Repository navigation
Functions with intersection arguments are not always typed coherently #9514
Description
TypeScript Version: 1.8.10
Code
// tl;dr:
// This should give a static error, but does not.
var monkeyAnalyzer: (input: Monkey) => void = (input: Monkey & Robot) => { console.log(input.FurColor, input.MotorCount); }
// Will definitely fail, as we can expect from the incompatible signatures of the declared type of monkeyAnalyzer and what we actually assign to it
monkeyAnalyzer(new Monkey());
// Verbose example
class Monkey
{
public FurColor: string;
}
class Robot
{
public MotorCount: number;
}
class MonkeyRobotAnalyzer
{
public static Analyze = (input: Monkey & Robot): void =>
{
console.log(input.FurColor, input.MotorCount);
}
}
class MonkeyManager
{
constructor(monkeyAnalyzer: (input: Monkey) => void)
{
monkeyAnalyzer(new Monkey());
}
}
MonkeyRobotAnalyzer.Analyze(new Monkey()); // Shows error as it should
var myManager = new MonkeyManager(MonkeyRobotAnalyzer.Analyze); // No error, but should not be allowed, signature of MonkeyRobotAnalyzer.Analyze is not compatible with (input: Monkey) => voidExpected behavior:
The incompatible signatures should result in a static error
Actual behavior:
We get no static error
Activity
HerringtonDarkholme commentedon Jul 5, 2016
This is a by-design behavior because of bivariant function, as stated in FAQ
Here you are using
(input: Monkey & Robot) => voidas(input: Monkey) => void, and the latter is assignable to the former. So the function call just type checks.Elephant-Vessel commentedon Jul 5, 2016
Oh my, would it be possible to get this as a compiler flag? This led to a lot of code being faulty and in need of refactoring in a system I'm working on due to a minor mistake in the type declaration of an input parameter.
It seems that the motivation for this behavior is to enable useful hacks at the cost of type safety. Isn't typescript all (well mostly) about the opposite? A small compiler flag at least? Pretty please?
HerringtonDarkholme commentedon Jul 5, 2016
Please see this discussion.
#1394