mirror of
https://github.com/ruby/ruby.git
synced 2022-11-09 12:17:21 -05:00
e9cb552ec9
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@61713 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
5621 lines
127 KiB
C
5621 lines
127 KiB
C
/**********************************************************************
|
|
|
|
numeric.c -
|
|
|
|
$Author$
|
|
created at: Fri Aug 13 18:33:09 JST 1993
|
|
|
|
Copyright (C) 1993-2007 Yukihiro Matsumoto
|
|
|
|
**********************************************************************/
|
|
|
|
#include "ruby/encoding.h"
|
|
#include "ruby/util.h"
|
|
#include "internal.h"
|
|
#include "id.h"
|
|
#include <assert.h>
|
|
#include <ctype.h>
|
|
#include <math.h>
|
|
#include <stdio.h>
|
|
|
|
#ifdef HAVE_FLOAT_H
|
|
#include <float.h>
|
|
#endif
|
|
|
|
#ifdef HAVE_IEEEFP_H
|
|
#include <ieeefp.h>
|
|
#endif
|
|
|
|
/* use IEEE 64bit values if not defined */
|
|
#ifndef FLT_RADIX
|
|
#define FLT_RADIX 2
|
|
#endif
|
|
#ifndef FLT_ROUNDS
|
|
#define FLT_ROUNDS 1
|
|
#endif
|
|
#ifndef DBL_MIN
|
|
#define DBL_MIN 2.2250738585072014e-308
|
|
#endif
|
|
#ifndef DBL_MAX
|
|
#define DBL_MAX 1.7976931348623157e+308
|
|
#endif
|
|
#ifndef DBL_MIN_EXP
|
|
#define DBL_MIN_EXP (-1021)
|
|
#endif
|
|
#ifndef DBL_MAX_EXP
|
|
#define DBL_MAX_EXP 1024
|
|
#endif
|
|
#ifndef DBL_MIN_10_EXP
|
|
#define DBL_MIN_10_EXP (-307)
|
|
#endif
|
|
#ifndef DBL_MAX_10_EXP
|
|
#define DBL_MAX_10_EXP 308
|
|
#endif
|
|
#ifndef DBL_DIG
|
|
#define DBL_DIG 15
|
|
#endif
|
|
#ifndef DBL_MANT_DIG
|
|
#define DBL_MANT_DIG 53
|
|
#endif
|
|
#ifndef DBL_EPSILON
|
|
#define DBL_EPSILON 2.2204460492503131e-16
|
|
#endif
|
|
|
|
#ifdef HAVE_INFINITY
|
|
#elif !defined(WORDS_BIGENDIAN) /* BYTE_ORDER == LITTLE_ENDIAN */
|
|
const union bytesequence4_or_float rb_infinity = {{0x00, 0x00, 0x80, 0x7f}};
|
|
#else
|
|
const union bytesequence4_or_float rb_infinity = {{0x7f, 0x80, 0x00, 0x00}};
|
|
#endif
|
|
|
|
#ifdef HAVE_NAN
|
|
#elif !defined(WORDS_BIGENDIAN) /* BYTE_ORDER == LITTLE_ENDIAN */
|
|
const union bytesequence4_or_float rb_nan = {{0x00, 0x00, 0xc0, 0x7f}};
|
|
#else
|
|
const union bytesequence4_or_float rb_nan = {{0x7f, 0xc0, 0x00, 0x00}};
|
|
#endif
|
|
|
|
#ifndef HAVE_ROUND
|
|
double
|
|
round(double x)
|
|
{
|
|
double f;
|
|
|
|
if (x > 0.0) {
|
|
f = floor(x);
|
|
x = f + (x - f >= 0.5);
|
|
}
|
|
else if (x < 0.0) {
|
|
f = ceil(x);
|
|
x = f - (f - x >= 0.5);
|
|
}
|
|
return x;
|
|
}
|
|
#endif
|
|
|
|
static double
|
|
round_half_up(double x, double s)
|
|
{
|
|
double f, xs = x * s;
|
|
|
|
f = round(xs);
|
|
if (s == 1.0) return f;
|
|
if (x > 0) {
|
|
if ((double)((f + 0.5) / s) <= x) f += 1;
|
|
x = f;
|
|
}
|
|
else {
|
|
if ((double)((f - 0.5) / s) >= x) f -= 1;
|
|
x = f;
|
|
}
|
|
return x;
|
|
}
|
|
|
|
static double
|
|
round_half_down(double x, double s)
|
|
{
|
|
double f, xs = x * s;
|
|
|
|
f = round(xs);
|
|
if (x > 0) {
|
|
if ((double)((f - 0.5) / s) >= x) f -= 1;
|
|
x = f;
|
|
}
|
|
else {
|
|
if ((double)((f + 0.5) / s) <= x) f += 1;
|
|
x = f;
|
|
}
|
|
return x;
|
|
}
|
|
|
|
static double
|
|
round_half_even(double x, double s)
|
|
{
|
|
double f, d, xs = x * s;
|
|
|
|
if (x > 0.0) {
|
|
f = floor(xs);
|
|
d = xs - f;
|
|
if (d > 0.5)
|
|
d = 1.0;
|
|
else if (d == 0.5 || ((double)((f + 0.5) / s) <= x))
|
|
d = fmod(f, 2.0);
|
|
else
|
|
d = 0.0;
|
|
x = f + d;
|
|
}
|
|
else if (x < 0.0) {
|
|
f = ceil(xs);
|
|
d = f - xs;
|
|
if (d > 0.5)
|
|
d = 1.0;
|
|
else if (d == 0.5 || ((double)((f - 0.5) / s) >= x))
|
|
d = fmod(-f, 2.0);
|
|
else
|
|
d = 0.0;
|
|
x = f - d;
|
|
}
|
|
return x;
|
|
}
|
|
|
|
static VALUE fix_uminus(VALUE num);
|
|
static VALUE fix_mul(VALUE x, VALUE y);
|
|
static VALUE fix_lshift(long, unsigned long);
|
|
static VALUE fix_rshift(long, unsigned long);
|
|
static VALUE int_pow(long x, unsigned long y);
|
|
static VALUE int_even_p(VALUE x);
|
|
static int int_round_zero_p(VALUE num, int ndigits);
|
|
VALUE rb_int_floor(VALUE num, int ndigits);
|
|
VALUE rb_int_ceil(VALUE num, int ndigits);
|
|
static VALUE flo_to_i(VALUE num);
|
|
static int float_round_overflow(int ndigits, int binexp);
|
|
static int float_round_underflow(int ndigits, int binexp);
|
|
|
|
static ID id_coerce, id_div, id_divmod;
|
|
#define id_to_i idTo_i
|
|
#define id_eq idEq
|
|
#define id_cmp idCmp
|
|
|
|
VALUE rb_cNumeric;
|
|
VALUE rb_cFloat;
|
|
VALUE rb_cInteger;
|
|
#ifndef RUBY_INTEGER_UNIFICATION
|
|
VALUE rb_cFixnum;
|
|
#endif
|
|
|
|
VALUE rb_eZeroDivError;
|
|
VALUE rb_eFloatDomainError;
|
|
|
|
static ID id_to, id_by;
|
|
|
|
void
|
|
rb_num_zerodiv(void)
|
|
{
|
|
rb_raise(rb_eZeroDivError, "divided by 0");
|
|
}
|
|
|
|
enum ruby_num_rounding_mode
|
|
rb_num_get_rounding_option(VALUE opts)
|
|
{
|
|
static ID round_kwds[1];
|
|
VALUE rounding;
|
|
VALUE str;
|
|
const char *s;
|
|
|
|
if (!NIL_P(opts)) {
|
|
if (!round_kwds[0]) {
|
|
round_kwds[0] = rb_intern_const("half");
|
|
}
|
|
if (!rb_get_kwargs(opts, round_kwds, 0, 1, &rounding)) goto noopt;
|
|
if (SYMBOL_P(rounding)) {
|
|
str = rb_sym2str(rounding);
|
|
}
|
|
else if (NIL_P(rounding)) {
|
|
goto noopt;
|
|
}
|
|
else if (!RB_TYPE_P(str = rounding, T_STRING)) {
|
|
str = rb_check_string_type(rounding);
|
|
if (NIL_P(str)) goto invalid;
|
|
}
|
|
s = RSTRING_PTR(str);
|
|
switch (RSTRING_LEN(str)) {
|
|
case 2:
|
|
if (rb_memcicmp(s, "up", 2) == 0)
|
|
return RUBY_NUM_ROUND_HALF_UP;
|
|
break;
|
|
case 4:
|
|
if (rb_memcicmp(s, "even", 4) == 0)
|
|
return RUBY_NUM_ROUND_HALF_EVEN;
|
|
if (strncasecmp(s, "down", 4) == 0)
|
|
return RUBY_NUM_ROUND_HALF_DOWN;
|
|
break;
|
|
}
|
|
invalid:
|
|
rb_raise(rb_eArgError, "invalid rounding mode: % "PRIsVALUE, rounding);
|
|
}
|
|
noopt:
|
|
return RUBY_NUM_ROUND_DEFAULT;
|
|
}
|
|
|
|
/* experimental API */
|
|
int
|
|
rb_num_to_uint(VALUE val, unsigned int *ret)
|
|
{
|
|
#define NUMERR_TYPE 1
|
|
#define NUMERR_NEGATIVE 2
|
|
#define NUMERR_TOOLARGE 3
|
|
if (FIXNUM_P(val)) {
|
|
long v = FIX2LONG(val);
|
|
#if SIZEOF_INT < SIZEOF_LONG
|
|
if (v > (long)UINT_MAX) return NUMERR_TOOLARGE;
|
|
#endif
|
|
if (v < 0) return NUMERR_NEGATIVE;
|
|
*ret = (unsigned int)v;
|
|
return 0;
|
|
}
|
|
|
|
if (RB_TYPE_P(val, T_BIGNUM)) {
|
|
if (BIGNUM_NEGATIVE_P(val)) return NUMERR_NEGATIVE;
|
|
#if SIZEOF_INT < SIZEOF_LONG
|
|
/* long is 64bit */
|
|
return NUMERR_TOOLARGE;
|
|
#else
|
|
/* long is 32bit */
|
|
if (rb_absint_size(val, NULL) > sizeof(int)) return NUMERR_TOOLARGE;
|
|
*ret = (unsigned int)rb_big2ulong((VALUE)val);
|
|
return 0;
|
|
#endif
|
|
}
|
|
return NUMERR_TYPE;
|
|
}
|
|
|
|
#define method_basic_p(klass) rb_method_basic_definition_p(klass, mid)
|
|
|
|
static inline int
|
|
int_pos_p(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return FIXNUM_POSITIVE_P(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return BIGNUM_POSITIVE_P(num);
|
|
}
|
|
rb_raise(rb_eTypeError, "not an Integer");
|
|
}
|
|
|
|
static inline int
|
|
int_neg_p(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return FIXNUM_NEGATIVE_P(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return BIGNUM_NEGATIVE_P(num);
|
|
}
|
|
rb_raise(rb_eTypeError, "not an Integer");
|
|
}
|
|
|
|
int
|
|
rb_num_negative_p(VALUE num)
|
|
{
|
|
return rb_num_negative_int_p(num);
|
|
}
|
|
|
|
static VALUE
|
|
num_funcall_op_0(VALUE x, VALUE arg, int recursive)
|
|
{
|
|
ID func = (ID)arg;
|
|
if (recursive) {
|
|
const char *name = rb_id2name(func);
|
|
if (ISALNUM(name[0])) {
|
|
rb_name_error(func, "%"PRIsVALUE".%"PRIsVALUE,
|
|
x, ID2SYM(func));
|
|
}
|
|
else if (name[0] && name[1] == '@' && !name[2]) {
|
|
rb_name_error(func, "%c%"PRIsVALUE,
|
|
name[0], x);
|
|
}
|
|
else {
|
|
rb_name_error(func, "%"PRIsVALUE"%"PRIsVALUE,
|
|
ID2SYM(func), x);
|
|
}
|
|
}
|
|
return rb_funcallv(x, func, 0, 0);
|
|
}
|
|
|
|
static VALUE
|
|
num_funcall0(VALUE x, ID func)
|
|
{
|
|
return rb_exec_recursive(num_funcall_op_0, x, (VALUE)func);
|
|
}
|
|
|
|
static void
|
|
num_funcall_op_1_recursion(VALUE x, ID func, VALUE y)
|
|
{
|
|
const char *name = rb_id2name(func);
|
|
if (ISALNUM(name[0])) {
|
|
rb_name_error(func, "%"PRIsVALUE".%"PRIsVALUE"(%"PRIsVALUE")",
|
|
x, ID2SYM(func), y);
|
|
}
|
|
else {
|
|
rb_name_error(func, "%"PRIsVALUE"%"PRIsVALUE"%"PRIsVALUE,
|
|
x, ID2SYM(func), y);
|
|
}
|
|
}
|
|
|
|
static VALUE
|
|
num_funcall_op_1(VALUE y, VALUE arg, int recursive)
|
|
{
|
|
ID func = (ID)((VALUE *)arg)[0];
|
|
VALUE x = ((VALUE *)arg)[1];
|
|
if (recursive) {
|
|
num_funcall_op_1_recursion(x, func, y);
|
|
}
|
|
return rb_funcall(x, func, 1, y);
|
|
}
|
|
|
|
static VALUE
|
|
num_funcall1(VALUE x, ID func, VALUE y)
|
|
{
|
|
VALUE args[2];
|
|
args[0] = (VALUE)func;
|
|
args[1] = x;
|
|
return rb_exec_recursive_paired(num_funcall_op_1, y, x, (VALUE)args);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.coerce(numeric) -> array
|
|
*
|
|
* If +numeric+ is the same type as +num+, returns an array
|
|
* <code>[numeric, num]</code>. Otherwise, returns an array with both
|
|
* +numeric+ and +num+ represented as Float objects.
