Incorrect handling of `end <id>`

[Glavo]

Compiler version

scala-3.0.1-RC1 (83acd4b)

Minimized code

This code shouldn't have been compiled, but it did:

class C {
}
end C

And automatic rewriting (-rewrite -no-indent) does not delete the end. For this code:

class C:
  val foo: Int = 0
end C

After compiling with scalac -rewrite -no-indent:

class C {
  val foo: Int = 0
}
end C

[som-snytt]

Braces and end markers are not mutually exclusive because:

Braces are not much better since a brace by itself also contains no information about what region is closed.

I don't have an opinion about behavior under -rewrite, but -endless would be a fun option. One could also pun about the "endless bikeshedding".

[tisonkun]

Braces are not much better since a brace by itself also contains no information about what region is closed.

@som-snytt most of modern editor should be able to help you find the information about what region is closed, and fold it.

[noti0na1]

@tisonkun It is difficult to balance how much information the source code carries and how much information the editor should display. Personally, I prefer using } over end when then the ending point is unclear.

[som-snytt]

Sorry, I didn't cite the reference doc for end: https://dotty.epfl.ch/docs/reference/other-new-features/indentation.html#the-end-marker

I understand the quote merely to imply that end is not intended as an alternative to brace, so I think the example ought to compile.

I agree there are varying preferences about the syntax. However, the greatest passions are reserved for comments, and while I am now a satisfied user of the end marker, I shall always recoil at the end comment:

class C {
  LOLOC()  // lots of lines of code
} // end C

It would be especially diabolical if the rewrite not only didn't remove the end marker, but put it behind a line comment after the brace.

[som-snytt]

Actually the bug is that empty template is not patched:

class C:
end C

or equivalently, empty template with colon must be disallowed.

#12185 (comment)

Source must compile if (optional) end marker is deleted.

#12918 (comment)

[som-snytt]

OK

package p {
}
end p

but not

package p:
end p

or

package p
end p

nor

package p
class C
end p

[som-snytt]

This was fixed to make it a code comment in #12954 but PR was not linked.

I shall always recoil at the end comment

is obviously no longer the case.