5

The various is... functions (e.g. isalpha, isdigit) in ctype.h aren't entirely predictable. They take int arguments but expect character values in the unsigned char range, so on a platform where char is signed, passing a char value directly could lead to undesirable sign extension. I believe that the typical approach to handling this is to explicitly cast to an unsigned char first.

Okay, but what is the proper, portable way to deal with the various isw... functions in wctype.h? wchar_t, like char, also may be signed or unsigned, but because wchar_t is itself a typedef, a typename of unsigned wchar_t is illegal.

4

2 に答える 2

2

Isn't that what wint_t is for? The iswXxxxx() functions take a wint_t type:

ISO 9899:1999 covers this in various sections, working backwards:

§7.25 Wide character classification and mapping utilities <wctype.h>

§7.25.2.1.1 The iswalnum function

Synopsis

#include <wctype.h>
int iswalnum(wint_t wc);

Description

The iswalnum function tests for any wide character for which iswalpha or iswdigit is true.

§7.24 Extended multibyte and wide character utilities <wchar.h>

§7.24.1 Introduction:

wint_t

which is an integer type unchanged by default argument promotions that can hold any value corresponding to members of the extended character set, as well as at least one value that does not correspond to any member of the extended character set (see WEOF below);269)

269) wchar_t and wint_t can be the same integer type.

The 'unchanged by default argument promotions' should mean that it has to be as big as an int, though it could be a short or unsigned short if sizeof(short) == sizeof(int) (which is seldom the case these days, though it was true for some 16-bit systems).

§7.17 Common definitions <stddef.h>

wchar_t

which is an integer type whose range of values can represent distinct codes for all members of the largest extended character set specified among the supported locales; the null character shall have the code value zero and each member of the basic character set shall have a code value equal to its value when used as the lone character in an integer character constant.

As long as the value passed to iswalnum() or its kin is a valid wchar_t or WEOF, the function will work correctly. If you manufactured the value out of thin air and manage to get the value wrong, you get undefined behaviour.

于 2012-05-06T05:34:46.207 に答える
2

Upon re-reading the ISO C99 specification regarding wctype.h, it states:

For all functions described in this subclause that accept an argument of type wint_t, the value shall be representable as a wchar_t or shall equal the value of the macro WEOF. If this argument has any other value, the behavior is undefined. (§7.25.1/5)

Contrast this with the corresponding note for ctype.h:

In all cases the argument is an int, the value of which shall be representable as an unsigned char or shall equal the value of the macro EOF. If the argument has any other value, the behavior is undefined. (§7.4/1)

(emphasis mine)

I think that it's also worth understanding the motivation for why the ctype.h functions require unsigned char representations. The standard requires that EOF be a negative int (§7.19.1/3), so the ctype.h functions use unsigned char representations to (try to) avoid potential ambiguity.

In contrast, that motivation doesn't exist for wctype.h functions. The standard makes no such requirement of WEOF, elaborated by footnote 270:

The value of the macro WEOF may differ from that of EOF and need not be negative.

because WEOF is already guaranteed to not conflict with any character represented by wchar_t (§7.24.1/3).

Therefore the wctype.h functions don't have or need any of the unsigned nonsense, and wchar_t values can be passed to them directly.

于 2012-05-06T06:03:35.853 に答える