Thread: warning: initializer element is not computable at load time

With MSVC I get:

main.cpp(55) : warning C4204: nonstandard extension used : non-constant aggregate initializer
main.cpp(55) : warning C4204: nonstandard extension used : non-constant aggregate initializer
main.cpp(84) : warning C4715: 'precedence' : not all control paths return a value

The first two warnings indicate that the standard way would be:

Code:

char operator[ 2 ];
operator[0] = operator1;
operator[1] = operator2;

The last warning should be wrong (the compiler doesn't realize that you have all cases covered), but you might use if...else if...else instead (or perhaps just return opValue[0] - opValue[1];

Originally Posted by yougene

1. Which part of the statement is the "initializer element" ?

The parts in red:

Code:

char operator[ 2 ] = { operator1, operator2 };

Originally Posted by yougene

2. What does it mean that it's "non-computable at load-time" ?

The values of those elements are only known at run time. It is not actually true in this case, but it could be true if you call precedence() using values obtained from the user.

Incidentally, note that precedence() has a control path that does not return a value.

>> Incidentally, note that precedence() has a control path that does not return a value. Well, technically, yes; practically, no. [edit] what the heck is up with the boards formatting engine? [/code]

Originally Posted by Sebastiani

Incidentally, note that precedence() has a control path that does not return a value. Well, technically, yes; practically, no.

Yeah, now I see that anon has covered that. My eyes were fooled by a sequence of if statements with returns but no return at the end. Consdering that, I would say that the code is wrong in a sense: the last if should be an else, or removed entirely.

Originally Posted by yougene

1. Will this cause problems if I do pass user-input?

Since a compiler extension is in use, no. However it may cause problems if you intend to port the code.

>> Consdering that, I would say that the code is wrong in a sense: the last if should be an else, or removed entirely. I agree completely. Besides, a slight change to the code could lead to a nasty bug. If nothing else, it's bad form...

Originally Posted by yougene

The way I did it has the possibility of not returning anything
Using else has the possibility of not returning a correct value

The possibly of not returning a correct value always exists, though it can be mitigated with repeatable tests (and formal verification, if necessary).

Incidentally, if you follow the convention of merely comparing with 0, you could just write:

Code:

return opValue[ 0 ] - opValue[ 1 ];

EDIT:
Concerning your original question, I think that it may be better to do it this way:

Code:

/* Return precedence value of operator1 as a positive integer,
 * or 0 if operator1 is invalid.
 */
int precedence( char operator1 )
{
    switch( operator1 )
    {
    case '+':
    case '-':
        return 1;
    case '*':
    case '/':
    case '%':
        return 2;
    case '^':
        return 3;
    case '(':
    case ')':
        return 4;
    default:
        return 0;
    }
}

/* Return a positive integer, zero, or a negative integer if the precedence of
 * operator1 is greater than, equal to, or less than that of operator2 respectively.
 * The operators are assumed to be valid.
 */
int comparePrecedence( char operator1, char operator2 )
{
    return precedence( operator1 ) - precedence( operator2 );
}

Code:

int isOperator( char c )
{
	if ( c == '(' || c == ')'
       || c == '+' || c == '-'
       || c == '*' || c == '/'
	  || c == '^' || c == '%' )
		return 1;

	else
		return 0;


}

Might I suggest:

Code:

int isOperator( char c )
{
    return ( c == '(' || c == ')'
        || c == '+' || c == '-'
        || c == '*' || c == '/'
        || c == '^' || c == '%' );
}

(Note the indentation, yours isn't very good!)
cpwiki.sf.net/Indentation

Or if you want to be fancy, you could use something like

Code:

int isOperator( char c ) {
    return strchr("()+-*/^%", c) != NULL;
}

Originally Posted by yougene

What would be good form?

I was originally going to use if, else if, else. Both have the same short-coming in my mind although in different ways.

The way I did it has the possibility of not returning anything
Using else has the possibility of not returning a correct value

As the program stands neither concern is valid since a number is always positive, 0, or negative.

My value judgment was which is worse? "Having a function not return anything which is easily caught or having a function return the wrong value (else statement) which could have been more difficult." The function is pretty trivial and it probably doesn't require this hypothetical but there it is. Is my reasoning partial or flawed?

Consider:

Code:

if (x) return x;
if (y) return y;
if (z) return z;

...and...

Code:

if (x) return x;
else if (y) return y;
else return z;

The latter form will stop the compiler from thinking there might another case you have forgotten. But they are also identical.
There is no drawback the the latter one, which you should be well aware of.

Code:

if(X) return x;
if(Y) return y;
return z;

will also get rid of the warning if you are sure that X, Y and Z represent all the available choices