|
|
*
|
|
* This coercion mechanism is used by Ruby to handle mixed-type numeric
|
|
* operations: it is intended to find a compatible common type between the two
|
|
* operands of the operator.
|
|
*
|
|
* 1.coerce(2.5) #=> [2.5, 1.0]
|
|
* 1.2.coerce(3) #=> [3.0, 1.2]
|
|
* 1.coerce(2) #=> [2, 1]
|
|
*/
|
|
|
|
static VALUE
|
|
num_coerce(VALUE x, VALUE y)
|
|
{
|
|
if (CLASS_OF(x) == CLASS_OF(y))
|
|
return rb_assoc_new(y, x);
|
|
x = rb_Float(x);
|
|
y = rb_Float(y);
|
|
return rb_assoc_new(y, x);
|
|
}
|
|
|
|
NORETURN(static void coerce_failed(VALUE x, VALUE y));
|
|
static void
|
|
coerce_failed(VALUE x, VALUE y)
|
|
{
|
|
if (SPECIAL_CONST_P(y) || BUILTIN_TYPE(y) == T_FLOAT) {
|
|
y = rb_inspect(y);
|
|
}
|
|
else {
|
|
y = rb_obj_class(y);
|
|
}
|
|
rb_raise(rb_eTypeError, "%"PRIsVALUE" can't be coerced into %"PRIsVALUE,
|
|
y, rb_obj_class(x));
|
|
}
|
|
|
|
static int
|
|
do_coerce(VALUE *x, VALUE *y, int err)
|
|
{
|
|
VALUE ary = rb_check_funcall(*y, id_coerce, 1, x);
|
|
if (ary == Qundef) {
|
|
if (err) {
|
|
coerce_failed(*x, *y);
|
|
}
|
|
return FALSE;
|
|
}
|
|
if (!err && NIL_P(ary)) {
|
|
return FALSE;
|
|
}
|
|
if (!RB_TYPE_P(ary, T_ARRAY) || RARRAY_LEN(ary) != 2) {
|
|
rb_raise(rb_eTypeError, "coerce must return [x, y]");
|
|
}
|
|
|
|
*x = RARRAY_AREF(ary, 0);
|
|
*y = RARRAY_AREF(ary, 1);
|
|
return TRUE;
|
|
}
|
|
|
|
VALUE
|
|
rb_num_coerce_bin(VALUE x, VALUE y, ID func)
|
|
{
|
|
do_coerce(&x, &y, TRUE);
|
|
return rb_funcall(x, func, 1, y);
|
|
}
|
|
|
|
VALUE
|
|
rb_num_coerce_cmp(VALUE x, VALUE y, ID func)
|
|
{
|
|
if (do_coerce(&x, &y, FALSE))
|
|
return rb_funcall(x, func, 1, y);
|
|
return Qnil;
|
|
}
|
|
|
|
VALUE
|
|
rb_num_coerce_relop(VALUE x, VALUE y, ID func)
|
|
{
|
|
VALUE c, x0 = x, y0 = y;
|
|
|
|
if (!do_coerce(&x, &y, FALSE) ||
|
|
NIL_P(c = rb_funcall(x, func, 1, y))) {
|
|
rb_cmperr(x0, y0);
|
|
return Qnil; /* not reached */
|
|
}
|
|
return c;
|
|
}
|
|
|
|
/*
|
|
* :nodoc:
|
|
*
|
|
* Trap attempts to add methods to Numeric objects. Always raises a TypeError.
|
|
*
|
|
* Numerics should be values; singleton_methods should not be added to them.
|
|
*/
|
|
|
|
static VALUE
|
|
num_sadded(VALUE x, VALUE name)
|
|
{
|
|
ID mid = rb_to_id(name);
|
|
/* ruby_frame = ruby_frame->prev; */ /* pop frame for "singleton_method_added" */
|
|
rb_remove_method_id(rb_singleton_class(x), mid);
|
|
rb_raise(rb_eTypeError,
|
|
"can't define singleton method \"%"PRIsVALUE"\" for %"PRIsVALUE,
|
|
rb_id2str(mid),
|
|
rb_obj_class(x));
|
|
|
|
UNREACHABLE;
|
|
}
|
|
|
|
#if 0
|
|
/*
|
|
* call-seq:
|
|
* num.clone(freeze: true) -> num
|
|
*
|
|
* Returns the receiver. +freeze+ cannot be +false+.
|
|
*/
|
|
static VALUE
|
|
num_clone(int argc, VALUE *argv, VALUE x)
|
|
{
|
|
return rb_immutable_obj_clone(argc, argv, x);
|
|
}
|
|
#else
|
|
# define num_clone rb_immutable_obj_clone
|
|
#endif
|
|
|
|
#if 0
|
|
/*
|
|
* call-seq:
|
|
* num.dup -> num
|
|
*
|
|
* Returns the receiver.
|
|
*/
|
|
static VALUE
|
|
num_dup(VALUE x)
|
|
{
|
|
return x;
|
|
}
|
|
#else
|
|
# define num_dup num_uplus
|
|
#endif
|
|
|
|
/*
|
|
* call-seq:
|
|
* +num -> num
|
|
*
|
|
* Unary Plus---Returns the receiver.
|
|
*/
|
|
|
|
static VALUE
|
|
num_uplus(VALUE num)
|
|
{
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.i -> Complex(0, num)
|
|
*
|
|
* Returns the corresponding imaginary number.
|
|
* Not available for complex numbers.
|
|
*
|
|
* -42.i #=> (0-42i)
|
|
* 2.0.i #=> (0+2.0i)
|
|
*/
|
|
|
|
static VALUE
|
|
num_imaginary(VALUE num)
|
|
{
|
|
return rb_complex_new(INT2FIX(0), num);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* -num -> numeric
|
|
*
|
|
* Unary Minus---Returns the receiver, negated.
|
|
*/
|
|
|
|
static VALUE
|
|
num_uminus(VALUE num)
|
|
{
|
|
VALUE zero;
|
|
|
|
zero = INT2FIX(0);
|
|
do_coerce(&zero, &num, TRUE);
|
|
|
|
return num_funcall1(zero, '-', num);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.fdiv(numeric) -> float
|
|
*
|
|
* Returns float division.
|
|
*/
|
|
|
|
static VALUE
|
|
num_fdiv(VALUE x, VALUE y)
|
|
{
|
|
return rb_funcall(rb_Float(x), '/', 1, y);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.div(numeric) -> integer
|
|
*
|
|
* Uses +/+ to perform division, then converts the result to an integer.
|
|
* Numeric does not define the +/+ operator; this is left to subclasses.
|
|
*
|
|
* Equivalent to <code>num.divmod(numeric)[0]</code>.
|
|
*
|
|
* See Numeric#divmod.
|
|
*/
|
|
|
|
static VALUE
|
|
num_div(VALUE x, VALUE y)
|
|
{
|
|
if (rb_equal(INT2FIX(0), y)) rb_num_zerodiv();
|
|
return rb_funcall(num_funcall1(x, '/', y), rb_intern("floor"), 0);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.modulo(numeric) -> real
|
|
*
|
|
* <code>x.modulo(y)</code> means <code>x-y*(x/y).floor</code>.
|
|
*
|
|
* Equivalent to <code>num.divmod(numeric)[1]</code>.
|
|
*
|
|
* See Numeric#divmod.
|
|
*/
|
|
|
|
static VALUE
|
|
num_modulo(VALUE x, VALUE y)
|
|
{
|
|
VALUE q = num_funcall1(x, id_div, y);
|
|
return rb_funcall(x, '-', 1,
|
|
rb_funcall(y, '*', 1, q));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.remainder(numeric) -> real
|
|
*
|
|
* <code>x.remainder(y)</code> means <code>x-y*(x/y).truncate</code>.
