Document additional optional braces conditions

[tgodzik]

Compiler version

3.0.0-RC2

Minimized code

The below examples work with optional braces, but are undocumented, which caused us to skip it while implementing the parser:

  for{ a <- 0 to 11 }
      println(a)
      println(a)

  for( a <- 0 to 11 )
      println(a)
      println(a)

  while( a < 10)
      a += 1
      println(a)

Are we missing anything?

Expectation

We should have optional braces properly documented, otherwise it's hard to build tooling around it, but also it might cause suprises to the users.

[bishabosha]

I would say it is documented in the syntax http://dotty.epfl.ch/docs/reference/syntax.html:

Expr1             ::=  ...
                    |  ‘while’ ‘(’ Expr ‘)’ {nl} Expr
BlockExpr         ::=  <<< (CaseClauses | Block) >>> // BlockExpr <- SimpleExpr <- PrefixExpr <- InfixExpr <- PostfixExpr <- Expr1 <- Expr
<<< ts >>>        ::=  ‘{’ ts ‘}’
                    |  indent ts outdent
<<< ts >>>        ::=  [nl] ‘{’ ts ‘}’
                    |  `:` indent ts outdent

[tgodzik]

I haven't noticed it there 🤔 and also it's still kind of weird to have the additional do, then keyword, but still be able to just not use them :(

I would however add it here for the users: http://dotty.epfl.ch/docs/reference/other-new-features/indentation.html

[tgodzik]

Ok, thanks @bishabosha for pointing me to the syntax, I haven't noticed that it's actually included there. I thought that the optional braces page was better to look at. It seems we don't have anything else besides these 3 I mentioned.

[odersky]

Good point. The optional braces after for(..._) are probably an oversight. At some point we allowed optional braces after
if (...) as well but then we retracted since we said that was just too much choice at the boundary between new and old syntax. It seems we overlooked for (...). I think we should probably be uniform and either accept indentation after if (...) again or drop it after for (...). And among these two choices my slight tendency at the moment would be to drop it after for (...). But for 3.0 everything should stay as it is, and we should get more experience what users think before making the next moves.

[tgodzik]

Good point. The optional braces after for(..._) are probably an oversight. At some point we allowed optional braces after
if (...) as well but then we retracted since we said that was just too much choice at the boundary between new and old syntax.

I think that we decided that if (cond) <indentation> should stay at least until 3.1 ?

This works in 3.0.0-RC2:

  if (true)
    println(1)
    println(2)

We can add scalafix rules for all the syntax changes by then, so in most cases this should be straightforward to migrate.

[odersky]

Ah, yes, you are right of course. I got my versions crossed. So yes, everything behaves in the same way and we are good.

[odersky]

I actually don't see a discussion of indentation after if (...) in http://dotty.epfl.ch/docs/reference/other-new-features/indentation.html either. @tgodzik do you agree that we need to document that case as well?

[tgodzik]

Sure! That makes sense.