www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - final switch traps and improvements

reply "bearophile" <bearophileHUGS lycos.com> writes:
Recently Ada 2012 was accepted as ISO standard, and on the good 
Lambda the Ultimate blog there is a small thread about it, with 
an interesting post:

http://lambda-the-ultimate.org/node/4661#comment-73731

The post makes two simple valid points, worth considering.

Currently in D this code compiles with no errors:


void main() {
     bool b;
     final switch (b) {
         case true:
             break;
     }
}


And gives at run-time:
core.exception.SwitchError test(3): No appropriate switch clause 
found

It's one of my top fifteen bug reports (it was mislabelled as 
enhancement request):
http://d.puremagic.com/issues/show_bug.cgi?id=5713

In my opinion It's a topic worth discussing and fixing.

The simplest solution is to statically disallow "final switch" to 
operate on anything bug enums.

A better solution is to make it work correctly where it can, and 
statically disallow the other situations. This means a final 
switch on bool must require both true and false cases. If the 
final switching is on a "uint % 3" expression then the compiler 
must statically require all the 0,1,2 cases and nothing else.


If part of all of this enhancement request gets accepted, then 
other useful final switch situations are worth supporting:
http://d.puremagic.com/issues/show_bug.cgi?id=596

One common use case is with Nullable, but syntactically speaking 
this is not good enough:

import std.typecons: Nullable;
void main() {
     auto n = Nullable!int(5);
     final switch (n) {
         case Nullable(false, _): ...
         case Nullable(true, _): ...
     }
}

Bye,
bearophile
Dec 23 2012
parent reply Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:
On 12/24/12 2:11 AM, bearophile wrote:

 Recently Ada 2012 was accepted as ISO standard, and on the good Lambda
 the Ultimate blog there is a small thread about it, with an interesting
 post:

 http://lambda-the-ultimate.org/node/4661#comment-73731

 The post makes two simple valid points, worth considering.

 Currently in D this code compiles with no errors:


 void main() {
 bool b;
 final switch (b) {
 case true:
 break;
 }
 }


 And gives at run-time:
 core.exception.SwitchError test(3): No appropriate switch clause found

 It's one of my top fifteen bug reports (it was mislabelled as
 enhancement request):
 http://d.puremagic.com/issues/show_bug.cgi?id=5713

 In my opinion It's a topic worth discussing and fixing.


Yes, this is a bug in the language design (as much as the "method is 
hidden" exception that I recall we got rid of).


 The simplest solution is to statically disallow "final switch" to
 operate on anything bug enums.


I'd say we eradicate the exception altogether. Again, it reflects 
incompleteness in language's approach. Requiring people to add a

default: {}

whenever they don't care about certain cases is a low price to pay for 
ensuring everybody stays in sync.


Andrei
Dec 24 2012
parent reply "Adam D. Ruppe" <destructionator gmail.com> writes:
On Monday, 24 December 2012 at 16:01:31 UTC, Andrei Alexandrescu 
wrote:

 I'd say we eradicate the exception altogether. Again, it 
 reflects incompleteness in language's approach. Requiring 
 people to add a


Note btw:

void main() {
         int a = 10;
         switch(a) {
                 case 10:
         }
}

test7.d(3): Error: non-final switch statement without a default 
is deprecated


final switch lets you get away without a default because it is 
supposed to ensure you have a case in there for everything.

That final switch(b) {case false: } compiles at all is the bug.
Dec 24 2012
parent Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:
On 12/24/12 11:10 AM, Adam D. Ruppe wrote:

 On Monday, 24 December 2012 at 16:01:31 UTC, Andrei Alexandrescu wrote:

 I'd say we eradicate the exception altogether. Again, it reflects
 incompleteness in language's approach. Requiring people to add a


 Note btw:

 void main() {
 int a = 10;
 switch(a) {
 case 10:
 }
 }

 test7.d(3): Error: non-final switch statement without a default is
 deprecated


Oh, that's great!


 final switch lets you get away without a default because it is supposed
 to ensure you have a case in there for everything.

 That final switch(b) {case false: } compiles at all is the bug.


Yah, sorry for getting out of touch with the state of the art.


Andrei
Dec 24 2012