Better handling of this mismatches

[mqudsi]

TypeScript Version: 2.8.3

Search Terms: this, intitle:this, this bind

function assert(expr: boolean, msg: string = "") {
	if (!expr) {
		console.log("assertion failed!");
		throw msg;
	}
}

class Foo {
	_bar: Bar;

	constructor() {
		this._bar = new Bar();
		this._bar.Callback = this.callback;
	}

	public run() {
		this._bar.Callback();
	}

	callback() {
		assert(!(this instanceof Bar), "\"this\" has unexpected type \"Bar\"");
		assert(this instanceof Foo, "\"this\" has unexpected type!");
		console.log("Test passed, \"this\" retains the expected type.");
	}
}

class Bar {
	public Callback: () => void;
	public run() {
		this.Callback();
	}
}

var f = new Foo();
f.run();

I realize the current behavior is by-design, this is a suggestion/request-for-comment to improve type safety here.

Expected behavior:

The code above (playground link) was written with the (incorrect/naive) assumption that at line 13 (this._bar.Callback = this.callback;) the function reference would be stored bound to the value of this at the time of execution of that line.

I understand why TypeScript can't ditch this horrible behavior and evaluate the code sanely for reasons of backwards compatibility when changing JS file extensions to .TS, but as this is a common type-safety issue, I feel like TS should take measures to mitigate the issue.

TypeScript should be able to catch the unsafe usage of this in the example above (it fails the assertions).

Additionally, when typing this.<TAB>, the suggestions generated are for values based on the current scope instead of being any or something else, leading one to believe the code is correct.

Suggestions:

• Perhaps a --strictThis option generating an error when this.xxx is assigned to an object of a different type from the type of this in the current scope?
• A new keyword, e.g. self, the value of which binds to the current value of this that can be used instead.

Actual behavior:
The line this._bar.Callback = this.callback; does not resolve the value of this at the time it is invoked, leading to the failure of the assertion.

Playground Link:
Link

Related Issues:

[mhegazy]

duplicate of #10288

[mqudsi]

Thanks, Mohamed! If it's OK with you, I'll copy my comments there.