8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering#include <errno.h>

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering#include <stdlib.h>

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering#include <inttypes.h>

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering#include <string.h>

329c542585cd92cb905990e3bf59eda16fd88cfbLennart Poettering#include <stdbool.h>

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering#include "utf8.h"

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering#include "util.h"

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poetteringstatic inline bool is_unicode_valid(uint32_t ch) {

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering if (ch >= 0x110000) /* End of unicode space */

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return false;

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering if ((ch & 0xFFFFF800) == 0xD800) /* Reserved area for UTF-16 */

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return false;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((ch >= 0xFDD0) && (ch <= 0xFDEF)) /* Reserved */

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return false;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((ch & 0xFFFE) == 0xFFFE) /* BOM (Byte Order Mark) */

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek return false;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return true;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering}

ecabcf8b6edcc856ec2fd5bd43fc675a8fe04731Lennart Poettering

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poetteringstatic bool is_unicode_control(uint32_t ch) {

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek /*

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering return (ch < ' ' && ch != '\t' && ch != '\n') ||

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering (0x7F <= ch && ch <= 0x9F);

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering}

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

4aa4d2ae9717d0f8656528a3197bbc0c256380b1Zbigniew Jędrzejewski-Szmekstatic int utf8_encoded_expected_len(const char *str) {

4aa4d2ae9717d0f8656528a3197bbc0c256380b1Zbigniew Jędrzejewski-Szmek unsigned char c = (unsigned char)str[0];

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering if (c < 0x80)

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering return 1;

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poettering if ((c & 0xe0) == 0xc0)

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return 2;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((c & 0xf0) == 0xe0)

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return 3;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((c & 0xf8) == 0xf0)

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return 4;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((c & 0xfc) == 0xf8)

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek return 5;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering if ((c & 0xfe) == 0xfc)

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return 6;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering return 0;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering}

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering

755bde375f4db393ad06e73340bfcf4d0cf91bb2Lennart Poetteringstatic int utf8_encoded_to_unichar(const char *str) {

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering int unichar;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering int len;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering int i;

cc56fafeebf814ef035e549115cf1850e6473fa5WaLyong Cho

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek len = utf8_encoded_expected_len(str);

da927ba997d68401563b927f92e6e40e021a8e5cMichal Schmidt switch (len) {

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek case 1:

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek return (int)str[0];

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering case 2:

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek unichar = str[0] & 0x1f;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering break;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering case 3:

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek unichar = (int)str[0] & 0x0f;

7dbb1d08f66cd44b1296be3ee8e3629b989e19a8Zbigniew Jędrzejewski-Szmek break;

8ea48dfcd33e8db0c01bf8c57c3bbcfdc3c86d4bLennart Poettering case 4:

unichar = (int)str[0] & 0x07;

break;

case 5:

unichar = (int)str[0] & 0x03;

break;

case 6:

unichar = (int)str[0] & 0x01;

break;

default:

return -1;

}

for (i = 1; i < len; i++) {

if (((int)str[i] & 0xc0) != 0x80)

return -1;

unichar <<= 6;

unichar |= (int)str[i] & 0x3f;

}

return unichar;

}

bool utf8_is_printable(const char* str, size_t length) {

const uint8_t *p;

assert(str);

for (p = (const uint8_t*) str; length; p++) {

int encoded_len = utf8_encoded_valid_unichar((const char *)p);

int32_t val = utf8_encoded_to_unichar((const char*)p);

if (encoded_len < 0 || val < 0 || is_unicode_control(val))

return false;

length -= encoded_len;

}

return true;

}

const char *utf8_is_valid(const char *str) {

const uint8_t *p;

assert(str);

for (p = (const uint8_t*) str; *p; ) {

int len = utf8_encoded_valid_unichar((const char *)p);

if (len < 0)

return NULL;

p += len;

}

return str;

}

char *ascii_is_valid(const char *str) {

const char *p;

assert(str);

for (p = str; *p; p++)

if ((unsigned char) *p >= 128)

return NULL;

return (char*) str;

}

char *ascii_filter(const char *str) {

const char *s;

char *r, *d;

size_t l;

assert(str);

l = strlen(str);

r = malloc(l + 1);

if (!r)

return NULL;

for (s = str, d = r; *s; s++)

if ((unsigned char) *s < 128)

*(d++) = *s;

*d = 0;

return r;

}

char *utf16_to_utf8(const void *s, size_t length) {

char *r;

const uint8_t *f;

uint8_t *t;

r = new(char, (length*3+1)/2 + 1);

if (!r)

return NULL;

t = (uint8_t*) r;

for (f = s; f < (const uint8_t*) s + length; f += 2) {

uint16_t c;

c = (f[1] << 8) | f[0];

if (c == 0) {

*t = 0;

return r;

} else if (c < 0x80) {

*(t++) = (uint8_t) c;

} else if (c < 0x800) {

*(t++) = (uint8_t) (0xc0 | (c >> 6));

*(t++) = (uint8_t) (0x80 | (c & 0x3f));

} else {

*(t++) = (uint8_t) (0xe0 | (c >> 12));

*(t++) = (uint8_t) (0x80 | ((c >> 6) & 0x3f));

*(t++) = (uint8_t) (0x80 | (c & 0x3f));

}

}

*t = 0;

return r;

}

static int utf8_unichar_to_encoded_len(int unichar) {

if (unichar < 0x80)

return 1;

if (unichar < 0x800)

return 2;

if (unichar < 0x10000)

return 3;

if (unichar < 0x200000)

return 4;

if (unichar < 0x4000000)

return 5;

return 6;

}

int utf8_encoded_valid_unichar(const char *str) {

int len;

int unichar;

int i;

len = utf8_encoded_expected_len(str);

if (len == 0)

return -1;

/* ascii is valid */

if (len == 1)

return 1;

/* check if expected encoded chars are available */

for (i = 0; i < len; i++)

if ((str[i] & 0x80) != 0x80)

return -1;

unichar = utf8_encoded_to_unichar(str);

/* check if encoded length matches encoded value */

if (utf8_unichar_to_encoded_len(unichar) != len)

return -1;

/* check if value has valid range */

if (!is_unicode_valid(unichar))

return -1;

return len;

}