| /* |
| * string.c : an XML string utilities module |
| * |
| * This module provides various utility functions for manipulating |
| * the xmlChar* type. All functions named xmlStr* have been moved here |
| * from the parser.c file (their original home). |
| * |
| * See Copyright for the status of this software. |
| * |
| * UTF8 string routines from: |
| * William Brack <wbrack@mmm.com.hk> |
| * |
| * daniel@veillard.com |
| */ |
| |
| #define IN_LIBXML |
| #include "libxml.h" |
| |
| #include <stdlib.h> |
| #include <string.h> |
| #include <libxml/xmlmemory.h> |
| #include <libxml/parserInternals.h> |
| #include <libxml/xmlstring.h> |
| |
| /************************************************************************ |
| * * |
| * Commodity functions to handle xmlChars * |
| * * |
| ************************************************************************/ |
| |
| /** |
| * xmlStrndup: |
| * @cur: the input xmlChar * |
| * @len: the len of @cur |
| * |
| * a strndup for array of xmlChar's |
| * |
| * Returns a new xmlChar * or NULL |
| */ |
| xmlChar * |
| xmlStrndup(const xmlChar *cur, int len) { |
| xmlChar *ret; |
| |
| if ((cur == NULL) || (len < 0)) return(NULL); |
| ret = (xmlChar *) xmlMallocAtomic((len + 1) * sizeof(xmlChar)); |
| if (ret == NULL) { |
| xmlErrMemory(NULL, NULL); |
| return(NULL); |
| } |
| memcpy(ret, cur, len * sizeof(xmlChar)); |
| ret[len] = 0; |
| return(ret); |
| } |
| |
| /** |
| * xmlStrdup: |
| * @cur: the input xmlChar * |
| * |
| * a strdup for array of xmlChar's. Since they are supposed to be |
| * encoded in UTF-8 or an encoding with 8bit based chars, we assume |
| * a termination mark of '0'. |
| * |
| * Returns a new xmlChar * or NULL |
| */ |
| xmlChar * |
| xmlStrdup(const xmlChar *cur) { |
| const xmlChar *p = cur; |
| |
| if (cur == NULL) return(NULL); |
| while (*p != 0) p++; /* non input consuming */ |
| return(xmlStrndup(cur, p - cur)); |
| } |
| |
| /** |
| * xmlCharStrndup: |
| * @cur: the input char * |
| * @len: the len of @cur |
| * |
| * a strndup for char's to xmlChar's |
| * |
| * Returns a new xmlChar * or NULL |
| */ |
| |
| xmlChar * |
| xmlCharStrndup(const char *cur, int len) { |
| int i; |
| xmlChar *ret; |
| |
| if ((cur == NULL) || (len < 0)) return(NULL); |
| ret = (xmlChar *) xmlMallocAtomic((len + 1) * sizeof(xmlChar)); |
| if (ret == NULL) { |
| xmlErrMemory(NULL, NULL); |
| return(NULL); |
| } |
| for (i = 0;i < len;i++) |
| ret[i] = (xmlChar) cur[i]; |
| ret[len] = 0; |
| return(ret); |
| } |
| |
| /** |
| * xmlCharStrdup: |
| * @cur: the input char * |
| * |
| * a strdup for char's to xmlChar's |
| * |
| * Returns a new xmlChar * or NULL |
| */ |
| |
| xmlChar * |
| xmlCharStrdup(const char *cur) { |
| const char *p = cur; |
| |
| if (cur == NULL) return(NULL); |
| while (*p != '\0') p++; /* non input consuming */ |
| return(xmlCharStrndup(cur, p - cur)); |
| } |
| |
| /** |
| * xmlStrcmp: |
| * @str1: the first xmlChar * |
