C conditional operator ('?') with empty second parameter

Question

Typically the '?' operator is used in the following form:

A ? B : C

However in cases where B = A I have seen the following abbreviation

A ? : C

This surprisingly works. Is it better to leave the second parameter in (style wise), or is their a chance certain compilers won't be able to handle this?

Solution 1

It is not permitted by the language C (as far as I know), but compilers such as gcc have the shortcut a?:c as an extension. a?:c means the same as a?a:c.

Solution 2

Its a gcc's extension

Conditionals with Omitted Operands

x ? : y is equivalent to x ? x : y

Solution 3

Unless I'm badly mistake, you're using a compiler extension (at a guess, gcc). I'm pretty sure the standard does not allow you to omit the second operand to the ternary operator.

Solution 4

I fill in a bit.

The standard uses the term conditional operator.

Syntax
  conditional-expression:
          logical-OR-expression
          logical-OR-expression ? expression : conditional-expression

A conditional expression does not yield an lvalue. Also; Wikipedia; Conditional

_{Note: I.e.: C++ has:
        logical-OR-expression ? expression : assignment-expression}

Constraints:
* The first operand shall have scalar type[1].
* One of the following shall hold for the second and third operands:
   — both operands have arithmetic type[2];
   — both operands have the same structure[3] or union type[4];
   — both operands have void type[5];
   — both operands are pointers to qualified or unqualified[6] versions of compatible
     types[7];
   — one operand is a pointer and the other is a null pointer constant[8]; or
   — one operand is a pointer to an object or incomplete type[9] and the other 
     is a pointer to a qualified or unqualified version of void.

Foot food:

[1] Scalar type     : Arithmetic types and pointer types.
[2] Arithmetic type : Integer and floating types.
[3] Structure type  : A sequentially allocated nonempty set of member objects (and, in
                     certain circumstances, an incomplete array), each of which has an
                     optionally specified name and possibly distinct type.
[4] Union type      : An overlapping nonempty set of member objects, each of which has
                     an optionally specified name and possibly distinct type.
[5] Void type       : An empty set of values; it is an incomplete type that cannot be
                     completed.
[6] Qualified type  : 1998 (const and volatile), 1999 (restrict), respectively 
                     2011 (_Atomic). *
[7] Compatible type : Their types are the same.
[8] Null ptr. const.: NULL; implementation-defined null pointer constant.
[9] Incomplete type : Types that describe objects but lack information needed to determine 
                      their sizes.

^* Type qualifiers in C

So: Not wise to use.

Solution 5

i did a little research in the web, acording to wikipedia, this behavior is supported by a GNU extension of C. http://en.wikipedia.org/wiki/%3F:#C

So it is very probable that other compilers consider this illegal. By the way, this operator is called ternary conditional so you can browse about it.

EDIT:

I checked in gcc and apple llvm and it works fine.

View more solutions

Comments

[Lion]

Looking like Groovy-like syntax.

[Celada]

The page you linked to contradicts itself. On the one hand it says "This example is perfectly equivalent to x ? x : y", which means x is evaluated twice, but on the other hand the last paragraph states that x would be evaluated only once, which would make it perfectly equivalent to x || y, not x ? x : y

[Prasoon Saurav]

@Celada : I think it means to say that x ? : y is roughly equivalent to x ? x : y except x gets evaluated only once in the former case.

[Prasoon Saurav]

@Celada : x || y evaluates to 0 or 1 which is not the case with this operator.

[Celada]

I stand corrected. I thought || returned value value of its first argument if the first argument was something that is considered true (nonzero).

[Jack]

It's one of ISO C99 features. It works fine for me on GCC 4.4.1<TDM-2 mingw32> and TCC(that's an ANSI C compiler) version 0.9.25 on Windows plataform.

[mbauman]

It means the same... with the caveat that a does not contain side-effects. a?:c only executes a once, whereas a?a:c would execute the side-effects of a twice.

[4LegsDrivenCat]

It works for Clang too.

[4LegsDrivenCat]

It works for Clang too, not only GCC.