Help Settle a Debate - C++

cmfarley19 · 12-15-2011, 08:34 AM

Is this legal:

Code:

switch ( variable )
{
case 0:
    ...
    break;
case 1:
    ...
   break;

if ( boolean )
{
   case 2:
      ...
      break;
   case 3:
      ...
      break;
}

case 4:
   ...
   break;
}

Note cases 2 & 3 are encapsulated within an if.

Thoughts or opinions?

We can not find documentation or references to support either side so if you know of any please let me know.

Cheers,

Chris

millgates · 12-15-2011, 08:45 AM

It compiles and runs, so I think it's legal.

Doc CPU · 12-15-2011, 08:50 AM

Hi there,

Quote:

Originally Posted by cmfarley19

Is this legal:

Code:

switch ( variable )
{
case 0:
    ...
    break;
case 1:
    ...
   break;

if ( boolean )
{
   case 2:
      ...
      break;
   case 3:
      ...
      break;
}

case 4:
   ...
   break;
}

no, it isn't - though I can't point you to a specification.

Quote:

Originally Posted by cmfarley19

Note cases 2 & 3 are encapsulated within an if.

That's the point. While any code is allowed (though useless) between a break and the next case, the case keyword is only allowed in a switch scope. In your sample, however, there are two cases as children of an if statement.

[X] Doc CPU

Aquarius_Girl · 12-15-2011, 08:50 AM

AFAIK, "variable" can't be a string or a floating point.

Doc CPU · 12-15-2011, 08:55 AM

Hi there,

Quote:

Originally Posted by Anisha Kaul

AFAIK, "variable" can't be a string or a floating point.

true, but this has nothing to do with the issue of nested scopes or cross-scope control structures.

[X] Doc CPU

Aquarius_Girl · 12-15-2011, 08:58 AM

Quote:

Originally Posted by Doc CPU

true, but this has nothing to do with the issue of nested scopes or cross-scope control structures.

Yeah, I had read the question incompletely. ;-)

SigTerm · 12-15-2011, 09:13 AM

2003 standard does not say it is invalid.

According to grammar (Annex A.5, ISO/IEC 14882:2003):

Code:

statement:
        labeled-statement
        expression-statement
        compound-statement
        selection-statement
        iteration-statement
        jump-statement
        declaration-statement
        try-block

labeled-statement:
        identifier : statement
        case constant-expression : statement
        default : statement

expression-statement:
        expression(opt);

compound-statement:
        { statement-seq(opt)}

statement-seq:
        statement
        statement-seq statement

selection-statement:
        if ( condition ) statement
        if ( condition ) statement else statement
        switch ( condition ) statement

condition:
        expression
        type-specifier-seq declarator = assignment-expression

it should be a valid expression.

switch is within "section-statement". Which can contain "labeled statement" and "section statement". "case" is "labeled statement". and "if" is "section statement".

Also, 6.4.2 (stmt.switch) states that:

Quote:

case and default labels in themselves do not alter the flow of control, which continues unimpeded
across such labels. To exit from a switch, see break, 6.6.1. [Note: usually, the substatement that is the
subject of a switch is compound and case and default labels appear on the top-level statements con-
tained within the (compound) substatement, but this is not required. Declarations can appear in the sub-
statement of a switch-statement. ]

Plus this compiles on g++/cl.exe:

Code:

int main(int argc, char** argv){
	switch(argc){
		case(0):
			break;
		if (true){
			case(1):
				break;
			case(2):
				break;
		}
	}

	return 0;
}

So it should be a perfectly valid expression.

Get yourself a copy of C++ standard, and read it next time instead of asking.

dcf · 12-15-2011, 09:35 AM

Yes. The code compiles under gcc as well as Visual C++. That does not mean it does what is implied. The 'if' statement is completely ignored by both compilers. Lint flags the 'if' statement as 'unreachable code.'

SigTerm: Rather than post comments about your interpretation of the standard, why don't you test the results?

cmfarley19 · 12-15-2011, 09:39 AM

So trying this code out in various compilers produces the same results.

Code:

int main(int argc, char* argv[])
{
 printf("Start\r\n");
 int idx;
 int full_switch=0;

 for (idx = 0; idx < 3; idx++)
 {
  switch (idx)
  {
  case 0:
   printf("case 0\r\n");
   break;
  if ( full_switch )
  {
  case 1:
   printf("case 1\r\n");
   break;
  }
  case 2:
   printf("case 2\r\n");
   break;
  }
 }

 full_switch = 1;

 for (idx = 0; idx < 3; idx++)
 {
  switch (idx)
  {
  case 0:
   printf("case 0\r\n");
   break;
  if ( full_switch )
  {
  case 1:
   printf("case 1\r\n");
   break;
  }
  case 2:
   printf("case 2\r\n");
   break;
  }
 }

 printf("End\r\n");
 return 0;
}

No compiler errors or warnings and produces the following results:

Code:

cmfarley@wombat2:~$ ./switch 
Start
case 0
case 1
case 2
case 0
case 1
case 2
End

dcf - I had not gone down the Lint route, but my testing showed the same. The if statement is ignored. Compiler optimization apparently.

SigTerm - No need to be a douche.

SigTerm · 12-15-2011, 09:56 AM

Quote:

Originally Posted by dcf

SigTerm: Rather than post comments about your interpretation of the standard, why don't you test the results?

Perhaps you missed "it compiles" part.

If block shouldn't be ignored, if there is no "break" before if. break terminates switch case.

Quote:

Originally Posted by cmfarley19

SigTerm - No need to be a douche.

I explained to YOU WHY this works, pointed you to relevant sections of standard, and this what I get?
Solve your problems yourself next time. There's plenty of documentation for everything - all at your fingertips, quickly available via any search engine.

cmfarley19 · 12-15-2011, 10:01 AM

Quote:

I explained to YOU WHY this works, pointed you to relevant sections of standard, and this what I get?

Perhaps you did. Had you left out the last line I would have considered it a fine response.

dcf · 12-15-2011, 10:04 AM

"Yes. The code compiles under gcc as well as Visual C++." Hmm. Nope. I don't think I missed that part. Compilers typically implement switch statement using a jump table. Think of it as a list of goto labels. Since the 'if' is smack in the middle of the list of labels it is unreacheable. Given that Visual C++ and gcc both compile the code and produce results that illustrate that the 'if' statement has no influence on the enclosed cases, indicates that Microsoft and GNU have a different interpretation of the standard than you do, SigTerm.

SigTerm · 12-15-2011, 10:06 AM

Quote:

Originally Posted by cmfarley19

Perhaps you did. Had you left out the last line I would have considered it a fine response.

Since you're registered since 2002 I suppose that you should be aware that "RTFM" is actually a useful advice and is not necessarily hostile.
Checking documentation takes minutes. Getting a meaningful reply from a thread - hours or days. So, yes, it makes sense to actually get a copy of C++ standard and read it when you're in trouble - if you value your time, that is.

Well, whatever. I'm outta here.

jthill · 12-15-2011, 10:19 AM

Hey, if you think your if-in-a-switch is dubious, get a load of Duff's device. The note's worth reading, including "but it compiles and runs just fine". You're in good company.

dcf · 12-15-2011, 10:31 AM

Wow! That Duff code is scary stuff. I can see how it works. But it is the kind of thing that is likely to trip up the unwary pretty easily. Or even the wary for that matter. I think the reason it works is that there are no breaks following the cases. The labels are still reachable destinations for the switch jumps, but there is no coding hanging out in space following a break that is inaccessible.