| * @str2: the second xmlChar * |
| * |
| * a strcmp for xmlChar's |
| * |
| * Returns the integer result of the comparison |
| */ |
| |
| int |
| xmlStrcmp(const xmlChar *str1, const xmlChar *str2) { |
| register int tmp; |
| |
| if (str1 == str2) return(0); |
| if (str1 == NULL) return(-1); |
| if (str2 == NULL) return(1); |
| do { |
| tmp = *str1++ - *str2; |
| if (tmp != 0) return(tmp); |
| } while (*str2++ != 0); |
| return 0; |
| } |
| |
| /** |
| * xmlStrEqual: |
| * @str1: the first xmlChar * |
| * @str2: the second xmlChar * |
| * |
| * Check if both string are equal of have same content |
| * Should be a bit more readable and faster than xmlStrEqual() |
| * |
| * Returns 1 if they are equal, 0 if they are different |
| */ |
| |
| int |
| xmlStrEqual(const xmlChar *str1, const xmlChar *str2) { |
| if (str1 == str2) return(1); |
| if (str1 == NULL) return(0); |
| if (str2 == NULL) return(0); |
| do { |
| if (*str1++ != *str2) return(0); |
| } while (*str2++); |
| return(1); |
| } |
| |
| /** |
| * xmlStrQEqual: |
| * @pref: the prefix of the QName |
| * @name: the localname of the QName |
| * @str: the second xmlChar * |
| * |
| * Check if a QName is Equal to a given string |
| * |
| * Returns 1 if they are equal, 0 if they are different |
| */ |
| |
| int |
| xmlStrQEqual(const xmlChar *pref, const xmlChar *name, const xmlChar *str) { |
| if (pref == NULL) return(xmlStrEqual(name, str)); |
| if (name == NULL) return(0); |
| if (str == NULL) return(0); |
| |
| do { |
| if (*pref++ != *str) return(0); |
| } while ((*str++) && (*pref)); |
| if (*str++ != ':') return(0); |
| do { |
| if (*name++ != *str) return(0); |
| } while (*str++); |
| return(1); |
| } |
| |
| /** |
| * xmlStrncmp: |
| * @str1: the first xmlChar * |
| * @str2: the second xmlChar * |
| * @len: the max comparison length |
| * |
| * a strncmp for xmlChar's |
| * |
| * Returns the integer result of the comparison |
| */ |
| |
| int |
| xmlStrncmp(const xmlChar *str1, const xmlChar *str2, int len) { |
| register int tmp; |
| |
| if (len <= 0) return(0); |
| if (str1 == str2) return(0); |
| if (str1 == NULL) return(-1); |
| if (str2 == NULL) return(1); |
| #ifdef __GNUC__ |
| tmp = strncmp(str1, str2, len); |
| return tmp; |
| #else |
| do { |
| tmp = *str1++ - *str2; |
| if (tmp != 0 || --len == 0) return(tmp); |
| } while (*str2++ != 0); |
| return 0; |
| #endif |
| } |
| |
| static const xmlChar casemap[256] = { |
| 0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07, |
| 0x08,0x09,0x0A,0x0B,0x0C,0x0D,0x0E,0x0F, |
| 0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17, |
| 0x18,0x19,0x1A,0x1B,0x1C,0x1D,0x1E,0x1F, |
| 0x20,0x21,0x22,0x23,0x24,0x25,0x26,0x27, |
| 0x28,0x29,0x2A,0x2B,0x2C,0x2D,0x2E,0x2F, |
| 0x30,0x31,0x32,0x33,0x34,0x35,0x36,0x37, |
| 0x38,0x39,0x3A,0x3B,0x3C,0x3D,0x3E,0x3F, |
| 0x40,0x61,0x62,0x63,0x64,0x65,0x66,0x67, |
| 0x68,0x69,0x6A,0x6B,0x6C,0x6D,0x6E,0x6F, |
