Mark "case" keyword as optional in pattern "match"

[jaisw7]

Instead of

val result = x match {
  case 0 => "Zero"
  case n if n > 0 => "Positive"
  case n if n < 0 => "Negative"
  case _ => "Unknown"
}

I believe

val result = x match {
  0 => "Zero"
  n if n > 0 => "Positive"
  n if n < 0 => "Negative"
  _ => "Unknown"
}

is more concise.

I was wondering if there is a particular reason for using case keyword e.g., it makes the grammar unambiguous? If not, can we keep it as optional? I can submit a patch if everyone is aligned.

[som-snytt]

The RHS is a block; case tells you where it ends without braces. I proposed omitting the wildcard, case => "Unknown", that was already too much (or too little). There was a pre-sip for x match p to mean case p => true case _ => false. I think there is an open discussion about improving guard syntax (for disjoint cases).

[jaisw7]

Got it. I realized that OCaml and Rust uses symbols for disambiguation:

let result = match x with 
|  0 -> "Zero"
|  n when n > 0 -> "Positive"
|  n when n < 0 -> "Negative"
|  _ -> "Unknown"

let result = match x {
    0 => "Zero"
  , n if n > 0 => "Positive"
  , n if n < 0 => "Negative"
  , _ => "Unknown"
}

I wish Scala was as minimal as Ocaml and/or Rust.

[spamegg1]

I also felt as @jaisw7 initially about the case keyword, but my view has changed over time. Sometimes things can be a bit too concise. Having a meaningful, readable (and still short) guard is nice. So many easy-to-make mistakes with forgetting SML / OCaml's |, not to mention dealing with the weird alignment / indentation.

As another, and arguably more important, defense of case, it creates conceptual consistency, which is especially useful in teaching. case being used in pattern matching, enums and case classes make things very clear and easy to understand. Moreover it perfectly matches (no pun intended) mathematical proof-writing where the word case is used, which makes the code match (again) the proof one-to-one.