Subversion Repositories HelenOS

Rev

Rev 4175 | Rev 4196 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

/*
 * Copyright (c) 2001-2004 Jakub Jermar
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * - Redistributions of source code must retain the above copyright
 *   notice, this list of conditions and the following disclaimer.
 * - Redistributions in binary form must reproduce the above copyright
 *   notice, this list of conditions and the following disclaimer in the
 *   documentation and/or other materials provided with the distribution.
 * - The name of the author may not be used to endorse or promote products
 *   derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

/** @addtogroup generic
 * @{
 */

/**
 * @file
 * @brief Miscellaneous functions.
 */

#include <string.h>
#include <print.h>
#include <cpu.h>
#include <arch/asm.h>
#include <arch.h>
#include <console/kconsole.h>

char invalch = '?';

/** Decode a single UTF-8 character from a NULL-terminated string.
 *
 * Decode a single UTF-8 character from a plain char NULL-terminated
 * string. Decoding starts at @index and this index is incremented
 * if the current UTF-8 string is encoded in more than a single byte.
 *
 * @param str   Plain character NULL-terminated string.
 * @param index Index (counted in plain characters) where to start
 *              the decoding.
 * @param limit Maximal allowed value of index.
 *
 * @return Decoded character in UTF-32 or '?' if the encoding is wrong.
 *
 */
wchar_t utf8_decode(const char *str, index_t *index, index_t limit)
{
    uint8_t c1;           /* First plain character from str */
    uint8_t c2;           /* Second plain character from str */
    uint8_t c3;           /* Third plain character from str */
    uint8_t c4;           /* Fourth plain character from str */
    
    if (*index > limit)
        return invalch;
    
    c1 = (uint8_t) str[*index];
    
    if ((c1 & 0x80) == 0) {
        /* Plain ASCII (code points 0 .. 127) */
        return (wchar_t) c1;
    }
    
    if ((c1 & 0xe0) == 0xc0) {
        /* Code points 128 .. 2047 */
        if (*index + 1 > limit)
            return invalch;
        
        c2 = (uint8_t) str[*index + 1];
        if ((c2 & 0xc0) == 0x80) {
            (*index)++;
            return ((wchar_t) ((c1 & 0x1f) << 6) | (c2 & 0x3f));
        } else
            return invalch;
    }
    
    if ((c1 & 0xf0) == 0xe0) {
        /* Code points 2048 .. 65535 */
        if (*index + 2 > limit)
            return invalch;
        
        c2 = (uint8_t) str[*index + 1];
        if ((c2 & 0xc0) == 0x80) {
            (*index)++;
            c3 = (uint8_t) str[*index + 1];
            if ((c3 & 0xc0) == 0x80) {
                (*index)++;
                return ((wchar_t) ((c1 & 0x0f) << 12) | ((c2 & 0x3f) << 6) | (c3 & 0x3f));
            } else
                return invalch;
        } else
            return invalch;
    }
    
    if ((c1 & 0xf8) == 0xf0) {
        /* Code points 65536 .. 1114111 */
        if (*index + 3 > limit)
            return invalch;
        
        c2 = (uint8_t) str[*index + 1];
        if ((c2 & 0xc0) == 0x80) {
            (*index)++;
            c3 = (uint8_t) str[*index + 1];
            if ((c3 & 0xc0) == 0x80) {
                (*index)++;
                c4 = (uint8_t) str[*index + 1];
                if ((c4 & 0xc0) == 0x80) {
                    (*index)++;
                    return ((wchar_t) ((c1 & 0x07) << 18) | ((c2 & 0x3f) << 12) | ((c3 & 0x3f) << 6) | (c4 & 0x3f));
                } else
                    return invalch;
            } else
                return invalch;
        } else
            return invalch;
    }
    
    return invalch;
}

/** Encode a single UTF-32 character as UTF-8
 *
 * Encode a single UTF-32 character as UTF-8 and store it into
 * the given buffer at @index. Encoding starts at @index and
 * this index is incremented if the UTF-8 character takes
 * more than a single byte.
 *
 * @param ch    Input UTF-32 character.
 * @param str   Output buffer.
 * @param index Index (counted in plain characters) where to start
 *              the encoding
 * @param limit Maximal allowed value of index.
 *
 * @return True if the character was encoded or false if there is not
 *         enought space in the output buffer or the character is invalid
 *         Unicode code point.
 *
 */
bool utf8_encode(const wchar_t ch, char *str, index_t *index, index_t limit)
{
    if (*index > limit)
        return false;
    
    if ((ch >= 0) && (ch <= 127)) {
        /* Plain ASCII (code points 0 .. 127) */
        str[*index] = ch & 0x7f;
        return true;
    }
    
    if ((ch >= 128) && (ch <= 2047)) {
        /* Code points 128 .. 2047 */
        if (*index + 1 > limit)
            return false;
        
        str[*index] = 0xc0 | ((ch >> 6) & 0x1f);
        (*index)++;
        str[*index] = 0x80 | (ch & 0x3f);
        return true;
    }
    
    if ((ch >= 2048) && (ch <= 65535)) {
        /* Code points 2048 .. 65535 */
        if (*index + 2 > limit)
            return false;
        
        str[*index] = 0xe0 | ((ch >> 12) & 0x0f);
        (*index)++;
        str[*index] = 0x80 | ((ch >> 6) & 0x3f);
        (*index)++;
        str[*index] = 0x80 | (ch & 0x3f);
        return true;
    }
    
