Merge mainline changes

[helenos.git] / uspace / lib / posix / stdlib / strtold.c

/*
 * Copyright (c) 2011 Jiri Zarevucky
 * 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 libposix
 * @{
 */
/** @file Backend for floating point conversions.
 */

#define LIBPOSIX_INTERNAL

/* Must be first. */
#include "../stdbool.h"

#include "../internal/common.h"
#include "../stdlib.h"

#include "../assert.h"
#include "../ctype.h"
#include "../stdint.h"
#include "../strings.h"
#include "../errno.h"
#include "../limits.h"

#include "../float.h"

#ifndef HUGE_VALL
        #define HUGE_VALL (+1.0l / +0.0l)
#endif

#ifndef abs
        #define abs(x) (((x) < 0) ? -(x) : (x))
#endif

/* If the constants are not defined, use double precision as default. */
#ifndef LDBL_MANT_DIG
        #define LDBL_MANT_DIG 53
#endif
#ifndef LDBL_MAX_EXP
        #define LDBL_MAX_EXP 1024
#endif
#ifndef LDBL_MIN_EXP
        #define LDBL_MIN_EXP (-1021)
#endif
#ifndef LDBL_DIG
        #define LDBL_DIG 15
#endif
#ifndef LDBL_MIN
        #define LDBL_MIN 2.2250738585072014E-308
#endif

/* power functions ************************************************************/

#if LDBL_MAX_EXP >= 16384
const int MAX_POW5 = 12;
#else
const int MAX_POW5 = 8;
#endif

/* The value at index i is approximately 5**(2**i). */
long double pow5[] = {
        0x5p0l,
        0x19p0l,
        0x271p0l,
        0x5F5E1p0l,
        0x2386F26FC1p0l,
        0x4EE2D6D415B85ACEF81p0l,
        0x184F03E93FF9F4DAA797ED6E38ED6p36l,
        0x127748F9301D319BF8CDE66D86D62p185l,
        0x154FDD7F73BF3BD1BBB77203731FDp482l,
#if LDBL_MAX_EXP >= 16384
        0x1C633415D4C1D238D98CAB8A978A0p1076l,
        0x192ECEB0D02EA182ECA1A7A51E316p2265l,
        0x13D1676BB8A7ABBC94E9A519C6535p4643l,
        0x188C0A40514412F3592982A7F0094p9398l,
#endif
};

#if LDBL_MAX_EXP >= 16384
const int MAX_POW2 = 15;
#else
const int MAX_POW2 = 9;
#endif

/* Powers of two. */
long double pow2[] = {
        0x1P1l,
        0x1P2l,
        0x1P4l,
        0x1P8l,
        0x1P16l,
        0x1P32l,
        0x1P64l,
        0x1P128l,
        0x1P256l,
        0x1P512l,
#if LDBL_MAX_EXP >= 16384
        0x1P1024l,
        0x1P2048l,
        0x1P4096l,
        0x1P8192l,
#endif
};

/**
 * Multiplies a number by a power of five.
 * The result may be inexact and may not be the best possible approximation.
 *
 * @param mant Number to be multiplied.
 * @param exp Base 5 exponent.
 * @return mant multiplied by 5**exp
 */
static long double mul_pow5(long double mant, int exp)
{
        if (mant == 0.0l || mant == HUGE_VALL) {
                return mant;
        }
        
        if (abs(exp) >> (MAX_POW5 + 1) != 0) {
                /* Too large exponent. */
                errno = ERANGE;
                return exp < 0 ? LDBL_MIN : HUGE_VALL;
        }
        
        if (exp < 0) {
                exp = abs(exp);
                for (int bit = 0; bit <= MAX_POW5; ++bit) {
                        /* Multiply by powers of five bit-by-bit. */
                        if (((exp >> bit) & 1) != 0) {
                                mant /= pow5[bit];
                                if (mant == 0.0l) {
                                        /* Underflow. */
                                        mant = LDBL_MIN;
                                        errno = ERANGE;
                                        break;
                                }
                        }
                }
        } else {
                for (int bit = 0; bit <= MAX_POW5; ++bit) {
                        /* Multiply by powers of five bit-by-bit. */
                        if (((exp >> bit) & 1) != 0) {
                                mant *= pow5[bit];
                                if (mant == HUGE_VALL) {
                                        /* Overflow. */
                                        errno = ERANGE;
                                        break;
                                }
                        }
                }
        }
        