|
|
*
|
|
* See Numeric#divmod.
|
|
*/
|
|
|
|
static VALUE
|
|
num_remainder(VALUE x, VALUE y)
|
|
{
|
|
VALUE z = num_funcall1(x, '%', y);
|
|
|
|
if ((!rb_equal(z, INT2FIX(0))) &&
|
|
((rb_num_negative_int_p(x) &&
|
|
rb_num_positive_int_p(y)) ||
|
|
(rb_num_positive_int_p(x) &&
|
|
rb_num_negative_int_p(y)))) {
|
|
return rb_funcall(z, '-', 1, y);
|
|
}
|
|
return z;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.divmod(numeric) -> array
|
|
*
|
|
* Returns an array containing the quotient and modulus obtained by dividing
|
|
* +num+ by +numeric+.
|
|
*
|
|
* If <code>q, r = x.divmod(y)</code>, then
|
|
*
|
|
* q = floor(x/y)
|
|
* x = q*y + r
|
|
*
|
|
* The quotient is rounded toward negative infinity, as shown in the
|
|
* following table:
|
|
*
|
|
* a | b | a.divmod(b) | a/b | a.modulo(b) | a.remainder(b)
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* 13 | 4 | 3, 1 | 3 | 1 | 1
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* 13 | -4 | -4, -3 | -4 | -3 | 1
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* -13 | 4 | -4, 3 | -4 | 3 | -1
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* -13 | -4 | 3, -1 | 3 | -1 | -1
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* 11.5 | 4 | 2, 3.5 | 2.875 | 3.5 | 3.5
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* 11.5 | -4 | -3, -0.5 | -2.875 | -0.5 | 3.5
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* -11.5 | 4 | -3, 0.5 | -2.875 | 0.5 | -3.5
|
|
* ------+-----+---------------+---------+-------------+---------------
|
|
* -11.5 | -4 | 2, -3.5 | 2.875 | -3.5 | -3.5
|
|
*
|
|
*
|
|
* Examples
|
|
*
|
|
* 11.divmod(3) #=> [3, 2]
|
|
* 11.divmod(-3) #=> [-4, -1]
|
|
* 11.divmod(3.5) #=> [3, 0.5]
|
|
* (-11).divmod(3.5) #=> [-4, 3.0]
|
|
* 11.5.divmod(3.5) #=> [3, 1.0]
|
|
*/
|
|
|
|
static VALUE
|
|
num_divmod(VALUE x, VALUE y)
|
|
{
|
|
return rb_assoc_new(num_div(x, y), num_modulo(x, y));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.real? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ is a real number (i.e. not Complex).
|
|
*/
|
|
|
|
static VALUE
|
|
num_real_p(VALUE num)
|
|
{
|
|
return Qtrue;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.integer? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ is an Integer.
|
|
*
|
|
* 1.0.integer? #=> false
|
|
* 1.integer? #=> true
|
|
*/
|
|
|
|
static VALUE
|
|
num_int_p(VALUE num)
|
|
{
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.abs -> numeric
|
|
* num.magnitude -> numeric
|
|
*
|
|
* Returns the absolute value of +num+.
|
|
*
|
|
* 12.abs #=> 12
|
|
* (-34.56).abs #=> 34.56
|
|
* -34.56.abs #=> 34.56
|
|
*
|
|
* Numeric#magnitude is an alias for Numeric#abs.
|
|
*/
|
|
|
|
static VALUE
|
|
num_abs(VALUE num)
|
|
{
|
|
if (rb_num_negative_int_p(num)) {
|
|
return num_funcall0(num, idUMinus);
|
|
}
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.zero? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ has a zero value.
|
|
*/
|
|
|
|
static VALUE
|
|
num_zero_p(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
if (FIXNUM_ZERO_P(num)) {
|
|
return Qtrue;
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
if (rb_bigzero_p(num)) {
|
|
/* this should not happen usually */
|
|
return Qtrue;
|
|
}
|
|
}
|
|
else if (rb_equal(num, INT2FIX(0))) {
|
|
return Qtrue;
|
|
}
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.nonzero? -> self or nil
|
|
*
|
|
* Returns +self+ if +num+ is not zero, +nil+ otherwise.
|
|
*
|
|
* This behavior is useful when chaining comparisons:
|
|
*
|
|
* a = %w( z Bb bB bb BB a aA Aa AA A )
|
|
* b = a.sort {|a,b| (a.downcase <=> b.downcase).nonzero? || a <=> b }
|
|
* b #=> ["A", "a", "AA", "Aa", "aA", "BB", "Bb", "bB", "bb", "z"]
|
|
*/
|
|
|
|
static VALUE
|
|
num_nonzero_p(VALUE num)
|
|
{
|
|
if (RTEST(num_funcall0(num, rb_intern("zero?")))) {
|
|
return Qnil;
|
|
}
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.finite? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ is a finite number, otherwise returns +false+.
|
|
*/
|
|
static VALUE
|
|
num_finite_p(VALUE num)
|
|
{
|
|
return Qtrue;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.infinite? -> -1, 1, or nil
|
|
*
|
|
* Returns +nil+, -1, or 1 depending on whether the value is
|
|
* finite, <code>-Infinity</code>, or <code>+Infinity</code>.
|
|
*/
|
|
static VALUE
|
|
num_infinite_p(VALUE num)
|
|
{
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.to_int -> integer
|
|
*
|
|
* Invokes the child class's +to_i+ method to convert +num+ to an integer.
|
|
*
|
|
* 1.0.class #=> Float
|
|
* 1.0.to_int.class #=> Integer
|
|
* 1.0.to_i.class #=> Integer
|
|
*/
|
|
|
|
static VALUE
|
|
num_to_int(VALUE num)
|
|
{
|
|
return num_funcall0(num, id_to_i);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.positive? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ is greater than 0.
|
|
*/
|
|
|
|
static VALUE
|
|
num_positive_p(VALUE num)
|
|
{
|
|
const ID mid = '>';
|
|
|
|
if (FIXNUM_P(num)) {
|
|
if (method_basic_p(rb_cInteger))
|
|
return (SIGNED_VALUE)num > (SIGNED_VALUE)INT2FIX(0) ? Qtrue : Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
if (method_basic_p(rb_cInteger))
|
|
return BIGNUM_POSITIVE_P(num) && !rb_bigzero_p(num) ? Qtrue : Qfalse;
|
|
}
|
|
return rb_num_compare_with_zero(num, mid);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.negative? -> true or false
|
|
*
|
|
* Returns +true+ if +num+ is less than 0.
|
|
*/
|
|
|
|
static VALUE
|
|
num_negative_p(VALUE num)
|
|
{
|
|
return rb_num_negative_int_p(num) ? Qtrue : Qfalse;
|
|
}
|
|
|
|
|
|
/********************************************************************
|
|
*
|
|
* Document-class: Float
|
|
*
|
|
* Float objects represent inexact real numbers using the native
|
|
* architecture's double-precision floating point representation.
|
|
*
|
|
* Floating point has a different arithmetic and is an inexact number.
|
|
* So you should know its esoteric system. See following:
|
|
*
|
|
* - http://docs.sun.com/source/806-3568/ncg_goldberg.html
|
|
* - http://wiki.github.com/rdp/ruby_tutorials_core/ruby-talk-faq#wiki-floats_imprecise
|
|
* - http://en.wikipedia.org/wiki/Floating_point#Accuracy_problems
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_new_in_heap(double d)
|
|
{
|
|
NEWOBJ_OF(flt, struct RFloat, rb_cFloat, T_FLOAT | (RGENGC_WB_PROTECTED_FLOAT ? FL_WB_PROTECTED : 0));
|
|
|
|
flt->float_value = d;
|
|
OBJ_FREEZE(flt);
|
|
return (VALUE)flt;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.to_s -> string
|
|
*
|
|
* Returns a string containing a representation of +self+.
|
|
* As well as a fixed or exponential form of the +float+,
|
|
* the call may return +NaN+, +Infinity+, and +-Infinity+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_to_s(VALUE flt)
|
|
{
|
|
enum {decimal_mant = DBL_MANT_DIG-DBL_DIG};
|
|
enum {float_dig = DBL_DIG+1};
|
|
char buf[float_dig + (decimal_mant + CHAR_BIT - 1) / CHAR_BIT + 10];
|
|
double value = RFLOAT_VALUE(flt);
|
|
VALUE s;
|
|
char *p, *e;
|
|
int sign, decpt, digs;
|
|
|
|
if (isinf(value)) {
|
|
static const char minf[] = "-Infinity";
|
|
const int pos = (value > 0); /* skip "-" */
|
|
return rb_usascii_str_new(minf+pos, strlen(minf)-pos);
|
|
}
|
|
else if (isnan(value))
|
|
return rb_usascii_str_new2("NaN");
|
|
|
|
p = ruby_dtoa(value, 0, 0, &decpt, &sign, &e);
|
|
s = sign ? rb_usascii_str_new_cstr("-") : rb_usascii_str_new(0, 0);
|
|
if ((digs = (int)(e - p)) >= (int)sizeof(buf)) digs = (int)sizeof(buf) - 1;
|
|
memcpy(buf, p, digs);
|
|
xfree(p);
|
|
if (decpt > 0) {
|
|
if (decpt < digs) {
|
|
memmove(buf + decpt + 1, buf + decpt, digs - decpt);
|
|
buf[decpt] = '.';
|
|
rb_str_cat(s, buf, digs + 1);
|
|
}
|
|
else if (decpt <= DBL_DIG) {
|
|
long len;
|
|
char *ptr;
|
|
rb_str_cat(s, buf, digs);
|
|
rb_str_resize(s, (len = RSTRING_LEN(s)) + decpt - digs + 2);
|
|
ptr = RSTRING_PTR(s) + len;
|
|
if (decpt > digs) {
|
|
memset(ptr, '0', decpt - digs);
|
|
ptr += decpt - digs;
|
|
}
|
|
memcpy(ptr, ".0", 2);
|
|
}
|
|
else {
|
|
goto exp;
|
|
}
|
|
}
|
|
else if (decpt > -4) {
|
|
long len;
|
|
char *ptr;
|
|
rb_str_cat(s, "0.", 2);
|
|
rb_str_resize(s, (len = RSTRING_LEN(s)) - decpt + digs);
|
|
ptr = RSTRING_PTR(s);
|
|
memset(ptr += len, '0', -decpt);
|
|
memcpy(ptr -= decpt, buf, digs);
|
|
}
|
|
else {
|
|
exp:
|
|
if (digs > 1) {
|
|
memmove(buf + 2, buf + 1, digs - 1);
|
|
}
|
|
else {
|
|
buf[2] = '0';
|
|
digs++;
|
|
}
|
|
buf[1] = '.';
|
|
rb_str_cat(s, buf, digs + 1);
|
|
rb_str_catf(s, "e%+03d", decpt - 1);
|
|
}
|
|
return s;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.coerce(numeric) -> array
|
|
*
|
|
* Returns an array with both +numeric+ and +float+ represented as Float
|
|
* objects.
|
|
*
|
|
* This is achieved by converting +numeric+ to a Float.
|
|
*
|
|
* 1.2.coerce(3) #=> [3.0, 1.2]
|
|
* 2.5.coerce(1.1) #=> [1.1, 2.5]
|
|
*/
|
|
|
|
static VALUE
|
|
flo_coerce(VALUE x, VALUE y)
|
|
{
|
|
return rb_assoc_new(rb_Float(y), x);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* -float -> float
|
|
*
|
|
* Returns +float+, negated.
