Rewrite to no-indent keeps end comment

[som-snytt]

Preserve end marker as end comment, since if it was worth having an end marker, it's worth commenting.

class C:
  def f = 42
end C

is rewritten to

class C {
  def f = 42
} // end C

[dwijnand]

LGTM

[som-snytt]

Deprecating "empty colonized template bodies" may be undesirable because accepting it was intentional; however, it is an excuse to use the phrase "empty colonized template bodies" outside a sci-fi context or academic paper.

[dwijnand]

Could you propose that separately then? Personally I don't like it, but I also don't like that class C: is bad syntax too - and changing that would make your consistency point.

[dwijnand]

^

[som-snytt]

@dwijnand Rebased and kept only the feature described.

I'll propose deprecating empty body after colon separately; I don't feel strongly about it, except that "end marker is optional" strikes me as a useful rule that should not be violated lightly. (Note to self for later.)

Also dropped converting the converse, end comment to end marker, because it was just a heuristic that did not verify the identifier. Probably auto-end-marking is a more useful and consistent feature, for which there is a ticket; maybe the conversion can be done then.

[dwijnand]

LGTM, thanks Som

[adpi2]

I suspect this to be the root cause of #17187

object A:
    object B:
        def a = 2

When rewriting to no-indent, we want to add }\n and then }\n at the exact same position. But this piece of code only keep the second patch, which leads to the incorrect:

object A {
    object B {
        def a = 2
}

[adpi2]

Tested also this:

class C:
  def f = 42

end C

And it is rewritten to this:

class C {
  def f = 42
}

end C

[som-snytt]

@adpi2 it looks like it introduced "replace" semantics for "patch at a point". I don't remember now why. There is a tweak to "overlaps" to include overlaps if exactly one point.