    if ((ch >= 65536) && (ch <= 1114111)) {
        /* Code points 65536 .. 1114111 */
        if (*index + 3 > limit)
            return false;
        
        str[*index] = 0xf0 | ((ch >> 18) & 0x07);
        (*index)++;
        str[*index] = 0x80 | ((ch >> 12) & 0x3f);
        (*index)++;
        str[*index] = 0x80 | ((ch >> 6) & 0x3f);
        (*index)++;
        str[*index] = 0x80 | (ch & 0x3f);
        return true;
    }
    
    return false;
}

/** Get bytes used by UTF-8 characters.
 *
 * Get the number of bytes (count of plain characters) which
 * are used by a given count of UTF-8 characters in a string.
 * As UTF-8 encoding is multibyte, there is no constant
 * correspondence between number of characters and used bytes.
 *
 * @param str   UTF-8 string to consider.
 * @param count Number of UTF-8 characters to count.
 *
 * @return Number of bytes used by the characters.
 *
 */
size_t utf8_count_bytes(const char *str, count_t count)
{
    size_t size = 0;
    index_t index = 0;
    
    while ((utf8_decode(str, &index, UTF8_NO_LIMIT) != 0) && (size < count)) {
        size++;
        index++;
    }
    
    return index;
}

/** Check whether character is plain ASCII.
 *
 * @return True if character is plain ASCII.
 *
 */
bool ascii_check(const wchar_t ch)
{
    if ((ch >= 0) && (ch <= 127))
        return true;
    
    return false;
}

/** Check whether character is Unicode.
 *
 * @return True if character is valid Unicode code point.
 *
 */
bool unicode_check(const wchar_t ch)
{
    if ((ch >= 0) && (ch <= 1114111))
        return true;
    
    return false;
}

/** Return number of plain characters in a string.
 *
 * @param str NULL-terminated string.
 *
 * @return Number of characters in str.
 *
 */
size_t strlen(const char *str)
{
    size_t size;
    for (size = 0; str[size]; size++);
    
    return size;
}

/** Return number of UTF-8 characters in a string.
 *
 * @param str NULL-terminated UTF-8 string.
 *
 * @return Number of UTF-8 characters in str.
 *
 */
size_t strlen_utf8(const char *str)
{
    size_t size = 0;
    index_t index = 0;
    
    while (utf8_decode(str, &index, UTF8_NO_LIMIT) != 0) {
        size++;
        index++;
    }
    
    return size;
}

/** Return number of UTF-32 characters in a string.
 *
 * @param str NULL-terminated UTF-32 string.
 *
 * @return Number of UTF-32 characters in str.
 *
 */
size_t strlen_utf32(const wchar_t *str)
{
    size_t size;
    for (size = 0; str[size]; size++);
    
    return size;
}

/** Compare two NULL terminated strings
 *
 * Do a char-by-char comparison of two NULL terminated strings.
 * The strings are considered equal iff they consist of the same
 * characters on the minimum of their lengths.
 *
 * @param src First string to compare.
 * @param dst Second string to compare.
 *
 * @return 0 if the strings are equal, -1 if first is smaller, 1 if second smaller.
 *
 */
int strcmp(const char *src, const char *dst)
{
    for (; *src && *dst; src++, dst++) {
        if (*src < *dst)
            return -1;
        if (*src > *dst)
            return 1;
    }
    if (*src == *dst)
        return 0;
    
    if (!*src)
        return -1;
    
    return 1;
}


/** Compare two NULL terminated strings
 *
 * Do a char-by-char comparison of two NULL terminated strings.
 * The strings are considered equal iff they consist of the same
 * characters on the minimum of their lengths and specified maximal
 * length.
 *
 * @param src First string to compare.
 * @param dst Second string to compare.
 * @param len Maximal length for comparison.
 *
 * @return 0 if the strings are equal, -1 if first is smaller, 1 if second smaller.
 *
 */
int strncmp(const char *src, const char *dst, size_t len)
{
    unsigned int i;
    
    for (i = 0; (*src) && (*dst) && (i < len); src++, dst++, i++) {
        if (*src < *dst)
            return -1;
        
        if (*src > *dst)
            return 1;
    }
    
    if (i == len || *src == *dst)
        return 0;
    
    if (!*src)
        return -1;
    
    return 1;
}



/** Copy NULL terminated string.
 *
 * Copy at most 'len' characters from string 'src' to 'dest'.
 * If 'src' is shorter than 'len', '\0' is inserted behind the
 * last copied character.
 *
 * @param src  Source string.
 * @param dest Destination buffer.
 * @param len  Size of destination buffer.
 *
 */
void strncpy(char *dest, const char *src, size_t len)
{
    unsigned int i;
    
    for (i = 0; i < len; i++) {
        if (!(dest[i] = src[i]))
            return;
    }
    
    dest[i - 1] = '\0';
}

/** Find first occurence of character in string.
 *
 * @param s String to search.
 * @param i Character to look for.
 *
 * @return Pointer to character in @a s or NULL if not found.
 */
extern char *strchr(const char *s, int i)
{
    while (*s != '\0') {
        if (*s == i)
            return (char *) s;
        ++s;
    }
    
    return NULL;
}

/** @}
 */