Two small changes to the pattern matching spec

[gafter]

Per #206 (comment)

The use of a pattern in a case was intended to be an extension to the existing alternative of a constant expression. That should be clarified, though. In addition, it was intended that a case-guard could be used with a constant-valued case. I will amend the spec to clarify.

The goto case statement would be relaxed in two ways. First, the expression no longer needs to be a constant. Second, the expression no longer needs to correspond directly to a case label. Logically the goto case statement would take an expression and dispatches the whole switch statement with the new expression. This needs to be added to the spec too.

[HaloFour]

@gafter

Example:

int i = 1;
bool b = false;
switch (i) {
    case 2 when b == true:
        Console.WriteLine("case 2 when b == true");
        break;
    case 2:
        Console.WriteLine("case 2");
        b = true;
        goto case (i + 1);
    default:
        Console.WriteLine("default");
        goto case 2;
};

Would print:

default
case 2
case 2 when b == true

[gafter]

@HaloFour That is a useful example for the spec, thanks!

[bondsbw]

Curious, does the following loop on i, or break in one of the other locations?

int i = ...;
switch (i)
{
    case i - 1:
       Console.WriteLine("i - 1");
       break;
    case i:
       Console.WriteLine("i");
       goto case i++;
    case i + 1:
       Console.WriteLine("i + 1");
       break;
}

Bigger question is, does it do what it really should do? Is there a preference?

EDIT: changed example to include i - 1 case

[alrz]

@bondsbw A constant-pattern doesn't accept a general variable, just "a literal, the name of a declared const variable, or an enumeration constant". However, earlier in the spec, it is defined as a shift-expression which is not correct I guess.

[gafter]

@alrz this is a constant expression, not a constant pattern. That is an existing production in the language grammar not intended to be changed by the pattern-matching spec. In any case, you are correct to say that you cannot have an expression that is not a constant.

[alrz]

@gafter Yes, I meant that, well, constant-pattern defined as a constant-expression too, according to the spec. But in the "Patterns" section it is defined differently which I believe, is not intended.

[bondsbw]

Ah, I misinterpreted "First, the expression no longer needs to be a constant." I thought that was intended for both goto case and case.

[gafter]

Fixed in branch features/patterns.