Order of evaluation of comma operator [duplicate]

What does the , operator do in C?


The expression:

(expression1,  expression2)

First expression1 is evaluated, then expression2 is evaluated, and the value of expression2 is returned for the whole expression.


I've seen used most in while loops:

string s;
while(read_string(s), s.len() > 5)
{
   //do something
}

It will do the operation, then do a test based on a side-effect. The other way would be to do it like this:

string s;
read_string(s);
while(s.len() > 5)
{
   //do something
   read_string(s);
}

The comma operator will evaluate the left operand, discard the result and then evaluate the right operand and that will be the result. The idiomatic use as noted in the link is when initializing the variables used in a for loop, and it gives the following example:

void rev(char *s, size_t len)
{
  char *first;
  for ( first = s, s += len - 1; s >= first; --s)
      /*^^^^^^^^^^^^^^^^^^^^^^^*/ 
      putchar(*s);
}

Otherwise there are not many great uses of the comma operator, although it is easy to abuse to generate code that is hard to read and maintain.

From the draft C99 standard the grammar is as follows:

expression:
  assignment-expression
  expression , assignment-expression

and paragraph 2 says:

The left operand of a comma operator is evaluated as a void expression; there is a sequence point after its evaluation. Then the right operand is evaluated; the result has its type and value. 97) If an attempt is made to modify the result of a comma operator or to access it after the next sequence point, the behavior is undefined.

Footnote 97 says:

A comma operator does not yield an lvalue.

which means you can not assign to the result of the comma operator.

It is important to note that the comma operator has the lowest precedence and therefore there are cases where using () can make a big difference, for example:

#include <stdio.h>

int main()
{
    int x, y ;

    x = 1, 2 ;
    y = (3,4) ;

    printf( "%d %d\n", x, y ) ;
}

will have the following output:

1 4

The comma operator combines the two expressions either side of it into one, evaluating them both in left-to-right order. The value of the right-hand side is returned as the value of the whole expression. (expr1, expr2) is like { expr1; expr2; } but you can use the result of expr2 in a function call or assignment.

It is often seen in for loops to initialise or maintain multiple variables like this:

for (low = 0, high = MAXSIZE; low < high; low = newlow, high = newhigh)
{
    /* do something with low and high and put new values
       in newlow and newhigh */
}

Apart from this, I've only used it "in anger" in one other case, when wrapping up two operations that should always go together in a macro. We had code that copied various binary values into a byte buffer for sending on a network, and a pointer maintained where we had got up to:

unsigned char outbuff[BUFFSIZE];
unsigned char *ptr = outbuff;

*ptr++ = first_byte_value;
*ptr++ = second_byte_value;

send_buff(outbuff, (int)(ptr - outbuff));

Where the values were shorts or ints we did this:

*((short *)ptr)++ = short_value;
*((int *)ptr)++ = int_value;

Later we read that this was not really valid C, because (short *)ptr is no longer an l-value and can't be incremented, although our compiler at the time didn't mind. To fix this, we split the expression in two:

*(short *)ptr = short_value;
ptr += sizeof(short);

However, this approach relied on all developers remembering to put both statements in all the time. We wanted a function where you could pass in the output pointer, the value and and the value's type. This being C, not C++ with templates, we couldn't have a function take an arbitrary type, so we settled on a macro:

#define ASSIGN_INCR(p, val, type)  ((*((type) *)(p) = (val)), (p) += sizeof(type))

By using the comma operator we were able to use this in expressions or as statements as we wished:

if (need_to_output_short)
    ASSIGN_INCR(ptr, short_value, short);

latest_pos = ASSIGN_INCR(ptr, int_value, int);

send_buff(outbuff, (int)(ASSIGN_INCR(ptr, last_value, int) - outbuff));

I'm not suggesting any of these examples are good style! Indeed, I seem to remember Steve McConnell's Code Complete advising against even using comma operators in a for loop: for readability and maintainability, the loop should be controlled by only one variable, and the expressions in the for line itself should only contain loop-control code, not other extra bits of initialisation or loop maintenance.


It causes the evaluation of multiple statements, but uses only the last one as a resulting value (rvalue, I think).

So...

int f() { return 7; }
int g() { return 8; }

int x = (printf("assigning x"), f(), g() );

should result in x being set to 8.