| 0x70,0x71,0x72,0x73,0x74,0x75,0x76,0x77, |
| 0x78,0x79,0x7A,0x7B,0x5C,0x5D,0x5E,0x5F, |
| 0x60,0x61,0x62,0x63,0x64,0x65,0x66,0x67, |
| 0x68,0x69,0x6A,0x6B,0x6C,0x6D,0x6E,0x6F, |
| 0x70,0x71,0x72,0x73,0x74,0x75,0x76,0x77, |
| 0x78,0x79,0x7A,0x7B,0x7C,0x7D,0x7E,0x7F, |
| 0x80,0x81,0x82,0x83,0x84,0x85,0x86,0x87, |
| 0x88,0x89,0x8A,0x8B,0x8C,0x8D,0x8E,0x8F, |
| 0x90,0x91,0x92,0x93,0x94,0x95,0x96,0x97, |
| 0x98,0x99,0x9A,0x9B,0x9C,0x9D,0x9E,0x9F, |
| 0xA0,0xA1,0xA2,0xA3,0xA4,0xA5,0xA6,0xA7, |
| 0xA8,0xA9,0xAA,0xAB,0xAC,0xAD,0xAE,0xAF, |
| 0xB0,0xB1,0xB2,0xB3,0xB4,0xB5,0xB6,0xB7, |
| 0xB8,0xB9,0xBA,0xBB,0xBC,0xBD,0xBE,0xBF, |
| 0xC0,0xC1,0xC2,0xC3,0xC4,0xC5,0xC6,0xC7, |
| 0xC8,0xC9,0xCA,0xCB,0xCC,0xCD,0xCE,0xCF, |
| 0xD0,0xD1,0xD2,0xD3,0xD4,0xD5,0xD6,0xD7, |
| 0xD8,0xD9,0xDA,0xDB,0xDC,0xDD,0xDE,0xDF, |
| 0xE0,0xE1,0xE2,0xE3,0xE4,0xE5,0xE6,0xE7, |
| 0xE8,0xE9,0xEA,0xEB,0xEC,0xED,0xEE,0xEF, |
| 0xF0,0xF1,0xF2,0xF3,0xF4,0xF5,0xF6,0xF7, |
| 0xF8,0xF9,0xFA,0xFB,0xFC,0xFD,0xFE,0xFF |
| }; |
| |
| /** |
| * xmlStrcasecmp: |
| * @str1: the first xmlChar * |
| * @str2: the second xmlChar * |
| * |
| * a strcasecmp for xmlChar's |
| * |
| * Returns the integer result of the comparison |
| */ |
| |
| int |
| xmlStrcasecmp(const xmlChar *str1, const xmlChar *str2) { |
| register int tmp; |
| |
| if (str1 == str2) return(0); |
| if (str1 == NULL) return(-1); |
| if (str2 == NULL) return(1); |
| do { |
| tmp = casemap[*str1++] - casemap[*str2]; |
| if (tmp != 0) return(tmp); |
| } while (*str2++ != 0); |
| return 0; |
| } |
| |
| /** |
| * xmlStrncasecmp: |
| * @str1: the first xmlChar * |
| * @str2: the second xmlChar * |
| * @len: the max comparison length |
| * |
| * a strncasecmp for xmlChar's |
| * |
| * Returns the integer result of the comparison |
| */ |
| |
| int |
| xmlStrncasecmp(const xmlChar *str1, const xmlChar *str2, int len) { |
| register int tmp; |
| |
| if (len <= 0) return(0); |
| if (str1 == str2) return(0); |
| if (str1 == NULL) return(-1); |
| if (str2 == NULL) return(1); |
| do { |
| tmp = casemap[*str1++] - casemap[*str2]; |
| if (tmp != 0 || --len == 0) return(tmp); |
| } while (*str2++ != 0); |
| return 0; |
| } |
| |
| /** |
| * xmlStrchr: |
| * @str: the xmlChar * array |
| * @val: the xmlChar to search |
| * |
| * a strchr for xmlChar's |
| * |
| * Returns the xmlChar * for the first occurrence or NULL. |
| */ |
| |
| const xmlChar * |
| xmlStrchr(const xmlChar *str, xmlChar val) { |
| if (str == NULL) return(NULL); |
| while (*str != 0) { /* non input consuming */ |
| if (*str == val) return((xmlChar *) str); |
| str++; |
| } |
| return(NULL); |
| } |
| |
| /** |
| * xmlStrstr: |
| * @str: the xmlChar * array (haystack) |
| * @val: the xmlChar to search (needle) |
| * |
| * a strstr for xmlChar's |
| * |
| * Returns the xmlChar * for the first occurrence or NULL. |
| */ |
| |
| const xmlChar * |
| xmlStrstr(const xmlChar *str, const xmlChar *val) { |
| int n; |
| |
