elsewhere and was surprised that I didn't get a type error.

[MartinJohns]

Duplicate of #20131. Used search terms: unary in:title

[danvk]

It's an oldie! It makes sense that this is less clear-cut than I thought because of Date and libraries like math.js.

Can't believe the word "negation" didn't appear anywhere in that issue!