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_uminus(VALUE flt)
|
|
{
|
|
return DBL2NUM(-RFLOAT_VALUE(flt));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float + other -> float
|
|
*
|
|
* Returns a new Float which is the sum of +float+ and +other+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_plus(VALUE x, VALUE y)
|
|
{
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) + (double)FIX2LONG(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) + rb_big2dbl(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) + RFLOAT_VALUE(y));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '+');
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float - other -> float
|
|
*
|
|
* Returns a new Float which is the difference of +float+ and +other+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_minus(VALUE x, VALUE y)
|
|
{
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) - (double)FIX2LONG(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) - rb_big2dbl(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) - RFLOAT_VALUE(y));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '-');
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float * other -> float
|
|
*
|
|
* Returns a new Float which is the product of +float+ and +other+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_mul(VALUE x, VALUE y)
|
|
{
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) * (double)FIX2LONG(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) * rb_big2dbl(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) * RFLOAT_VALUE(y));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '*');
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float / other -> float
|
|
*
|
|
* Returns a new Float which is the result of dividing +float+ by +other+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_div(VALUE x, VALUE y)
|
|
{
|
|
long f_y;
|
|
double d;
|
|
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
f_y = FIX2LONG(y);
|
|
return DBL2NUM(RFLOAT_VALUE(x) / (double)f_y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
d = rb_big2dbl(y);
|
|
return DBL2NUM(RFLOAT_VALUE(x) / d);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM(RFLOAT_VALUE(x) / RFLOAT_VALUE(y));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '/');
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.fdiv(numeric) -> float
|
|
* float.quo(numeric) -> float
|
|
*
|
|
* Returns <code>float / numeric</code>, same as Float#/.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_quo(VALUE x, VALUE y)
|
|
{
|
|
return num_funcall1(x, '/', y);
|
|
}
|
|
|
|
static void
|
|
flodivmod(double x, double y, double *divp, double *modp)
|
|
{
|
|
double div, mod;
|
|
|
|
if (isnan(y)) {
|
|
/* y is NaN so all results are NaN */
|
|
if (modp) *modp = y;
|
|
if (divp) *divp = y;
|
|
return;
|
|
}
|
|
if (y == 0.0) rb_num_zerodiv();
|
|
if ((x == 0.0) || (isinf(y) && !isinf(x)))
|
|
mod = x;
|
|
else {
|
|
#ifdef HAVE_FMOD
|
|
mod = fmod(x, y);
|
|
#else
|
|
double z;
|
|
|
|
modf(x/y, &z);
|
|
mod = x - z * y;
|
|
#endif
|
|
}
|
|
if (isinf(x) && !isinf(y))
|
|
div = x;
|
|
else {
|
|
div = (x - mod) / y;
|
|
if (modp && divp) div = round(div);
|
|
}
|
|
if (y*mod < 0) {
|
|
mod += y;
|
|
div -= 1.0;
|
|
}
|
|
if (modp) *modp = mod;
|
|
if (divp) *divp = div;
|
|
}
|
|
|
|
/*
|
|
* Returns the modulo of division of x by y.
|
|
* An error will be raised if y == 0.
|
|
*/
|
|
|
|
double
|
|
ruby_float_mod(double x, double y)
|
|
{
|
|
double mod;
|
|
flodivmod(x, y, 0, &mod);
|
|
return mod;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float % other -> float
|
|
* float.modulo(other) -> float
|
|
*
|
|
* Returns the modulo after division of +float+ by +other+.
|
|
*
|
|
* 6543.21.modulo(137) #=> 104.21000000000004
|
|
* 6543.21.modulo(137.24) #=> 92.92999999999961
|
|
*/
|
|
|
|
static VALUE
|
|
flo_mod(VALUE x, VALUE y)
|
|
{
|
|
double fy;
|
|
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
fy = (double)FIX2LONG(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
fy = rb_big2dbl(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
fy = RFLOAT_VALUE(y);
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '%');
|
|
}
|
|
return DBL2NUM(ruby_float_mod(RFLOAT_VALUE(x), fy));
|
|
}
|
|
|
|
static VALUE
|
|
dbl2ival(double d)
|
|
{
|
|
if (FIXABLE(d)) {
|
|
return LONG2FIX((long)d);
|
|
}
|
|
return rb_dbl2big(d);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.divmod(numeric) -> array
|
|
*
|
|
* See Numeric#divmod.
|
|
*
|
|
* 42.0.divmod(6) #=> [7, 0.0]
|
|
* 42.0.divmod(5) #=> [8, 2.0]
|
|
*/
|
|
|
|
static VALUE
|
|
flo_divmod(VALUE x, VALUE y)
|
|
{
|
|
double fy, div, mod;
|
|
volatile VALUE a, b;
|
|
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
fy = (double)FIX2LONG(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
fy = rb_big2dbl(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
fy = RFLOAT_VALUE(y);
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, id_divmod);
|
|
}
|
|
flodivmod(RFLOAT_VALUE(x), fy, &div, &mod);
|
|
a = dbl2ival(div);
|
|
b = DBL2NUM(mod);
|
|
return rb_assoc_new(a, b);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float ** other -> float
|
|
*
|
|
* Raises +float+ to the power of +other+.
|
|
*
|
|
* 2.0**3 #=> 8.0
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_pow(VALUE x, VALUE y)
|
|
{
|
|
double dx, dy;
|
|
if (RB_TYPE_P(y, T_FIXNUM)) {
|
|
dx = RFLOAT_VALUE(x);
|
|
dy = (double)FIX2LONG(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
dx = RFLOAT_VALUE(x);
|
|
dy = rb_big2dbl(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
dx = RFLOAT_VALUE(x);
|
|
dy = RFLOAT_VALUE(y);
|
|
if (dx < 0 && dy != round(dy))
|
|
return num_funcall1(rb_complex_raw1(x), idPow, y);
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, idPow);
|
|
}
|
|
return DBL2NUM(pow(dx, dy));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.eql?(numeric) -> true or false
|
|
*
|
|
* Returns +true+ if +num+ and +numeric+ are the same type and have equal
|
|
* values. Contrast this with Numeric#==, which performs type conversions.
|
|
*
|
|
* 1 == 1.0 #=> true
|
|
* 1.eql?(1.0) #=> false
|
|
* 1.0.eql?(1.0) #=> true
|
|
*/
|
|
|
|
static VALUE
|
|
num_eql(VALUE x, VALUE y)
|
|
{
|
|
if (TYPE(x) != TYPE(y)) return Qfalse;
|
|
|
|
if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_eql(x, y);
|
|
}
|
|
|
|
return rb_equal(x, y);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* number <=> other -> 0 or nil
|
|
*
|
|
* Returns zero if +number+ equals +other+, otherwise returns +nil+.
|
|
*/
|
|
|
|
static VALUE
|
|
num_cmp(VALUE x, VALUE y)
|
|
{
|
|
if (x == y) return INT2FIX(0);
|
|
return Qnil;
|
|
}
|
|
|
|
static VALUE
|
|
num_equal(VALUE x, VALUE y)
|
|
{
|
|
VALUE result;
|
|
if (x == y) return Qtrue;
|
|
result = num_funcall1(y, id_eq, x);
|
|
if (RTEST(result)) return Qtrue;
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float == obj -> true or false
|
|
*
|
|
* Returns +true+ only if +obj+ has the same value as +float+.
|
|
* Contrast this with Float#eql?, which requires +obj+ to be a Float.
|
|
*
|
|
* 1.0 == 1 #=> true
|
|
*
|
|
* The result of <code>NaN == NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_equal(VALUE x, VALUE y)
|
|
{
|
|
volatile double a, b;
|
|
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_integer_float_eq(y, x);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(b)) return Qfalse;
|
|
#endif
|
|
}
|
|
else {
|
|
return num_equal(x, y);
|
|
}
|
|
a = RFLOAT_VALUE(x);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a)) return Qfalse;
|
|
#endif
|
|
return (a == b)?Qtrue:Qfalse;
|
|
}
|
|
|
|
#define flo_eq rb_float_equal
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.hash -> integer
|
|
*
|
|
* Returns a hash code for this float.
|
|
*
|
|
* See also Object#hash.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_hash(VALUE num)
|
|
{
|
|
return rb_dbl_hash(RFLOAT_VALUE(num));
|
|
}
|
|
|
|
VALUE
|
|
rb_dbl_hash(double d)
|
|
{
|
|
return LONG2FIX(rb_dbl_long_hash(d));
|
|
}
|
|
|
|
VALUE
|
|
rb_dbl_cmp(double a, double b)
|
|
{
|
|
if (isnan(a) || isnan(b)) return Qnil;
|
|
if (a == b) return INT2FIX(0);
|
|
if (a > b) return INT2FIX(1);
|
|
if (a < b) return INT2FIX(-1);
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float <=> real -> -1, 0, +1, or nil
|
|
*
|
|
* Returns -1, 0, or +1 depending on whether +float+ is
|
|
* less than, equal to, or greater than +real+.
|
|
* This is the basis for the tests in the Comparable module.
|
|
*
|
|
* The result of <code>NaN <=> NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*
|
|
* +nil+ is returned if the two values are incomparable.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_cmp(VALUE x, VALUE y)
|
|
{
|
|
double a, b;
|
|
VALUE i;
|
|
|
|
a = RFLOAT_VALUE(x);
|
|
if (isnan(a)) return Qnil;
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE rel = rb_integer_float_cmp(y, x);
|
|
if (FIXNUM_P(rel))
|
|
return INT2FIX(-FIX2INT(rel));
|
|
return rel;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
}
|
|
else {
|
|
if (isinf(a) && (i = rb_check_funcall(y, rb_intern("infinite?"), 0, 0)) != Qundef) {
|
|
if (RTEST(i)) {
|
|
int j = rb_cmpint(i, x, y);
|
|
j = (a > 0.0) ? (j > 0 ? 0 : +1) : (j < 0 ? 0 : -1);
|
|
return INT2FIX(j);
|
|
}
|
|
if (a > 0.0) return INT2FIX(1);
|
|
return INT2FIX(-1);
|
|
}
|
|
return rb_num_coerce_cmp(x, y, id_cmp);
|
|
}
|
|
return rb_dbl_cmp(a, b);
|
|
}
|
|
|
|
int
|
|
rb_float_cmp(VALUE x, VALUE y)
|
|
{
|
|
return NUM2INT(flo_cmp(x, y));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float > real -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is greater than +real+.