| if (str == NULL) return(NULL); |
| if (val == NULL) return(NULL); |
| n = xmlStrlen(val); |
| |
| if (n == 0) return(str); |
| while (*str != 0) { /* non input consuming */ |
| if (*str == *val) { |
| if (!xmlStrncmp(str, val, n)) return((const xmlChar *) str); |
| } |
| str++; |
| } |
| return(NULL); |
| } |
| |
| /** |
| * xmlStrcasestr: |
| * @str: the xmlChar * array (haystack) |
| * @val: the xmlChar to search (needle) |
| * |
| * a case-ignoring strstr for xmlChar's |
| * |
| * Returns the xmlChar * for the first occurrence or NULL. |
| */ |
| |
| const xmlChar * |
| xmlStrcasestr(const xmlChar *str, xmlChar *val) { |
| int n; |
| |
| if (str == NULL) return(NULL); |
| if (val == NULL) return(NULL); |
| n = xmlStrlen(val); |
| |
| if (n == 0) return(str); |
| while (*str != 0) { /* non input consuming */ |
| if (casemap[*str] == casemap[*val]) |
| if (!xmlStrncasecmp(str, val, n)) return(str); |
| str++; |
| } |
| return(NULL); |
| } |
| |
| /** |
| * xmlStrsub: |
| * @str: the xmlChar * array (haystack) |
| * @start: the index of the first char (zero based) |
| * @len: the length of the substring |
| * |
| * Extract a substring of a given string |
| * |
| * Returns the xmlChar * for the first occurrence or NULL. |
| */ |
| |
| xmlChar * |
| xmlStrsub(const xmlChar *str, int start, int len) { |
| int i; |
| |
| if (str == NULL) return(NULL); |
| if (start < 0) return(NULL); |
| if (len < 0) return(NULL); |
| |
| for (i = 0;i < start;i++) { |
| if (*str == 0) return(NULL); |
| str++; |
| } |
| if (*str == 0) return(NULL); |
| return(xmlStrndup(str, len)); |
| } |
| |
| /** |
| * xmlStrlen: |
| * @str: the xmlChar * array |
| * |
| * length of a xmlChar's string |
| * |
| * Returns the number of xmlChar contained in the ARRAY. |
| */ |
| |
| int |
| xmlStrlen(const xmlChar *str) { |
| int len = 0; |
| |
| if (str == NULL) return(0); |
| while (*str != 0) { /* non input consuming */ |
| str++; |
| len++; |
| } |
| return(len); |
| } |
| |
| /** |
| * xmlStrncat: |
| * @cur: the original xmlChar * array |
| * @add: the xmlChar * array added |
| * @len: the length of @add |
| * |
| * a strncat for array of xmlChar's, it will extend @cur with the len |
| * first bytes of @add. |
| * |
| * Returns a new xmlChar *, the original @cur is reallocated if needed |
| * and should not be freed |
| */ |
| |
| xmlChar * |
| xmlStrncat(xmlChar *cur, const xmlChar *add, int len) { |
| int size; |
| xmlChar *ret; |
| |
| if ((add == NULL) || (len == 0)) |
| return(cur); |
| if (cur == NULL) |
| return(xmlStrndup(add, len)); |
| |
| size = xmlStrlen(cur); |
| ret = (xmlChar *) xmlRealloc(cur, (size + len + 1) * sizeof(xmlChar)); |
| if (ret == NULL) { |
| xmlErrMemory(NULL, NULL); |
| return(cur); |
| } |
| memcpy(&ret[size], add, len * sizeof(xmlChar)); |
| ret[size + len] = 0; |
| return(ret); |
| } |
| |
| /** |
| * xmlStrncatNew: |
| * @str1: first xmlChar string |
| * @str2: second xmlChar string |
