Switch syntax; pedagogical nightmare?
Brian Goetz
brian.goetz at oracle.com
Fri Aug 11 16:26:32 UTC 2017
I would call this “mostly trivial” :) The problem with this example is that it falls afoul of definite assignment; c and d are not definitely assigned at the point of use. We already make extensive use of definite assignment analysis; this is a simple application of this concept.
> On Aug 3, 2017, at 8:25 AM, Tagir Valeev <amaembo at gmail.com> wrote:
>
> Actually fall through case is not so trivial. Consider
>
> switch (x) {
> case Foo(var a, var b):
> System.out.println(a+":"+b);
> //fallthru
> case Bar(var c, var d):
> System.out.println(c+":"+d);
> break;
> }
>
> So when we fall through, c and d are not defined. Thus this code should result in compilation error. Seems that the scenarios when fall through is actually possible are quite rare.
>
> Tagir Valeev.
>
> 3 авг. 2017 г. 6:43 AM пользователь "Brian Goetz" <brian.goetz at oracle.com <mailto:brian.goetz at oracle.com>> написал:
>
> > I just watched the Java One talk by Brien, and can't help but feel that
> > overloading the 'switch' keyword for pattern matching is a hell of a
> > mistake. I realize the appeal in that it doesn't introduce a new keyword
> > due to backwards source compatibility, but feel like it's a pedagogical
> > nightmare.
> >
> > Having the break;s left off is clearly desirable for pattern matching that
> > involves destructing, but you are going to be leading so many newbies to
> > using switch statements (as opposed to expressions) without break,
> > expecting it to behave similar to the new expression usage of switch.
>
> I think you have misinterpreted what is being proposed. Statement switch with patterns has exactly the same control flow behavior as it does without patterns — there is no relationship between whether the case labels are patterns and the break/fallthrough behavior.
>
> Similarly, if we introduce a switch *expression* (whose syntax would differ from statement switch), again this has nothing to do with patterns; it will behave the same way with patterns and without.
>
> So, the sky is not falling ;)
>
More information about the amber-dev
mailing list