|
|
*
|
|
* The result of <code>NaN > NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_gt(VALUE x, VALUE y)
|
|
{
|
|
double a, b;
|
|
|
|
a = RFLOAT_VALUE(x);
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE rel = rb_integer_float_cmp(y, x);
|
|
if (FIXNUM_P(rel))
|
|
return -FIX2INT(rel) > 0 ? Qtrue : Qfalse;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(b)) return Qfalse;
|
|
#endif
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, '>');
|
|
}
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a)) return Qfalse;
|
|
#endif
|
|
return (a > b)?Qtrue:Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float >= real -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is greater than or equal to +real+.
|
|
*
|
|
* The result of <code>NaN >= NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_ge(VALUE x, VALUE y)
|
|
{
|
|
double a, b;
|
|
|
|
a = RFLOAT_VALUE(x);
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE rel = rb_integer_float_cmp(y, x);
|
|
if (FIXNUM_P(rel))
|
|
return -FIX2INT(rel) >= 0 ? Qtrue : Qfalse;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(b)) return Qfalse;
|
|
#endif
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, idGE);
|
|
}
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a)) return Qfalse;
|
|
#endif
|
|
return (a >= b)?Qtrue:Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float < real -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is less than +real+.
|
|
*
|
|
* The result of <code>NaN < NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_lt(VALUE x, VALUE y)
|
|
{
|
|
double a, b;
|
|
|
|
a = RFLOAT_VALUE(x);
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE rel = rb_integer_float_cmp(y, x);
|
|
if (FIXNUM_P(rel))
|
|
return -FIX2INT(rel) < 0 ? Qtrue : Qfalse;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(b)) return Qfalse;
|
|
#endif
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, '<');
|
|
}
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a)) return Qfalse;
|
|
#endif
|
|
return (a < b)?Qtrue:Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float <= real -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is less than or equal to +real+.
|
|
*
|
|
* The result of <code>NaN <= NaN</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_le(VALUE x, VALUE y)
|
|
{
|
|
double a, b;
|
|
|
|
a = RFLOAT_VALUE(x);
|
|
if (RB_TYPE_P(y, T_FIXNUM) || RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE rel = rb_integer_float_cmp(y, x);
|
|
if (FIXNUM_P(rel))
|
|
return -FIX2INT(rel) <= 0 ? Qtrue : Qfalse;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(b)) return Qfalse;
|
|
#endif
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, idLE);
|
|
}
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a)) return Qfalse;
|
|
#endif
|
|
return (a <= b)?Qtrue:Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.eql?(obj) -> true or false
|
|
*
|
|
* Returns +true+ only if +obj+ is a Float with the same value as +float+.
|
|
* Contrast this with Float#==, which performs type conversions.
|
|
*
|
|
* 1.0.eql?(1) #=> false
|
|
*
|
|
* The result of <code>NaN.eql?(NaN)</code> is undefined,
|
|
* so an implementation-dependent value is returned.
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_eql(VALUE x, VALUE y)
|
|
{
|
|
if (RB_TYPE_P(y, T_FLOAT)) {
|
|
double a = RFLOAT_VALUE(x);
|
|
double b = RFLOAT_VALUE(y);
|
|
#if defined(_MSC_VER) && _MSC_VER < 1300
|
|
if (isnan(a) || isnan(b)) return Qfalse;
|
|
#endif
|
|
if (a == b)
|
|
return Qtrue;
|
|
}
|
|
return Qfalse;
|
|
}
|
|
|
|
#define flo_eql rb_float_eql
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.to_f -> self
|
|
*
|
|
* Since +float+ is already a Float, returns +self+.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_to_f(VALUE num)
|
|
{
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.abs -> float
|
|
* float.magnitude -> float
|
|
*
|
|
* Returns the absolute value of +float+.
|
|
*
|
|
* (-34.56).abs #=> 34.56
|
|
* -34.56.abs #=> 34.56
|
|
* 34.56.abs #=> 34.56
|
|
*
|
|
* Float#magnitude is an alias for Float#abs.
|
|
*/
|
|
|
|
VALUE
|
|
rb_float_abs(VALUE flt)
|
|
{
|
|
double val = fabs(RFLOAT_VALUE(flt));
|
|
return DBL2NUM(val);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.zero? -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is 0.0.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_zero_p(VALUE num)
|
|
{
|
|
if (RFLOAT_VALUE(num) == 0.0) {
|
|
return Qtrue;
|
|
}
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.nan? -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is an invalid IEEE floating point number.
|
|
*
|
|
* a = -1.0 #=> -1.0
|
|
* a.nan? #=> false
|
|
* a = 0.0/0.0 #=> NaN
|
|
* a.nan? #=> true
|
|
*/
|
|
|
|
static VALUE
|
|
flo_is_nan_p(VALUE num)
|
|
{
|
|
double value = RFLOAT_VALUE(num);
|
|
|
|
return isnan(value) ? Qtrue : Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.infinite? -> -1, 1, or nil
|
|
*
|
|
* Returns +nil+, -1, or 1 depending on whether the value is
|
|
* finite, <code>-Infinity</code>, or <code>+Infinity</code>.
|
|
*
|
|
* (0.0).infinite? #=> nil
|
|
* (-1.0/0.0).infinite? #=> -1
|
|
* (+1.0/0.0).infinite? #=> 1
|
|
*/
|
|
|
|
VALUE
|
|
rb_flo_is_infinite_p(VALUE num)
|
|
{
|
|
double value = RFLOAT_VALUE(num);
|
|
|
|
if (isinf(value)) {
|
|
return INT2FIX( value < 0 ? -1 : 1 );
|
|
}
|
|
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.finite? -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is a valid IEEE floating point number,
|
|
* i.e. it is not infinite and Float#nan? is +false+.
|
|
*/
|
|
|
|
VALUE
|
|
rb_flo_is_finite_p(VALUE num)
|
|
{
|
|
double value = RFLOAT_VALUE(num);
|
|
|
|
#ifdef HAVE_ISFINITE
|
|
if (!isfinite(value))
|
|
return Qfalse;
|
|
#else
|
|
if (isinf(value) || isnan(value))
|
|
return Qfalse;
|
|
#endif
|
|
|
|
return Qtrue;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.next_float -> float
|
|
*
|
|
* Returns the next representable floating point number.
|
|
*
|
|
* Float::MAX.next_float and Float::INFINITY.next_float is Float::INFINITY.
|
|
*
|
|
* Float::NAN.next_float is Float::NAN.
|
|
*
|
|
* For example:
|
|
*
|
|
* 0.01.next_float #=> 0.010000000000000002
|
|
* 1.0.next_float #=> 1.0000000000000002
|
|
* 100.0.next_float #=> 100.00000000000001
|
|
*
|
|
* 0.01.next_float - 0.01 #=> 1.734723475976807e-18
|
|
* 1.0.next_float - 1.0 #=> 2.220446049250313e-16
|
|
* 100.0.next_float - 100.0 #=> 1.4210854715202004e-14
|
|
*
|
|
* f = 0.01; 20.times { printf "%-20a %s\n", f, f.to_s; f = f.next_float }
|
|
* #=> 0x1.47ae147ae147bp-7 0.01
|
|
* # 0x1.47ae147ae147cp-7 0.010000000000000002
|
|
* # 0x1.47ae147ae147dp-7 0.010000000000000004
|
|
* # 0x1.47ae147ae147ep-7 0.010000000000000005
|
|
* # 0x1.47ae147ae147fp-7 0.010000000000000007
|
|
* # 0x1.47ae147ae148p-7 0.010000000000000009
|
|
* # 0x1.47ae147ae1481p-7 0.01000000000000001
|
|
* # 0x1.47ae147ae1482p-7 0.010000000000000012
|
|
* # 0x1.47ae147ae1483p-7 0.010000000000000014
|
|
* # 0x1.47ae147ae1484p-7 0.010000000000000016
|
|
* # 0x1.47ae147ae1485p-7 0.010000000000000018
|
|
* # 0x1.47ae147ae1486p-7 0.01000000000000002
|
|
* # 0x1.47ae147ae1487p-7 0.010000000000000021
|
|
* # 0x1.47ae147ae1488p-7 0.010000000000000023
|
|
* # 0x1.47ae147ae1489p-7 0.010000000000000024
|
|
* # 0x1.47ae147ae148ap-7 0.010000000000000026
|
|
* # 0x1.47ae147ae148bp-7 0.010000000000000028
|
|
* # 0x1.47ae147ae148cp-7 0.01000000000000003
|
|
* # 0x1.47ae147ae148dp-7 0.010000000000000031
|
|
* # 0x1.47ae147ae148ep-7 0.010000000000000033
|
|
*
|
|
* f = 0.0
|
|
* 100.times { f += 0.1 }
|
|
* f #=> 9.99999999999998 # should be 10.0 in the ideal world.
|
|
* 10-f #=> 1.9539925233402755e-14 # the floating point error.
|
|
* 10.0.next_float-10 #=> 1.7763568394002505e-15 # 1 ulp (unit in the last place).
|
|
* (10-f)/(10.0.next_float-10) #=> 11.0 # the error is 11 ulp.
|
|
* (10-f)/(10*Float::EPSILON) #=> 8.8 # approximation of the above.
|
|
* "%a" % 10 #=> "0x1.4p+3"
|
|
* "%a" % f #=> "0x1.3fffffffffff5p+3" # the last hex digit is 5. 16 - 5 = 11 ulp.
|
|
*/
|
|
static VALUE
|
|
flo_next_float(VALUE vx)
|
|
{
|
|
double x, y;
|
|
x = NUM2DBL(vx);
|
|
y = nextafter(x, INFINITY);
|
|
return DBL2NUM(y);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.prev_float -> float
|
|
*
|
|
* Returns the previous representable floating point number.
|
|
*
|
|
* (-Float::MAX).prev_float and (-Float::INFINITY).prev_float is -Float::INFINITY.
|
|
*
|
|
* Float::NAN.prev_float is Float::NAN.