| * @len: the len of @str2 |
| * |
| * same as xmlStrncat, but creates a new string. The original |
| * two strings are not freed. |
| * |
| * Returns a new xmlChar * or NULL |
| */ |
| xmlChar * |
| xmlStrncatNew(const xmlChar *str1, const xmlChar *str2, int len) { |
| int size; |
| xmlChar *ret; |
| |
| if ((str2 == NULL) || (len == 0)) |
| return(xmlStrdup(str1)); |
| if (str1 == NULL) |
| return(xmlStrndup(str2, len)); |
| |
| size = xmlStrlen(str1); |
| ret = (xmlChar *) xmlMalloc((size + len + 1) * sizeof(xmlChar)); |
| if (ret == NULL) { |
| xmlErrMemory(NULL, NULL); |
| return(xmlStrndup(str1, size)); |
| } |
| memcpy(ret, str1, size * sizeof(xmlChar)); |
| memcpy(&ret[size], str2, len * sizeof(xmlChar)); |
| ret[size + len] = 0; |
| return(ret); |
| } |
| |
| /** |
| * xmlStrcat: |
| * @cur: the original xmlChar * array |
| * @add: the xmlChar * array added |
| * |
| * a strcat for array of xmlChar's. Since they are supposed to be |
| * encoded in UTF-8 or an encoding with 8bit based chars, we assume |
| * a termination mark of '0'. |
| * |
| * Returns a new xmlChar * containing the concatenated string. |
| */ |
| xmlChar * |
| xmlStrcat(xmlChar *cur, const xmlChar *add) { |
| const xmlChar *p = add; |
| |
| if (add == NULL) return(cur); |
| if (cur == NULL) |
| return(xmlStrdup(add)); |
| |
| while (*p != 0) p++; /* non input consuming */ |
| return(xmlStrncat(cur, add, p - add)); |
| } |
| |
| /** |
| * xmlStrPrintf: |
| * @buf: the result buffer. |
| * @len: the result buffer length. |
| * @msg: the message with printf formatting. |
| * @...: extra parameters for the message. |
| * |
| * Formats @msg and places result into @buf. |
| * |
| * Returns the number of characters written to @buf or -1 if an error occurs. |
| */ |
| int |
| xmlStrPrintf(xmlChar *buf, int len, const xmlChar *msg, ...) { |
| va_list args; |
| int ret; |
| |
| if((buf == NULL) || (msg == NULL)) { |
| return(-1); |
| } |
| |
| va_start(args, msg); |
| ret = vsnprintf((char *) buf, len, (const char *) msg, args); |
| va_end(args); |
| buf[len - 1] = 0; /* be safe ! */ |
| |
| return(ret); |
| } |
| |
| /** |
| * xmlStrVPrintf: |
| * @buf: the result buffer. |
| * @len: the result buffer length. |
| * @msg: the message with printf formatting. |
| * @ap: extra parameters for the message. |
| * |
| * Formats @msg and places result into @buf. |
| * |
| * Returns the number of characters written to @buf or -1 if an error occurs. |
| */ |
| int |
| xmlStrVPrintf(xmlChar *buf, int len, const xmlChar *msg, va_list ap) { |
| int ret; |
| |
| if((buf == NULL) || (msg == NULL)) { |
| return(-1); |
| } |
| |
| ret = vsnprintf((char *) buf, len, (const char *) msg, ap); |
| buf[len - 1] = 0; /* be safe ! */ |
| |
| return(ret); |
| } |
| |
| /************************************************************************ |
| * * |
| * Generic UTF8 handling routines * |
| * * |
