Regression in 2.8.0+ with ternaries in a lambda

[boris-petrov]

Prettier-Java 2.8.1

Playground link

--print-width 50

Input:

public abstract class Foo {
  public String foo() {
	foo(() -> aaaaaaaaaaa &&
		bbbbbbbbbb &&
		cccccccccccccc
			? 1
			: 2);
  }
}

Output:

public abstract class Foo {

	public String foo() {
		foo(() ->
			aaaaaaaaaaa &&
			bbbbbbbbbb &&
			cccccccccccccc
				? 1
				: 2
		);
	}
}

Expected behavior:

public abstract class Foo {

	public String foo() {
		foo(() ->
			aaaaaaaaaaa &&
				bbbbbbbbbb &&
				cccccccccccccc
				? 1
				: 2
		);
	}
}

Similar to #802.

cc @jtkiesel

[jtkiesel]

While this did change between versions, it seems the new behavior is how Prettier prints the equivalent input in TypeScript.

[boris-petrov]

Oh wow, but... why? How does this make sense? Do you think this is a bug in Prettier?

[jtkiesel]

I don't think so, I'm guessing it's done this way because as soon as a second parameter is added to the function, the entire lambda is indented (example), so there isn't really a need to indent the binary expression to make it obvious where each method parameter begins (which is my guess as to why the binary expression of a ternary is indented when part of a function).

[boris-petrov]

My idea was that every time there is a "continuation" of a line on a new line - it should be indented. Which is true in practically all cases. I think this is the first case where that doesn't hold. But maybe I've had a wrong expectation all along. Anyway, thanks for the discussion and sorry for the fuss!