SIP-77 - End Markers for Method Blocks

[soronpo]

No description provided.

[bishabosha]

Couple of examples:

• Metals (via SemanticDB) - references to method symbol to include the new end marker
• Presentation Compiler? if that also reports symbol occurrences in source files

[prolativ]

Suggested change

	- **Implicit `apply` calls**: Use the name of the object/class that owns the `apply` method when it's called implicitly.
	- **Implicit `apply` calls**: Use the name of the object/class instance that owns the `apply` method when it's called implicitly.

I believe this was intended for cases like

class Foo:
  def apply(block: => Unit): Unit = ()

val foo = new Foo

foo:
  // do something
end foo

[soronpo]

Yes, I'll add this example

[prolativ]

What about curried methods as they were not mentioned explicitly here? E.g. should this work?

def foo(bar: String)(baz: String) = ???

foo("abc"):
  "xyz"
end foo

And similarly

class CurriedFoo(bar: String):
  def apply(baz: String) = ???
  
def foo(bar: String) = CurriedFoo(bar)

foo("abc"):
  "xyz"
end foo

[soronpo]

E.g. should this work?

Yes. Nothing explicitly is speced to prevent such application.

[odersky]

End markers for method calls fill an important hole.

End markers for new are in fact already implemented. This works

val foo = 
  new Foo:
    // do something
    // overrides, vals, defs...
  end new

And so does this:

val foo = new Foo:
  // do something
  // overrides, vals, defs...
end foo

I don't think we need another case were we also allow end new in the second case.

I am not sure about disambiguation:

def foo = bar:
  ...
end bar
end foo

is this supposed to work? It would be the first instance where we allow multiple ends on the same indentation level.

[bjornregnell]

I don't think end on the same indentation level should work as it does not give a strong visual clue on the nesting order.

[soronpo]

As discussed, I opened a ticket for end new not working under all conditions: scala/scala3#24250 and I'm removing anonymous class cases from this SIP.

[prolativ]

I think all the examples we give (except for some counterexamples) should be valid scala code (currently or after the SIP would become a part of the language). This snippet below is not one, because fewer-braces syntax requires non-empy code blocks (and comments don't count as code in this sence).
The same applies for some other examples below

[soronpo]

Implementation scala/scala3#24251