Allow implicit call with class with no body

[helixbass]

Fixes #5052

Tweaks addImplicitBracesAndParens() to correctly close an inline implicit call which has an argument that's a class with no body

So now eg f class A compiles

The issue was that it was generating the implicit opening ( but failing to generate the closing ) at the end of the line since when it saw class it pushed a CONTROL token that expected to see an INDENT-OUTDENT pair before unblocking the closing of the implicit call

[Inve1951]

this syntax for a function call used in the test is currently a hack

f
  b: 1
  class A
  a

[helixbass]

@Inve1951 what do you mean exactly? I included that test case to check that in a multiline implicit call (with an implicit object first argument) the new code didn't trigger a premature ending of the implicit call

[Inve1951]

@helixbass The only time where the compiler allows for a function call where the arguments start on a new line is when the first argument is an implicit object. All of the following examples throw at compile time with unexpected indentation:

f
  a

f
  class A

f
  1
  b: 3

f
  {b: 3}

[helixbass]

@Inve1951 yup the fact that an indented implicit object as the first arg can start an implicit call is a special case. That's why I thought it was worth testing since it's a different way for an implicit call and an "inline" class to interact

[GeoffreyBooth]

@Inve1951 or anyone else, is this okay to merge in?

[Inve1951]

I'm good with it. My only concern is that the test relies on the implicit object starting a call, which I'm not sure about whether that's a supported feature or not. But I'm certainly in favor to have that as a language feature. In fact I'd expect the same from the broken cases I demonstrated in my previous comment.

[vendethiel]

My only concern is that the test relies on the implicit object starting a call, which I'm not sure about whether that's a supported feature or not.

It's very much a special case. Your "broken cases" do not compile on purpose.

[GeoffreyBooth]

@vendethiel are there any issues with this PR, or is it okay to merge in?

[vendethiel]

I'm very bad at rewriter stuff, it the tests pass then I'd say go for it.