|
|
*
|
|
* For example:
|
|
*
|
|
* 0.01.prev_float #=> 0.009999999999999998
|
|
* 1.0.prev_float #=> 0.9999999999999999
|
|
* 100.0.prev_float #=> 99.99999999999999
|
|
*
|
|
* 0.01 - 0.01.prev_float #=> 1.734723475976807e-18
|
|
* 1.0 - 1.0.prev_float #=> 1.1102230246251565e-16
|
|
* 100.0 - 100.0.prev_float #=> 1.4210854715202004e-14
|
|
*
|
|
* f = 0.01; 20.times { printf "%-20a %s\n", f, f.to_s; f = f.prev_float }
|
|
* #=> 0x1.47ae147ae147bp-7 0.01
|
|
* # 0x1.47ae147ae147ap-7 0.009999999999999998
|
|
* # 0x1.47ae147ae1479p-7 0.009999999999999997
|
|
* # 0x1.47ae147ae1478p-7 0.009999999999999995
|
|
* # 0x1.47ae147ae1477p-7 0.009999999999999993
|
|
* # 0x1.47ae147ae1476p-7 0.009999999999999992
|
|
* # 0x1.47ae147ae1475p-7 0.00999999999999999
|
|
* # 0x1.47ae147ae1474p-7 0.009999999999999988
|
|
* # 0x1.47ae147ae1473p-7 0.009999999999999986
|
|
* # 0x1.47ae147ae1472p-7 0.009999999999999985
|
|
* # 0x1.47ae147ae1471p-7 0.009999999999999983
|
|
* # 0x1.47ae147ae147p-7 0.009999999999999981
|
|
* # 0x1.47ae147ae146fp-7 0.00999999999999998
|
|
* # 0x1.47ae147ae146ep-7 0.009999999999999978
|
|
* # 0x1.47ae147ae146dp-7 0.009999999999999976
|
|
* # 0x1.47ae147ae146cp-7 0.009999999999999974
|
|
* # 0x1.47ae147ae146bp-7 0.009999999999999972
|
|
* # 0x1.47ae147ae146ap-7 0.00999999999999997
|
|
* # 0x1.47ae147ae1469p-7 0.009999999999999969
|
|
* # 0x1.47ae147ae1468p-7 0.009999999999999967
|
|
*/
|
|
static VALUE
|
|
flo_prev_float(VALUE vx)
|
|
{
|
|
double x, y;
|
|
x = NUM2DBL(vx);
|
|
y = nextafter(x, -INFINITY);
|
|
return DBL2NUM(y);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.floor([ndigits]) -> integer or float
|
|
*
|
|
* Returns the largest number less than or equal to +float+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns a floating point number when +ndigits+ is positive,
|
|
* otherwise returns an integer.
|
|
*
|
|
* 1.2.floor #=> 1
|
|
* 2.0.floor #=> 2
|
|
* (-1.2).floor #=> -2
|
|
* (-2.0).floor #=> -2
|
|
*
|
|
* 1.234567.floor(2) #=> 1.23
|
|
* 1.234567.floor(3) #=> 1.234
|
|
* 1.234567.floor(4) #=> 1.2345
|
|
* 1.234567.floor(5) #=> 1.23456
|
|
*
|
|
* 34567.89.floor(-5) #=> 0
|
|
* 34567.89.floor(-4) #=> 30000
|
|
* 34567.89.floor(-3) #=> 34000
|
|
* 34567.89.floor(-2) #=> 34500
|
|
* 34567.89.floor(-1) #=> 34560
|
|
* 34567.89.floor(0) #=> 34567
|
|
* 34567.89.floor(1) #=> 34567.8
|
|
* 34567.89.floor(2) #=> 34567.89
|
|
* 34567.89.floor(3) #=> 34567.89
|
|
*
|
|
* Note that the limited precision of floating point arithmetic
|
|
* might lead to surprising results:
|
|
*
|
|
* (0.3 / 0.1).floor #=> 2 (!)
|
|
*/
|
|
|
|
static VALUE
|
|
flo_floor(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
double number, f;
|
|
int ndigits = 0;
|
|
|
|
if (rb_check_arity(argc, 0, 1)) {
|
|
ndigits = NUM2INT(argv[0]);
|
|
}
|
|
number = RFLOAT_VALUE(num);
|
|
if (number == 0.0) {
|
|
return ndigits > 0 ? DBL2NUM(number) : INT2FIX(0);
|
|
}
|
|
if (ndigits > 0) {
|
|
int binexp;
|
|
frexp(number, &binexp);
|
|
if (float_round_overflow(ndigits, binexp)) return num;
|
|
if (number > 0.0 && float_round_underflow(ndigits, binexp))
|
|
return DBL2NUM(0.0);
|
|
f = pow(10, ndigits);
|
|
f = floor(number * f) / f;
|
|
return DBL2NUM(f);
|
|
}
|
|
else {
|
|
num = dbl2ival(floor(number));
|
|
if (ndigits < 0) num = rb_int_floor(num, ndigits);
|
|
return num;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.ceil([ndigits]) -> integer or float
|
|
*
|
|
* Returns the smallest number greater than or equal to +float+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns a floating point number when +ndigits+ is positive,
|
|
* otherwise returns an integer.
|
|
*
|
|
* 1.2.ceil #=> 2
|
|
* 2.0.ceil #=> 2
|
|
* (-1.2).ceil #=> -1
|
|
* (-2.0).ceil #=> -2
|
|
*
|
|
* 1.234567.ceil(2) #=> 1.24
|
|
* 1.234567.ceil(3) #=> 1.235
|
|
* 1.234567.ceil(4) #=> 1.2346
|
|
* 1.234567.ceil(5) #=> 1.23457
|
|
*
|
|
* 34567.89.ceil(-5) #=> 100000
|
|
* 34567.89.ceil(-4) #=> 40000
|
|
* 34567.89.ceil(-3) #=> 35000
|
|
* 34567.89.ceil(-2) #=> 34600
|
|
* 34567.89.ceil(-1) #=> 34570
|
|
* 34567.89.ceil(0) #=> 34568
|
|
* 34567.89.ceil(1) #=> 34567.9
|
|
* 34567.89.ceil(2) #=> 34567.89
|
|
* 34567.89.ceil(3) #=> 34567.89
|
|
*
|
|
* Note that the limited precision of floating point arithmetic
|
|
* might lead to surprising results:
|
|
*
|
|
* (2.1 / 0.7).ceil #=> 4 (!)
|
|
*/
|
|
|
|
static VALUE
|
|
flo_ceil(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
double number, f;
|
|
int ndigits = 0;
|
|
|
|
if (rb_check_arity(argc, 0, 1)) {
|
|
ndigits = NUM2INT(argv[0]);
|
|
}
|
|
number = RFLOAT_VALUE(num);
|
|
if (number == 0.0) {
|
|
return ndigits > 0 ? DBL2NUM(number) : INT2FIX(0);
|
|
}
|
|
if (ndigits > 0) {
|
|
int binexp;
|
|
frexp(number, &binexp);
|
|
if (float_round_overflow(ndigits, binexp)) return num;
|
|
if (number < 0.0 && float_round_underflow(ndigits, binexp))
|
|
return DBL2NUM(0.0);
|
|
f = pow(10, ndigits);
|
|
f = ceil(number * f) / f;
|
|
return DBL2NUM(f);
|
|
}
|
|
else {
|
|
num = dbl2ival(ceil(number));
|
|
if (ndigits < 0) num = rb_int_ceil(num, ndigits);
|
|
return num;
|
|
}
|
|
}
|
|
|
|
static int
|
|
int_round_zero_p(VALUE num, int ndigits)
|
|
{
|
|
long bytes;
|
|
/* If 10**N / 2 > num, then return 0 */
|
|
/* We have log_256(10) > 0.415241 and log_256(1/2) = -0.125, so */
|
|
if (FIXNUM_P(num)) {
|
|
bytes = sizeof(long);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
bytes = rb_big_size(num);
|
|
}
|
|
else {
|
|
bytes = NUM2LONG(rb_funcall(num, idSize, 0));
|
|
}
|
|
return (-0.415241 * ndigits - 0.125 > bytes);
|
|
}
|
|
|
|
static SIGNED_VALUE
|
|
int_round_half_even(SIGNED_VALUE x, SIGNED_VALUE y)
|
|
{
|
|
SIGNED_VALUE z = +(x + y / 2) / y;
|
|
if ((z * y - x) * 2 == y) {
|
|
z &= ~1;
|
|
}
|
|
return z * y;
|
|
}
|
|
|
|
static SIGNED_VALUE
|
|
int_round_half_up(SIGNED_VALUE x, SIGNED_VALUE y)
|
|
{
|
|
return (x + y / 2) / y * y;
|
|
}
|
|
|
|
static SIGNED_VALUE
|
|
int_round_half_down(SIGNED_VALUE x, SIGNED_VALUE y)
|
|
{
|
|
return (x + y / 2 - 1) / y * y;
|
|
}
|
|
|
|
static int
|
|
int_half_p_half_even(VALUE num, VALUE n, VALUE f)
|
|
{
|
|
return (int)rb_int_odd_p(rb_int_idiv(n, f));
|
|
}
|
|
|
|
static int
|
|
int_half_p_half_up(VALUE num, VALUE n, VALUE f)
|
|
{
|
|
return int_pos_p(num);
|
|
}
|
|
|
|
static int
|
|
int_half_p_half_down(VALUE num, VALUE n, VALUE f)
|
|
{
|
|
return int_neg_p(num);
|
|
}
|
|
|
|
/*
|
|
* Assumes num is an Integer, ndigits <= 0
|
|
*/
|
|
VALUE
|
|
rb_int_round(VALUE num, int ndigits, enum ruby_num_rounding_mode mode)
|
|
{
|
|
VALUE n, f, h, r;
|
|
|
|
if (int_round_zero_p(num, ndigits)) {
|
|
return INT2FIX(0);
|
|
}
|
|
|
|
f = int_pow(10, -ndigits);
|
|
if (FIXNUM_P(num) && FIXNUM_P(f)) {
|
|
SIGNED_VALUE x = FIX2LONG(num), y = FIX2LONG(f);
|
|
int neg = x < 0;
|
|
if (neg) x = -x;
|
|
x = ROUND_CALL(mode, int_round, (x, y));
|
|
if (neg) x = -x;
|
|
return LONG2NUM(x);
|
|
}
|
|
if (RB_TYPE_P(f, T_FLOAT)) {
|
|
/* then int_pow overflow */
|
|
return INT2FIX(0);
|
|
}
|
|
h = rb_int_idiv(f, INT2FIX(2));
|
|
r = rb_int_modulo(num, f);
|
|
n = rb_int_minus(num, r);
|
|
r = rb_int_cmp(r, h);
|
|
if (FIXNUM_POSITIVE_P(r) ||
|
|
(FIXNUM_ZERO_P(r) && ROUND_CALL(mode, int_half_p, (num, n, f)))) {
|
|
n = rb_int_plus(n, f);
|
|
}
|
|
return n;
|
|
}
|
|
|
|
VALUE
|
|
rb_int_floor(VALUE num, int ndigits)
|
|
{
|
|
VALUE f;
|
|
|
|
if (int_round_zero_p(num, ndigits))
|
|
return INT2FIX(0);
|
|
f = int_pow(10, -ndigits);
|
|
if (FIXNUM_P(num) && FIXNUM_P(f)) {
|
|
SIGNED_VALUE x = FIX2LONG(num), y = FIX2LONG(f);
|
|
int neg = x < 0;
|
|
if (neg) x = -x + y - 1;
|
|
x = x / y * y;
|
|
if (neg) x = -x;
|
|
return LONG2NUM(x);
|
|
}
|
|
if (RB_TYPE_P(f, T_FLOAT)) {
|
|
/* then int_pow overflow */
|
|
return INT2FIX(0);
|
|
}
|
|
return rb_int_minus(num, rb_int_modulo(num, f));
|
|
}
|
|
|
|
VALUE
|
|
rb_int_ceil(VALUE num, int ndigits)
|
|
{
|
|
VALUE f;
|
|
|
|
if (int_round_zero_p(num, ndigits))
|
|
return INT2FIX(0);
|
|
f = int_pow(10, -ndigits);
|
|
if (FIXNUM_P(num) && FIXNUM_P(f)) {
|
|
SIGNED_VALUE x = FIX2LONG(num), y = FIX2LONG(f);
|
|
int neg = x < 0;
|
|
if (neg) x = -x;
|
|
else x += y - 1;
|
|
x = (x / y) * y;
|
|
if (neg) x = -x;
|
|
return LONG2NUM(x);
|
|
}
|
|
if (RB_TYPE_P(f, T_FLOAT)) {
|
|
/* then int_pow overflow */
|
|
return INT2FIX(0);
|
|
}
|
|
return rb_int_plus(num, rb_int_minus(f, rb_int_modulo(num, f)));
|
|
}
|
|
|
|
VALUE
|
|
rb_int_truncate(VALUE num, int ndigits)
|
|
{
|
|
VALUE f;
|
|
VALUE m;
|
|
|
|
if (int_round_zero_p(num, ndigits))
|
|
return INT2FIX(0);
|
|
f = int_pow(10, -ndigits);
|
|
if (FIXNUM_P(num) && FIXNUM_P(f)) {
|
|
SIGNED_VALUE x = FIX2LONG(num), y = FIX2LONG(f);
|
|
int neg = x < 0;
|
|
if (neg) x = -x;
|
|
x = x / y * y;
|
|
if (neg) x = -x;
|
|
return LONG2NUM(x);
|
|
}
|
|
if (RB_TYPE_P(f, T_FLOAT)) {
|
|
/* then int_pow overflow */
|
|
return INT2FIX(0);
|
|
}
|
|
m = rb_int_modulo(num, f);
|
|
if (int_neg_p(num)) {
|
|
return rb_int_plus(num, rb_int_minus(f, m));
|
|
}
|
|
else {
|
|
return rb_int_minus(num, m);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.round([ndigits] [, half: mode]) -> integer or float
|
|
*
|
|
* Returns +float+ rounded to the nearest value with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns a floating point number when +ndigits+ is positive,
|
|
* otherwise returns an integer.
