NAME

mbrtoc16, mbrtoc32, mbrtowc, mbrtowc_l — convert characters to wide characters

SYNOPSIS

#include <wchar.h>

size_t
mbrtowc(wchar_t *restrict pwc, const char *restrict str, size_t len, mstate_t *restrict ps);

#include <wchar.h>
#include <xlocale.h>

size_t
mbrtowc_l(wchar_t *restrict pwc, const char *restrict str, size_t len, mstate_t *restrict ps, locale_t loc);

#include <uchar.h>

size_t
mbrtoc16(char16_t *restrict p16c, const char *restrict str, size_t len, mbstate_t *restrict ps);

size_t
mbrtoc32(char32_t *restrict p32c, const char *restrict str, size_t len, mbstate_t *restrict ps);

DESCRIPTION

The mbrtoc16(), mbrtoc32(), mbrtowc(), and mbrtowc_l() functions convert character sequences, which may contain multi-byte characters, into different character formats. The functions work in the following formats:

mbrtoc16(): A UTF-16 code sequence, where every code point is represented by one or two char16_t. The UTF-16 encoding will encode certain Unicode code points as a pair of two 16-bit code sequences, commonly referred to as a surrogate pair.
mbrtoc32(): A UTF-32 code sequence, where every code point is represented by a single char32_t.
mbrtowc(), mbrtowc_l(): Wide characters, being a 32-bit value where every code point is represented by a single wchar_t. While the wchar_t and char32_t are different types, in this implementation, they are similar encodings.

The functions consume up to len characters from the string str and accumulate them in ps until a valid character is found, which is influenced by the LC_CTYPE category of the current locale. For example, in the C locale, only ASCII characters are recognized, while in a UTF-8 based locale like en_US.UTF-8, UTF-8 multi-byte character sequences that represent Unicode code points are recognized. The mbrtowc_l() function uses the locale passed in loc rather than the locale of the current thread.

When a valid character sequence has been found, it is converted to either a 16-bit character sequence for mbrtoc16() or a 32-bit character sequence for mbrtoc32() and will be stored in p16c and p32c respectively.

The ps argument represents a multi-byte conversion state which can be used across multiple calls to a given function (but not mixed between functions). These allow for characters to be consumed from subsequent buffers, e.g. different values of str. The functions may be called from multiple threads as long as they use unique values for ps. If ps is NULL, then a function-specific buffer will be used for the conversion state; however, this is stored between all threads and its use is not recommended.

When using these functions, more than one character may be output for a given set of consumed input characters. An example of this is when a given code point is represented as a set of surrogate pairs in UTF-16, which require two 16-bit characters to represent a code point. When this occurs, the functions return the special return value -3.

The functions all have a special behavior when NULL is passed for str. They instead will treat it as though pwc, p16c, or p32c were NULL, str had been passed as the empty string, "" and the length, len, would appear as the value 1. In other words, the functions would be called as:

mbrtowc(NULL, "", 1, ps)
mbrtowc_l(NULL, "", 1, ps)
mbrtoc16(NULL, "", 1, ps)
mbrtoc32(NULL, "", 1, ps)

Locale Details

Not all locales in the system are Unicode based locales. For example, ISO 8859 family locales have code points with values that do not match their counterparts in Unicode. When using these functions with non-Unicode based locales, the code points returned will be those determined by the locale. They will not be converted to the corresponding Unicode code point. For example, if using the Euro sign in ISO 8859-15, these functions might return the code point 0xa4 and not the Unicode value 0x20ac.

Regardless of the locale, the characters returned will be encoded as though the code point were the corresponding value in Unicode. This means that if a locale returns a value that would be a surrogate pair in the UTF-16 encoding, it will still be encoded as a UTF-16 character.

This behavior of the mbrtoc16() and mbrtoc32() functions should not be relied upon, is not portable, and subject to change for non-Unicode locales.

RETURN VALUES

The mbrtoc16(), mbrtoc32(), mbrtowc(), and mbrtowc_l() functions return the following values:

0: len or fewer bytes of str were consumed and the null wide character was written into the wide character buffer (pwc, p16c, p32c).
between 1 and len: The specified number of bytes were consumed and a single character was written into the wide character buffer (pwc, p16c, p32c).
(size_t)-1: An encoding error has occurred. The next len bytes of str do not contribute to a valid character. errno has been set to EILSEQ. No data was written into the wide character buffer (pwc, p16c, p32c).
(size_t)-2: len bytes of str were consumed, but a complete multi-byte character sequence has not been found and no data was written into the wide character buffer (pwc, p16c, p32c).
(size_t)-3: A character has been written into the wide character buffer (pwc, p16c, p32c). This character was from a previous call (such as another part of a UTF-16 surrogate pair) and no input was consumed. This is limited to the mbrtoc16() and mbrtoc32() functions.

EXAMPLES

Example 1 Using the mbrtoc32() function to convert a multibyte string.

#include <locale.h>
#include <stdlib.h>
#include <string.h>
#include <err.h>
#include <stdio.h>
#include <uchar.h>

int
main(void)
{
	mbstate_t mbs;
	char32_t out;
	size_t ret;
	const char *uchar_str = "\xe5\x85\x89";

	(void) memset(&mbs, 0, sizeof (mbs));
	(void) setlocale(LC_CTYPE, "en_US.UTF-8");
	ret = mbrtoc32(&out, uchar_str, strlen(uchar_str), &mbs);
	if (ret != strlen(uchar_str)) {
		errx(EXIT_FAILURE, "failed to convert string, got %zd",
		    ret);
	}

	(void) printf("Converted %zu bytes into UTF-32 character "
	    "0x%x0, ret, out);
	return (0);
}

When compiled and run, this produces:

$ ./a.out
Converted 3 bytes into UTF-32 character 0x5149

Example 2 Handling surrogate pairs from the mbrtoc16() function.

#include <locale.h>
#include <stdlib.h>
#include <string.h>
#include <err.h>
#include <stdio.h>
#include <uchar.h>

int
main(void)
{
        mbstate_t mbs;
        char16_t first, second;
        size_t ret;
        const char *uchar_str = "\xf0\x9f\x92\xa9";

        (void) memset(&mbs, 0, sizeof (mbs));
        (void) setlocale(LC_CTYPE, "en_US.UTF-8");
        ret = mbrtoc16(&first, uchar_str, strlen(uchar_str), &mbs);
        if (ret != strlen(uchar_str)) {
                errx(EXIT_FAILURE, "failed to convert string, got %zd",
                    ret);
        }

        ret = mbrtoc16(&second, "", 0, &mbs);
        if (ret != (size_t)-3) {
                errx(EXIT_FAILURE, "didn't get second surrogate pair, "
                    "got %zd", ret);
        }

        (void) printf("UTF-16 surrogates: 0x%x 0x%x0, first, second);
        return (0);
}

When compiled and run, this produces:

$ ./a.out
UTF-16 surrogates: 0xd83d 0xdca9

ERRORS

The mbrtoc16(), mbrtoc32(), mbrtowc(), and mbrtowc_l() functions will fail if:

EINVAL: The conversion state in ps is invalid.
EILSEQ: An invalid character sequence has been detected.

MT-LEVEL

The mbrtoc16(), mbrtoc32(), mbrtowc(), and mbrtowc_l() functions are MT-Safe as long as different mbstate_t structures are passed in ps. If ps is NULL or different threads use the same value for ps, then the functions are Unsafe.

INTERFACE STABILITY

Committed