| * From rfc2044: encoding of the Unicode values on UTF-8: * |
| * * |
| * UCS-4 range (hex.) UTF-8 octet sequence (binary) * |
| * 0000 0000-0000 007F 0xxxxxxx * |
| * 0000 0080-0000 07FF 110xxxxx 10xxxxxx * |
| * 0000 0800-0000 FFFF 1110xxxx 10xxxxxx 10xxxxxx * |
| * * |
| * I hope we won't use values > 0xFFFF anytime soon ! * |
| * * |
| ************************************************************************/ |
| |
| |
| /** |
| * xmlUTF8Size: |
| * @utf: pointer to the UTF8 character |
| * |
| * calculates the internal size of a UTF8 character |
| * |
| * returns the numbers of bytes in the character, -1 on format error |
| */ |
| int |
| xmlUTF8Size(const xmlChar *utf) { |
| xmlChar mask; |
| int len; |
| |
| if (utf == NULL) |
| return -1; |
| if (*utf < 0x80) |
| return 1; |
| /* check valid UTF8 character */ |
| if (!(*utf & 0x40)) |
| return -1; |
| /* determine number of bytes in char */ |
| len = 2; |
| for (mask=0x20; mask != 0; mask>>=1) { |
| if (!(*utf & mask)) |
| return len; |
| len++; |
| } |
| return -1; |
| } |
| |
| /** |
| * xmlUTF8Charcmp: |
| * @utf1: pointer to first UTF8 char |
| * @utf2: pointer to second UTF8 char |
| * |
| * compares the two UCS4 values |
| * |
| * returns result of the compare as with xmlStrncmp |
| */ |
| int |
| xmlUTF8Charcmp(const xmlChar *utf1, const xmlChar *utf2) { |
| |
| if (utf1 == NULL ) { |
| if (utf2 == NULL) |
| return 0; |
| return -1; |
| } |
| return xmlStrncmp(utf1, utf2, xmlUTF8Size(utf1)); |
| } |
| |
| /** |
| * xmlUTF8Strlen: |
| * @utf: a sequence of UTF-8 encoded bytes |
| * |
| * compute the length of an UTF8 string, it doesn't do a full UTF8 |
| * checking of the content of the string. |
| * |
| * Returns the number of characters in the string or -1 in case of error |
| */ |
| int |
| xmlUTF8Strlen(const xmlChar *utf) { |
| int ret = 0; |
| |
| if (utf == NULL) |
| return(-1); |
| |
| while (*utf != 0) { |
| if (utf[0] & 0x80) { |
| if ((utf[1] & 0xc0) != 0x80) |
| return(-1); |
| if ((utf[0] & 0xe0) == 0xe0) { |
| if ((utf[2] & 0xc0) != 0x80) |
| return(-1); |
| if ((utf[0] & 0xf0) == 0xf0) { |
| if ((utf[0] & 0xf8) != 0xf0 || (utf[3] & 0xc0) != 0x80) |
| return(-1); |
| utf += 4; |
| } else { |
| utf += 3; |
| } |
| } else { |
| utf += 2; |
| } |
| } else { |
| utf++; |
| } |
| ret++; |
| } |
| return(ret); |
| } |
| |
| /** |
| * xmlGetUTF8Char: |
| * @utf: a sequence of UTF-8 encoded bytes |
| * @len: a pointer to @bytes len |
| * |
| * Read one UTF8 Char from @utf |
| * |
| * Returns the char value or -1 in case of error, and updates *len with the |
| * number of bytes consumed |
| */ |
| int |
| xmlGetUTF8Char(const unsigned char *utf, int *len) { |
| unsigned int c; |
| |
| if (utf == NULL) |
| goto error; |
| if (len == NULL) |
| goto error; |
| if (*len < 1) |
| goto error; |
| |
| c = utf[0]; |
| if (c & 0x80) { |
| if (*len < 2) |
| goto error; |
| if ((utf[1] & 0xc0) != 0x80) |
| goto error; |
| if ((c & 0xe0) == 0xe0) { |
