JEP proposed to target JDK 12: 325: Switch Expressions (Preview)

[Ben Evans]

I tend to agree with Stephen. In particular, as Brian has so correctly
pointed out, most Java developers over-focus on syntax.

Therefore, it seems likely that, despite marking a syntax-introducing
feature as Preview, early adopters will come to rely on it, and even
put code that uses it into production. This then raises the barrier to
change the syntax of the Preview Feature. This seems to me to be more
pronounced in the case of core language features, as opposed to
incubator libraries and modules (e.g. the HTTP client).

I also want to ask about the balance we're striking between existing
Java developers and newcomers. I would argue that existing Java
developers really don't care whether we have a new keyword ('match')
for an expression form providing equivalent functionality to a switch
statement - existing code needs to be refactored in either case to
take advantage of the new capability.

On the other hand, I've also been spending a lot of time teaching
newcomers to the Java lanaguage over the last year. Given how my
students have struggled with default fallthrough, I can't see how
introducing a syntax form where the fallthrough behaviour depends upon
a minor syntax variation of the switch label is anything other than a
compounding of the original sin of importing default fallthrough from
C/C++. A separate keyword makes it entirely clear that this is a
different language feature and so different behaviour is to be
expected.

I like this version of the proposal better than previous discussions,
but I'm sure in my own mind that as proposed it will be a source of
new bugs that could be avoided. By introducing a new keyword we can
maintain a clean split between the legacy switch and a syntax that can
be cleanly upgraded to full match expressions over subsequent Java
versions (which is actually a much better fit, to my mind, to the
Preview Feature mechanism).

Thanks,

Ben
On Thu, 23 Aug 2018 at 08:18, Stephen Colebourne <scolebourne at joda.org> wrote:
>
> As I have expressed previously on amber-dev, I find the design adopted
> here to be overly complex and with a poor syntax choice. There are
> aspects to like, and the proposal is better than earlier versions. I
> also acknowledge that this is a proposal to create a preview language
> feature, however in all likelihood, there will be no further changes
> to the syntax/design once integrated as a preview.
>
> Some specific objections:
>
> (FYI, "classic" uses colon and has fallthrough-by-default, "enhanced"
> uses arrow -> and does not allow fall-through).
>
> - the original goal was to add one new language feature (expression
> switch), yet the solution adds three (classic expression switch,
> enhanced statement switch & enhanced expression switch). This
> unnecessarily complicates Java for little gain IMO.
>
> - all the evidence I've seen suggests that fallback in switch is very
> rare, and I'd strongly argue that most developers consider
> fallthrough-by-default to be a poor feature of Java. Adding a new
> language feature (classic expression switch) which adds more
> fallthrough-by-default is a mistake IMO. It is likely that style
> guides and best practice will recommend against using classic
> expression switch. I do not accept that consistency ("it makes a 2x2
> grid") is a sufficient reason to add a new language feature that will
> effectively be dead on arrival.
>
> - the use of -> as the separator in enhanced switch is undesirable
> IMO. Objections to the syntax have also been made by some expert group
> members. The particular problem is that it means something different
> to the use of -> in lambda expressions. This can result in confusing
> semi-puzzlers, such as "case RED -> YELLOW -> 6";.
>
> - the vast majority of existing switch statements would be more safely
> expressed using the enhanced form. As such, the enhanced form
> effectively replaces the classic form in Java - its that much better
> that classic statement switch will just be abandoned by developers
> IMO. Given this and the pointlessness of adding classic expression
> switch, the argument for using the keyword `switch` for the enhanced
> forms is weak IMO. At my presentations on the topic, the vast majority
> of developers have come to the conclusion that the enhanced form
> should use a different keyword, for example `match`, not `switch`.
> Doing so allows the colon to be used instead of the arrow:
>
> var name = match (trafficLight) {
>   case RED: "Red";
>   case YELLOW: "Yellow";
>   case GREEN: "Green";
> };
>
> - Adding a new keyword/symbol for the expression form (and potentially
> declaring it an ExpressionStatement instead of having a separate
> statement form) would IMO be a simpler approach to the problem space,
> potentially adding just 1 new form rather than 3 as proposed by the
> JEP.
>
>
> In summary, the world will not end if the JEP 325 progresses, and this
> proposal is much better than previous ones. However I believe that by
> taking the route of expanding `switch` rather than using a different
> keyword/symbol the result has been greater complexity than is
> necessary to meet the actual goal of "expression switch".
>
> thanks
> Stephen
>
>
> On 17 August 2018 at 18:44,  <mark.reinhold at oracle.com> wrote:
> > The following JEP is proposed to target JDK 12:
> >
> >   325: Switch Expressions (Preview)
> >        http://openjdk.java.net/jeps/325
> >
> > Feedback on this proposal is more than welcome, as are reasoned
> > objections.  If no such objections are raised by 23:00 UTC on Friday,
> > 24 August, or if they’re raised and then satisfactorily answered, then
> > per the JEP 2.0 process proposal [1] I’ll target this JEP to JDK 12.
> >
> > - Mark
> >
> >
> > [1] http://cr.openjdk.java.net/~mr/jep/jep-2.0-02.html