Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Are compound statements (blocks) surrounded by parens expressions in ANSI C?

Browsing the Linux kernel sources I found some piece of code where a block of statements surrounded by parenthesis is treated as a expression a la lisp (or ML), that is, an expression which value is the value of the last statement.

For example:

int a = ({
    int i;
    int t = 1;
    for (i = 2; i<5; i++) {
        t*=i;
    }
    t;
});

I've been looking at the ANSI C grammar trying to figure out how this piece of code would fit in the parse tree, but I haven't been successful.

So, does anybody know if this behaviour is mandated by the standard or is just a peculiarity of GCC?

Update: I've tried with the flag -pedantic and the compiler now gives me a warning:

warning: ISO C forbids braced-groups within expressions
like image 626
fortran Avatar asked Aug 06 '09 10:08

fortran


2 Answers

This is not standard C. It is a gcc extension called statement expressions. You can find the complete list of C extensions here. This is actually one of the many gcc extensions used in the Linux kernel and it seems like clang supports this too and although it is not explicitly named in the document.

As you observed the last expression serves as the value of the expression, the document says (emphasis mine):

The last thing in the compound statement should be an expression followed by a semicolon; the value of this subexpression serves as the value of the entire construct. (If you use some other kind of statement last within the braces, the construct has type void, and thus effectively no value.)

One of the main benefits would be to make safe macros that would avoid multiple evaluations of arguments with side effects. The example given uses this unsafe macro:

#define max(a,b) ((a) > (b) ? (a) : (b)) 

which evaluates either a or b twice and can be rewritten to eliminate this problem using statement expressions as follows:

#define maxint(a,b) \    ({int _a = (a), _b = (b); _a > _b ? _a : _b; })  

Note, the need to explicitly use int which can fixed using another gcc extension Typeof:

#define max(a,b) \    ({ typeof (a) _a = (a), _b = (b); _a > _b ? _a : _b; })  

Note that clang also supports typeof.

like image 104
Shafik Yaghmour Avatar answered Sep 28 '22 00:09

Shafik Yaghmour


It's called "braced-group within expression".

It's not allowed by ANSI/ISO C nor C++ but gcc supports it.

like image 21
laalto Avatar answered Sep 28 '22 01:09

laalto