| if (*len < 3) |
| goto error; |
| if ((utf[2] & 0xc0) != 0x80) |
| goto error; |
| if ((c & 0xf0) == 0xf0) { |
| if (*len < 4) |
| goto error; |
| if ((c & 0xf8) != 0xf0 || (utf[3] & 0xc0) != 0x80) |
| goto error; |
| *len = 4; |
| /* 4-byte code */ |
| c = (utf[0] & 0x7) << 18; |
| c |= (utf[1] & 0x3f) << 12; |
| c |= (utf[2] & 0x3f) << 6; |
| c |= utf[3] & 0x3f; |
| } else { |
| /* 3-byte code */ |
| *len = 3; |
| c = (utf[0] & 0xf) << 12; |
| c |= (utf[1] & 0x3f) << 6; |
| c |= utf[2] & 0x3f; |
| } |
| } else { |
| /* 2-byte code */ |
| *len = 2; |
| c = (utf[0] & 0x1f) << 6; |
| c |= utf[1] & 0x3f; |
| } |
| } else { |
| /* 1-byte code */ |
| *len = 1; |
| } |
| return(c); |
| |
| error: |
| *len = 0; |
| return(-1); |
| } |
| |
| /** |
| * xmlCheckUTF8: |
| * @utf: Pointer to putative UTF-8 encoded string. |
| * |
| * Checks @utf for being valid UTF-8. @utf is assumed to be |
| * null-terminated. This function is not super-strict, as it will |
| * allow longer UTF-8 sequences than necessary. Note that Java is |
| * capable of producing these sequences if provoked. Also note, this |
| * routine checks for the 4-byte maximum size, but does not check for |
| * 0x10ffff maximum value. |
| * |
| * Return value: true if @utf is valid. |
| **/ |
| int |
| xmlCheckUTF8(const unsigned char *utf) |
| { |
| int ix; |
| unsigned char c; |
| |
| for (ix = 0; (c = utf[ix]);) { |
| if (c & 0x80) { |
| if ((utf[ix + 1] & 0xc0) != 0x80) |
| return(0); |
| if ((c & 0xe0) == 0xe0) { |
| if ((utf[ix + 2] & 0xc0) != 0x80) |
| return(0); |
| if ((c & 0xf0) == 0xf0) { |
| if ((c & 0xf8) != 0xf0 || (utf[ix + 3] & 0xc0) != 0x80) |
| return(0); |
| ix += 4; |
| /* 4-byte code */ |
| } else |
| /* 3-byte code */ |
| ix += 3; |
| } else |
| /* 2-byte code */ |
| ix += 2; |
| } else |
| /* 1-byte code */ |
| ix++; |
| } |
| return(1); |
| } |
| |
| /** |
| * xmlUTF8Strsize: |
| * @utf: a sequence of UTF-8 encoded bytes |
| * @len: the number of characters in the array |
| * |
| * storage size of an UTF8 string |
| * |
| * Returns the storage size of |
| * the first 'len' characters of ARRAY |
| * |
| */ |
| |
| int |
| xmlUTF8Strsize(const xmlChar *utf, int len) { |
| const xmlChar *ptr=utf; |
| xmlChar ch; |
| |
| if (len <= 0) |
| return(0); |
| |
| while ( len-- > 0) { |
| if ( !*ptr ) |
| break; |
| if ( (ch = *ptr++) & 0x80) |
| while ( (ch<<=1) & 0x80 ) |
| ptr++; |
| } |
| return (ptr - utf); |
| } |
| |
| |
| /** |
| * xmlUTF8Strndup: |
| * @utf: the input UTF8 * |
| * @len: the len of @utf (in chars) |
| * |
| * a strndup for array of UTF8's |
| * |
| * Returns a new UTF8 * or NULL |
| */ |
| xmlChar * |
| xmlUTF8Strndup(const xmlChar *utf, int len) { |
| xmlChar *ret; |
| int i; |
| |
| if ((utf == NULL) || (len < 0)) return(NULL); |
| i = xmlUTF8Strsize(utf, len); |
| ret = (xmlChar *) xmlMallocAtomic((i + 1) * sizeof(xmlChar)); |
| if (ret == NULL) { |
| xmlGenericError(xmlGenericErrorContext, |