|
|
*
|
|
* 1.4.round #=> 1
|
|
* 1.5.round #=> 2
|
|
* 1.6.round #=> 2
|
|
* (-1.5).round #=> -2
|
|
*
|
|
* 1.234567.round(2) #=> 1.23
|
|
* 1.234567.round(3) #=> 1.235
|
|
* 1.234567.round(4) #=> 1.2346
|
|
* 1.234567.round(5) #=> 1.23457
|
|
*
|
|
* 34567.89.round(-5) #=> 0
|
|
* 34567.89.round(-4) #=> 30000
|
|
* 34567.89.round(-3) #=> 35000
|
|
* 34567.89.round(-2) #=> 34600
|
|
* 34567.89.round(-1) #=> 34570
|
|
* 34567.89.round(0) #=> 34568
|
|
* 34567.89.round(1) #=> 34567.9
|
|
* 34567.89.round(2) #=> 34567.89
|
|
* 34567.89.round(3) #=> 34567.89
|
|
*
|
|
* If the optional +half+ keyword argument is given,
|
|
* numbers that are half-way between two possible rounded values
|
|
* will be rounded according to the specified tie-breaking +mode+:
|
|
*
|
|
* * <code>:up</code> or +nil+: round half away from zero (default)
|
|
* * <code>:down</code>: round half toward zero
|
|
* * <code>:even</code>: round half toward the nearest even number
|
|
*
|
|
* 2.5.round(half: :up) #=> 3
|
|
* 2.5.round(half: :down) #=> 2
|
|
* 2.5.round(half: :even) #=> 2
|
|
* 3.5.round(half: :up) #=> 4
|
|
* 3.5.round(half: :down) #=> 3
|
|
* 3.5.round(half: :even) #=> 4
|
|
* (-2.5).round(half: :up) #=> -3
|
|
* (-2.5).round(half: :down) #=> -2
|
|
* (-2.5).round(half: :even) #=> -2
|
|
*/
|
|
|
|
static VALUE
|
|
flo_round(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
double number, f, x;
|
|
VALUE nd, opt;
|
|
int ndigits = 0;
|
|
enum ruby_num_rounding_mode mode;
|
|
|
|
if (rb_scan_args(argc, argv, "01:", &nd, &opt)) {
|
|
ndigits = NUM2INT(nd);
|
|
}
|
|
mode = rb_num_get_rounding_option(opt);
|
|
number = RFLOAT_VALUE(num);
|
|
if (number == 0.0) {
|
|
return ndigits > 0 ? DBL2NUM(number) : INT2FIX(0);
|
|
}
|
|
if (ndigits < 0) {
|
|
return rb_int_round(flo_to_i(num), ndigits, mode);
|
|
}
|
|
if (ndigits == 0) {
|
|
x = ROUND_CALL(mode, round, (number, 1.0));
|
|
return dbl2ival(x);
|
|
}
|
|
if (isfinite(number)) {
|
|
int binexp;
|
|
frexp(number, &binexp);
|
|
if (float_round_overflow(ndigits, binexp)) return num;
|
|
if (float_round_underflow(ndigits, binexp)) return DBL2NUM(0);
|
|
f = pow(10, ndigits);
|
|
x = ROUND_CALL(mode, round, (number, f));
|
|
return DBL2NUM(x / f);
|
|
}
|
|
return num;
|
|
}
|
|
|
|
static int
|
|
float_round_overflow(int ndigits, int binexp)
|
|
{
|
|
enum {float_dig = DBL_DIG+2};
|
|
|
|
/* Let `exp` be such that `number` is written as:"0.#{digits}e#{exp}",
|
|
i.e. such that 10 ** (exp - 1) <= |number| < 10 ** exp
|
|
Recall that up to float_dig digits can be needed to represent a double,
|
|
so if ndigits + exp >= float_dig, the intermediate value (number * 10 ** ndigits)
|
|
will be an integer and thus the result is the original number.
|
|
If ndigits + exp <= 0, the result is 0 or "1e#{exp}", so
|
|
if ndigits + exp < 0, the result is 0.
|
|
We have:
|
|
2 ** (binexp-1) <= |number| < 2 ** binexp
|
|
10 ** ((binexp-1)/log_2(10)) <= |number| < 10 ** (binexp/log_2(10))
|
|
If binexp >= 0, and since log_2(10) = 3.322259:
|
|
10 ** (binexp/4 - 1) < |number| < 10 ** (binexp/3)
|
|
floor(binexp/4) <= exp <= ceil(binexp/3)
|
|
If binexp <= 0, swap the /4 and the /3
|
|
So if ndigits + floor(binexp/(4 or 3)) >= float_dig, the result is number
|
|
If ndigits + ceil(binexp/(3 or 4)) < 0 the result is 0
|
|
*/
|
|
if (ndigits >= float_dig - (binexp > 0 ? binexp / 4 : binexp / 3 - 1)) {
|
|
return TRUE;
|
|
}
|
|
return FALSE;
|
|
}
|
|
|
|
static int
|
|
float_round_underflow(int ndigits, int binexp)
|
|
{
|
|
if (ndigits < - (binexp > 0 ? binexp / 3 + 1 : binexp / 4)) {
|
|
return TRUE;
|
|
}
|
|
return FALSE;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.to_i -> integer
|
|
* float.to_int -> integer
|
|
*
|
|
* Returns the +float+ truncated to an Integer.
|
|
*
|
|
* 1.2.to_i #=> 1
|
|
* (-1.2).to_i #=> -1
|
|
*
|
|
* Note that the limited precision of floating point arithmetic
|
|
* might lead to surprising results:
|
|
*
|
|
* (0.3 / 0.1).to_i #=> 2 (!)
|
|
*
|
|
* #to_int is an alias for #to_i.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_to_i(VALUE num)
|
|
{
|
|
double f = RFLOAT_VALUE(num);
|
|
|
|
if (f > 0.0) f = floor(f);
|
|
if (f < 0.0) f = ceil(f);
|
|
|
|
return dbl2ival(f);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.truncate([ndigits]) -> integer or float
|
|
*
|
|
* Returns +float+ truncated (toward zero) to
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns a floating point number when +ndigits+ is positive,
|
|
* otherwise returns an integer.
|
|
*
|
|
* 2.8.truncate #=> 2
|
|
* (-2.8).truncate #=> -2
|
|
* 1.234567.truncate(2) #=> 1.23
|
|
* 34567.89.truncate(-2) #=> 34500
|
|
*
|
|
* Note that the limited precision of floating point arithmetic
|
|
* might lead to surprising results:
|
|
*
|
|
* (0.3 / 0.1).truncate #=> 2 (!)
|
|
*/
|
|
static VALUE
|
|
flo_truncate(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
if (signbit(RFLOAT_VALUE(num)))
|
|
return flo_ceil(argc, argv, num);
|
|
else
|
|
return flo_floor(argc, argv, num);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.positive? -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is greater than 0.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_positive_p(VALUE num)
|
|
{
|
|
double f = RFLOAT_VALUE(num);
|
|
return f > 0.0 ? Qtrue : Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* float.negative? -> true or false
|
|
*
|
|
* Returns +true+ if +float+ is less than 0.
|
|
*/
|
|
|
|
static VALUE
|
|
flo_negative_p(VALUE num)
|
|
{
|
|
double f = RFLOAT_VALUE(num);
|
|
return f < 0.0 ? Qtrue : Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.floor([ndigits]) -> integer or float
|
|
*
|
|
* Returns the largest number less than or equal to +num+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* Numeric implements this by converting its value to a Float and
|
|
* invoking Float#floor.
|
|
*/
|
|
|
|
static VALUE
|
|
num_floor(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
return flo_floor(argc, argv, rb_Float(num));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.ceil([ndigits]) -> integer or float
|
|
*
|
|
* Returns the smallest number greater than or equal to +num+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* Numeric implements this by converting its value to a Float and
|
|
* invoking Float#ceil.