        return mant;
}

/**
 * Multiplies a number by a power of two. This is always exact.
 *
 * @param mant Number to be multiplied.
 * @param exp Base 2 exponent.
 * @return mant multiplied by 2**exp.
 */
static long double mul_pow2(long double mant, int exp)
{
        if (mant == 0.0l || mant == HUGE_VALL) {
                return mant;
        }
        
        if (exp > LDBL_MAX_EXP || exp < LDBL_MIN_EXP) {
                errno = ERANGE;
                return exp < 0 ? LDBL_MIN : HUGE_VALL;
        }
        
        if (exp < 0) {
                exp = abs(exp);
                for (int i = 0; i <= MAX_POW2; ++i) {
                        if (((exp >> i) & 1) != 0) {
                                mant /= pow2[i];
                                if (mant == 0.0l) {
                                        mant = LDBL_MIN;
                                        errno = ERANGE;
                                        break;
                                }
                        }
                }
        } else {
                for (int i = 0; i <= MAX_POW2; ++i) {
                        if (((exp >> i) & 1) != 0) {
                                mant *= pow2[i];
                                if (mant == HUGE_VALL) {
                                        errno = ERANGE;
                                        break;
                                }
                        }
                }
        }
        
        return mant;
}

/* end power functions ********************************************************/



/**
 * Convert decimal string representation of the floating point number.
 * Function expects the string pointer to be already pointed at the first
 * digit (i.e. leading optional sign was already consumed by the caller).
 * 
 * @param sptr Pointer to the storage of the string pointer. Upon successful
 *     conversion, the string pointer is updated to point to the first
 *     unrecognized character.
 * @return An approximate representation of the input floating-point number.
 */
static long double parse_decimal(const char **sptr)
{
        assert(sptr != NULL);
        assert (*sptr != NULL);
        
        const int DEC_BASE = 10;
        const char DECIMAL_POINT = '.';
        const char EXPONENT_MARK = 'e';
        
        const char *str = *sptr;
        long double significand = 0;
        long exponent = 0;
        
        /* number of digits parsed so far */
        int parsed_digits = 0;
        bool after_decimal = false;
        
        while (isdigit(*str) || (!after_decimal && *str == DECIMAL_POINT)) {
                if (*str == DECIMAL_POINT) {
                        after_decimal = true;
                        str++;
                        continue;
                }
                
                if (parsed_digits == 0 && *str == '0') {
                        /* Nothing, just skip leading zeros. */
                } else if (parsed_digits < LDBL_DIG) {
                        significand = significand * DEC_BASE + (*str - '0');
                        parsed_digits++;
                } else {
                        exponent++;
                }
                
                if (after_decimal) {
                        /* Decrement exponent if we are parsing the fractional part. */
                        exponent--;
                }
                
                str++;
        }
        
        /* exponent */
        if (tolower(*str) == EXPONENT_MARK) {
                str++;
                
                /* Returns MIN/MAX value on error, which is ok. */
                long exp = strtol(str, (char **) &str, DEC_BASE);
                
                if (exponent > 0 && exp > LONG_MAX - exponent) {
                        exponent = LONG_MAX;
                } else if (exponent < 0 && exp < LONG_MIN - exponent) {
                        exponent = LONG_MIN;
                } else {
                        exponent += exp;
                }
        }
        
        *sptr = str;
        
        /* Return multiplied by a power of ten. */
        return mul_pow2(mul_pow5(significand, exponent), exponent);
}

/**
 * Derive a hexadecimal digit from its character representation.
 * 
 * @param ch Character representation of the hexadecimal digit.
 * @return Digit value represented by an integer.
 */
static inline int hex_value(char ch)
{
        if (ch <= '9') {
                return ch - '0';
        } else {
                return 10 + tolower(ch) - 'a';
        }
}

/**
 * Convert hexadecimal string representation of the floating point number.
 * Function expects the string pointer to be already pointed at the first
 * digit (i.e. leading optional sign and 0x prefix were already consumed
 * by the caller).
 *
 * @param sptr Pointer to the storage of the string pointer. Upon successful
 *     conversion, the string pointer is updated to point to the first
 *     unrecognized character.
 * @return Representation of the input floating-point number.
 */
static long double parse_hexadecimal(const char **sptr)
{
        assert(sptr != NULL && *sptr != NULL);
        
        const int DEC_BASE = 10;
        const int HEX_BASE = 16;
        const char DECIMAL_POINT = '.';
        const char EXPONENT_MARK = 'p';
        
        const char *str = *sptr;
        long double significand = 0;
        long exponent = 0;
        