| "malloc of %ld byte failed\n", |
| (len + 1) * (long)sizeof(xmlChar)); |
| return(NULL); |
| } |
| memcpy(ret, utf, i * sizeof(xmlChar)); |
| ret[i] = 0; |
| return(ret); |
| } |
| |
| /** |
| * xmlUTF8Strpos: |
| * @utf: the input UTF8 * |
| * @pos: the position of the desired UTF8 char (in chars) |
| * |
| * a function to provide the equivalent of fetching a |
| * character from a string array |
| * |
| * Returns a pointer to the UTF8 character or NULL |
| */ |
| xmlChar * |
| xmlUTF8Strpos(const xmlChar *utf, int pos) { |
| xmlChar ch; |
| |
| if (utf == NULL) return(NULL); |
| if ( (pos < 0) || (pos >= xmlUTF8Strlen(utf)) ) |
| return(NULL); |
| while (pos--) { |
| if ((ch=*utf++) == 0) return(NULL); |
| if ( ch & 0x80 ) { |
| /* if not simple ascii, verify proper format */ |
| if ( (ch & 0xc0) != 0xc0 ) |
| return(NULL); |
| /* then skip over remaining bytes for this char */ |
| while ( (ch <<= 1) & 0x80 ) |
| if ( (*utf++ & 0xc0) != 0x80 ) |
| return(NULL); |
| } |
| } |
| return((xmlChar *)utf); |
| } |
| |
| /** |
| * xmlUTF8Strloc: |
| * @utf: the input UTF8 * |
| * @utfchar: the UTF8 character to be found |
| * |
| * a function to provide the relative location of a UTF8 char |
| * |
| * Returns the relative character position of the desired char |
| * or -1 if not found |
| */ |
| int |
| xmlUTF8Strloc(const xmlChar *utf, const xmlChar *utfchar) { |
| int i, size; |
| xmlChar ch; |
| |
| if (utf==NULL || utfchar==NULL) return -1; |
| size = xmlUTF8Strsize(utfchar, 1); |
| for(i=0; (ch=*utf) != 0; i++) { |
| if (xmlStrncmp(utf, utfchar, size)==0) |
| return(i); |
| utf++; |
| if ( ch & 0x80 ) { |
| /* if not simple ascii, verify proper format */ |
| if ( (ch & 0xc0) != 0xc0 ) |
| return(-1); |
| /* then skip over remaining bytes for this char */ |
| while ( (ch <<= 1) & 0x80 ) |
| if ( (*utf++ & 0xc0) != 0x80 ) |
| return(-1); |
| } |
| } |
| |
| return(-1); |
| } |
| /** |
| * xmlUTF8Strsub: |
| * @utf: a sequence of UTF-8 encoded bytes |
| * @start: relative pos of first char |
| * @len: total number to copy |
| * |
| * Create a substring from a given UTF-8 string |
| * Note: positions are given in units of UTF-8 chars |
| * |
| * Returns a pointer to a newly created string |
| * or NULL if any problem |
| */ |
| |
| xmlChar * |
| xmlUTF8Strsub(const xmlChar *utf, int start, int len) { |
| int i; |
| xmlChar ch; |
| |
| if (utf == NULL) return(NULL); |
| if (start < 0) return(NULL); |
| if (len < 0) return(NULL); |
| |
| /* |
| * Skip over any leading chars |
| */ |
| for (i = 0;i < start;i++) { |
| if ((ch=*utf++) == 0) return(NULL); |
| if ( ch & 0x80 ) { |
| /* if not simple ascii, verify proper format */ |
| if ( (ch & 0xc0) != 0xc0 ) |
| return(NULL); |
| /* then skip over remaining bytes for this char */ |
| while ( (ch <<= 1) & 0x80 ) |
| if ( (*utf++ & 0xc0) != 0x80 ) |
| return(NULL); |
| } |
| } |
| |
| return(xmlUTF8Strndup(utf, len)); |
| } |