Question on sealed types in pattern switch

[Brian Goetz]

> When a sealed type is used in a pattern switch, AFAIK the current
> proposal is intended to make it exhaustive with nulls rejected (just
> like enums):

That was the theory.  But, as it turns out, the analogy with enums is 
imperfect in exactly this way.  Still assessing the alternatives.

>   Shape s = ... // Shape is a sealed type of Circle, Rectangle and Hexagon
>   var result = switch (aShape) {
>     case Circle c -> ...
>     case Rectangle r -> ...
>     case Hexagon h -> ...
> }
>
> Being exhaustive and rejecting null seems like eminently desirable
> behaviour. here.

Yes, and so far, no problem.  None of the patterns is total, there's no 
`case null`, so, like with enums, null will be rejected by the switch 
with no rule changes.

> But what does this simple refectoring do?

The short answer is it is not necessarily as simple a refactoring as you 
think.

As I said, still assessing the alternatives.  Stay tuned.