        /* number of bits parsed so far */
        int parsed_bits = 0;
        bool after_decimal = false;
        
        while (posix_isxdigit(*str) || (!after_decimal && *str == DECIMAL_POINT)) {
                if (*str == DECIMAL_POINT) {
                        after_decimal = true;
                        str++;
                        continue;
                }
                
                if (parsed_bits == 0 && *str == '0') {
                        /* Nothing, just skip leading zeros. */
                } else if (parsed_bits <= LDBL_MANT_DIG) {
                        significand = significand * HEX_BASE + hex_value(*str);
                        parsed_bits += 4;
                } else {
                        exponent += 4;
                }
                
                if (after_decimal) {
                        exponent -= 4;
                }
                
                str++;
        }
        
        /* exponent */
        if (tolower(*str) == EXPONENT_MARK) {
                str++;
                
                /* Returns MIN/MAX value on error, which is ok. */
                long exp = strtol(str, (char **) &str, DEC_BASE);
                
                if (exponent > 0 && exp > LONG_MAX - exponent) {
                        exponent = LONG_MAX;
                } else if (exponent < 0 && exp < LONG_MIN - exponent) {
                        exponent = LONG_MIN;
                } else {
                        exponent += exp;
                }
        }
        
        *sptr = str;
        
        /* Return multiplied by a power of two. */
        return mul_pow2(significand, exponent);
}

/**
 * Converts a string representation of a floating-point number to
 * its native representation. Largely POSIX compliant, except for
 * locale differences (always uses '.' at the moment) and rounding.
 * Decimal strings are NOT guaranteed to be correctly rounded. This function
 * should return a good enough approximation for most purposes but if you
 * depend on a precise conversion, use hexadecimal representation.
 * Hexadecimal strings are currently always rounded towards zero, regardless
 * of the current rounding mode.
 *
 * @param nptr Input string.
 * @param endptr If non-NULL, *endptr is set to the position of the first
 *     unrecognized character.
 * @return An approximate representation of the input floating-point number.
 */
long double posix_strtold(const char *restrict nptr, char **restrict endptr)
{
        assert(nptr != NULL);
        
        const int RADIX = '.';
        
        /* minus sign */
        bool negative = false;
        /* current position in the string */
        int i = 0;
        
        /* skip whitespace */
        while (isspace(nptr[i])) {
                i++;
        }
        
        /* parse sign */
        switch (nptr[i]) {
        case '-':
                negative = true;
                /* fallthrough */
        case '+':
                i++;
        }
        
        /* check for NaN */
        if (posix_strncasecmp(&nptr[i], "nan", 3) == 0) {
                // FIXME: return NaN
                // TODO: handle the parenthesised case
                
                if (endptr != NULL) {
                        *endptr = (char *) nptr;
                }
                errno = EINVAL;
                return 0;
        }
        
        /* check for Infinity */
        if (posix_strncasecmp(&nptr[i], "inf", 3) == 0) {
                i += 3;
                if (posix_strncasecmp(&nptr[i], "inity", 5) == 0) {
                        i += 5;
                }
                
                if (endptr != NULL) {
                        *endptr = (char *) &nptr[i];
                }
                return negative ? -HUGE_VALL : +HUGE_VALL;
        }

        /* check for a hex number */
        if (nptr[i] == '0' && tolower(nptr[i + 1]) == 'x' &&
            (posix_isxdigit(nptr[i + 2]) ||
            (nptr[i + 2] == RADIX && posix_isxdigit(nptr[i + 3])))) {
                i += 2;
                
                const char *ptr = &nptr[i];
                /* this call sets errno if appropriate. */
                long double result = parse_hexadecimal(&ptr);
                if (endptr != NULL) {
                        *endptr = (char *) ptr;
                }
                return negative ? -result : result;
        }
        
        /* check for a decimal number */
        if (isdigit(nptr[i]) || (nptr[i] == RADIX && isdigit(nptr[i + 1]))) {
                const char *ptr = &nptr[i];
                /* this call sets errno if appropriate. */
                long double result = parse_decimal(&ptr);
                if (endptr != NULL) {
                        *endptr = (char *) ptr;
                }
                return negative ? -result : result;
        }
        
        /* nothing to parse */
        if (endptr != NULL) {
                *endptr = (char *) nptr;
        }
        errno = EINVAL;
        return 0;
}

/** @}
 */