Generic selection
Provides a way to choose one of several expressions at compile time, based on a type of a controlling expression
Contents |
[edit] Syntax
_Generic ( controlling-expression , association-list )
|
(since C11) | ||||||||
where association-list is a comma-separated list of associations, each of which has the syntax
type-name : expression
|
|||||||||
default : expression
|
|||||||||
where
| type-name | - | any complete object type that isn't variably-modified (that is, not VLA or pointer to VLA). |
| controlling-expression | - | any expression (except for the comma operator) whose type must be compatible with one of the type-names if the default association is not used
|
| expression | - | any expression (except for the comma operator) of any type and value category |
No two type-names in the association-list may specify compatible types. There may be only one association that uses the keyword default. If default is not used and none of the type-names are compatible with the type of the controlling expression, the program will not compile.
[edit] Explanation
The type of controlling-expression (after applying lvalue conversions if applicable and if supported -- see notes below), is compared with type-names from the list of associations.
If the type is compatible with the type-name of one of the associations, then the type, value, and value category of the generic selection are the type, value, and value category of the expression that appears after the colon for that type-name.
If none of the type-names are compatible with the type of the controlling-expression, and the default association is provided, then the type, value, and value category of the generic selection are the type, value, and value category of the expression after the default : label.
[edit] Notes
The controlling-expression and the expressions of the selections that are not chosen are are never evaluated.
The lvalue conversions of the controlling expression are still an open issue (DR 431/n1930): clang does not perform them and takes the types of lvalues as-is: "abc" matches char[4] and not char*, (int const){0} matches const int and not int). Meanwhile, gcc performs the lvalue conversion of the controlling expression (in type domain only: it discards the top-level cvr-qualifiers and atomicity or applies array to pointer/function-to-pointer transformations to the type of the controlling expression, without initiating any side-effects or calculating any values). For gcc, "abc" matches char* and not char[4] and (int const){0} matches int, and not const int.
All value categories, including function designators and void expressions, are allowed as expressions in a generic selection, and if selected, the generic selection itself has the same value category.
The type-generic math macros from <tgmath.h>, introduced in C99, were implemented in compiler-specific manner. Generic selections, introduced in C11, gave the programmers the ability to write similar type-dependent code.
Generic selection is similar to overloading in C++ (where one of several functions is chosen at compile time based on the types of the arguments), except that it makes the selection between arbitrary expressions, which, unlike function overloads, may have different return types and even different value categories.
[edit] Keywords
[edit] Example
#include <stdio.h> #include <math.h> // Possible implementation of the tgmath.h macro cbrt #define cbrt(X) _Generic((X), \ long double: cbrtl, \ default: cbrt, \ /*for clang*/ const float: cbrtf, \ float: cbrtf \ )(X) int main(void) { double x = 8.0; const float y = 3.375; printf("cbrt(8.0) = %f\n", cbrt(x)); // selects the default cbrt printf("cbrtf(3.375) = %f\n", cbrt(y)); // gcc: converts const float to float, // then selects cbrtf // clang: selects cbrtf for const float }
Output:
cbrt(8.0) = 2.000000 cbrtf(3.375) = 1.500000
[edit] References
- C11 standard (ISO/IEC 9899:2011):
- 6.5.1.1 Generic selection (p: 78-79)