|
|
*/
|
|
|
|
static VALUE
|
|
num_ceil(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
return flo_ceil(argc, argv, rb_Float(num));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.round([ndigits]) -> integer or float
|
|
*
|
|
* Returns +num+ rounded to the nearest value with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* Numeric implements this by converting its value to a Float and
|
|
* invoking Float#round.
|
|
*/
|
|
|
|
static VALUE
|
|
num_round(int argc, VALUE* argv, VALUE num)
|
|
{
|
|
return flo_round(argc, argv, rb_Float(num));
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.truncate([ndigits]) -> integer or float
|
|
*
|
|
* Returns +num+ truncated (toward zero) to
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* Numeric implements this by converting its value to a Float and
|
|
* invoking Float#truncate.
|
|
*/
|
|
|
|
static VALUE
|
|
num_truncate(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
return flo_truncate(argc, argv, rb_Float(num));
|
|
}
|
|
|
|
static double
|
|
ruby_float_step_size(double beg, double end, double unit, int excl)
|
|
{
|
|
const double epsilon = DBL_EPSILON;
|
|
double n = (end - beg)/unit;
|
|
double err = (fabs(beg) + fabs(end) + fabs(end-beg)) / fabs(unit) * epsilon;
|
|
|
|
if (isinf(unit)) {
|
|
return unit > 0 ? beg <= end : beg >= end;
|
|
}
|
|
if (unit == 0) {
|
|
return INFINITY;
|
|
}
|
|
if (err>0.5) err=0.5;
|
|
if (excl) {
|
|
if (n<=0) return 0;
|
|
if (n<1)
|
|
n = 0;
|
|
else
|
|
n = floor(n - err);
|
|
}
|
|
else {
|
|
if (n<0) return 0;
|
|
n = floor(n + err);
|
|
}
|
|
return n+1;
|
|
}
|
|
|
|
int
|
|
ruby_float_step(VALUE from, VALUE to, VALUE step, int excl)
|
|
{
|
|
if (RB_TYPE_P(from, T_FLOAT) || RB_TYPE_P(to, T_FLOAT) || RB_TYPE_P(step, T_FLOAT)) {
|
|
double beg = NUM2DBL(from);
|
|
double end = NUM2DBL(to);
|
|
double unit = NUM2DBL(step);
|
|
double n = ruby_float_step_size(beg, end, unit, excl);
|
|
long i;
|
|
|
|
if (isinf(unit)) {
|
|
/* if unit is infinity, i*unit+beg is NaN */
|
|
if (n) rb_yield(DBL2NUM(beg));
|
|
}
|
|
else if (unit == 0) {
|
|
VALUE val = DBL2NUM(beg);
|
|
for (;;)
|
|
rb_yield(val);
|
|
}
|
|
else {
|
|
for (i=0; i<n; i++) {
|
|
double d = i*unit+beg;
|
|
if (unit >= 0 ? end < d : d < end) d = end;
|
|
rb_yield(DBL2NUM(d));
|
|
}
|
|
}
|
|
return TRUE;
|
|
}
|
|
return FALSE;
|
|
}
|
|
|
|
VALUE
|
|
ruby_num_interval_step_size(VALUE from, VALUE to, VALUE step, int excl)
|
|
{
|
|
if (FIXNUM_P(from) && FIXNUM_P(to) && FIXNUM_P(step)) {
|
|
long delta, diff;
|
|
|
|
diff = FIX2LONG(step);
|
|
if (diff == 0) {
|
|
return DBL2NUM(INFINITY);
|
|
}
|
|
delta = FIX2LONG(to) - FIX2LONG(from);
|
|
if (diff < 0) {
|
|
diff = -diff;
|
|
delta = -delta;
|
|
}
|
|
if (excl) {
|
|
delta--;
|
|
}
|
|
if (delta < 0) {
|
|
return INT2FIX(0);
|
|
}
|
|
return ULONG2NUM(delta / diff + 1UL);
|
|
}
|
|
else if (RB_TYPE_P(from, T_FLOAT) || RB_TYPE_P(to, T_FLOAT) || RB_TYPE_P(step, T_FLOAT)) {
|
|
double n = ruby_float_step_size(NUM2DBL(from), NUM2DBL(to), NUM2DBL(step), excl);
|
|
|
|
if (isinf(n)) return DBL2NUM(n);
|
|
if (POSFIXABLE(n)) return LONG2FIX(n);
|
|
return rb_dbl2big(n);
|
|
}
|
|
else {
|
|
VALUE result;
|
|
ID cmp = '>';
|
|
switch (rb_cmpint(rb_num_coerce_cmp(step, INT2FIX(0), id_cmp), step, INT2FIX(0))) {
|
|
case 0: return DBL2NUM(INFINITY);
|
|
case -1: cmp = '<'; break;
|
|
}
|
|
if (RTEST(rb_funcall(from, cmp, 1, to))) return INT2FIX(0);
|
|
result = rb_funcall(rb_funcall(to, '-', 1, from), id_div, 1, step);
|
|
if (!excl || RTEST(rb_funcall(rb_funcall(from, '+', 1, rb_funcall(result, '*', 1, step)), cmp, 1, to))) {
|
|
result = rb_funcall(result, '+', 1, INT2FIX(1));
|
|
}
|
|
return result;
|
|
}
|
|
}
|
|
|
|
static int
|
|
num_step_negative_p(VALUE num)
|
|
{
|
|
const ID mid = '<';
|
|
VALUE zero = INT2FIX(0);
|
|
VALUE r;
|
|
|
|
if (FIXNUM_P(num)) {
|
|
if (method_basic_p(rb_cInteger))
|
|
return (SIGNED_VALUE)num < 0;
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
if (method_basic_p(rb_cInteger))
|
|
return BIGNUM_NEGATIVE_P(num);
|
|
}
|
|
|
|
r = rb_check_funcall(num, '>', 1, &zero);
|
|
if (r == Qundef) {
|
|
coerce_failed(num, INT2FIX(0));
|
|
}
|
|
return !RTEST(r);
|
|
}
|
|
|
|
static int
|
|
num_step_scan_args(int argc, const VALUE *argv, VALUE *to, VALUE *step)
|
|
{
|
|
VALUE hash;
|
|
int desc;
|
|
|
|
argc = rb_scan_args(argc, argv, "02:", to, step, &hash);
|
|
if (!NIL_P(hash)) {
|
|
ID keys[2];
|
|
VALUE values[2];
|
|
keys[0] = id_to;
|
|
keys[1] = id_by;
|
|
rb_get_kwargs(hash, keys, 0, 2, values);
|
|
if (values[0] != Qundef) {
|
|
if (argc > 0) rb_raise(rb_eArgError, "to is given twice");
|
|
*to = values[0];
|
|
}
|
|
if (values[1] != Qundef) {
|
|
if (argc > 1) rb_raise(rb_eArgError, "step is given twice");
|
|
*step = values[1];
|
|
}
|
|
}
|
|
else {
|
|
/* compatibility */
|
|
if (argc > 1 && NIL_P(*step)) {
|
|
rb_raise(rb_eTypeError, "step must be numeric");
|
|
}
|
|
if (rb_equal(*step, INT2FIX(0))) {
|
|
rb_raise(rb_eArgError, "step can't be 0");
|
|
}
|
|
}
|
|
if (NIL_P(*step)) {
|
|
*step = INT2FIX(1);
|
|
}
|
|
desc = num_step_negative_p(*step);
|
|
if (NIL_P(*to)) {
|
|
*to = desc ? DBL2NUM(-INFINITY) : DBL2NUM(INFINITY);
|
|
}
|
|
return desc;
|
|
}
|
|
|
|
static VALUE
|
|
num_step_size(VALUE from, VALUE args, VALUE eobj)
|
|
{
|
|
VALUE to, step;
|
|
int argc = args ? RARRAY_LENINT(args) : 0;
|
|
const VALUE *argv = args ? RARRAY_CONST_PTR(args) : 0;
|
|
|
|
num_step_scan_args(argc, argv, &to, &step);
|
|
|
|
return ruby_num_interval_step_size(from, to, step, FALSE);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* num.step(by: step, to: limit) {|i| block } -> self
|
|
* num.step(by: step, to: limit) -> an_enumerator
|
|
* num.step(limit=nil, step=1) {|i| block } -> self
|
|
* num.step(limit=nil, step=1) -> an_enumerator
|
|
*
|
|
* Invokes the given block with the sequence of numbers starting at +num+,
|
|
* incremented by +step+ (defaulted to +1+) on each call.
|
|
*
|
|
* The loop finishes when the value to be passed to the block is greater than
|
|
* +limit+ (if +step+ is positive) or less than +limit+ (if +step+ is
|
|
* negative), where +limit+ is defaulted to infinity.
|
|
*
|
|
* In the recommended keyword argument style, either or both of
|
|
* +step+ and +limit+ (default infinity) can be omitted. In the
|
|
* fixed position argument style, zero as a step
|
|
* (i.e. <code>num.step(limit, 0)</code>) is not allowed for historical
|
|
* compatibility reasons.
|
|
*
|
|
* If all the arguments are integers, the loop operates using an integer
|
|
* counter.
|
|
*
|
|
* If any of the arguments are floating point numbers, all are converted
|
|
* to floats, and the loop is executed
|
|
* <i>floor(n + n*Float::EPSILON) + 1</i> times,
|
|
* where <i>n = (limit - num)/step</i>.
|
|
*
|
|
* Otherwise, the loop starts at +num+, uses either the
|
|
* less-than (<code><</code>) or greater-than (<code>></code>) operator
|
|
* to compare the counter against +limit+,
|
|
* and increments itself using the <code>+</code> operator.
|
|
*
|
|
* If no block is given, an Enumerator is returned instead.
|
|
*
|
|
* For example:
|
|
*
|
|
* p 1.step.take(4)
|
|
* p 10.step(by: -1).take(4)
|
|
* 3.step(to: 5) {|i| print i, " " }
|
|
* 1.step(10, 2) {|i| print i, " " }
|
|
* Math::E.step(to: Math::PI, by: 0.2) {|f| print f, " " }
|
|
*
|
|
* Will produce:
|
|
*
|
|
* [1, 2, 3, 4]
|
|
* [10, 9, 8, 7]
|
|
* 3 4 5
|
|
* 1 3 5 7 9
|
|
* 2.718281828459045 2.9182818284590453 3.118281828459045
|
|
*/
|
|
|
|
static VALUE
|
|
num_step(int argc, VALUE *argv, VALUE from)
|
|
{
|
|
VALUE to, step;
|
|
int desc, inf;
|
|
|
|
RETURN_SIZED_ENUMERATOR(from, argc, argv, num_step_size);
|
|
|
|
desc = num_step_scan_args(argc, argv, &to, &step);
|
|
if (rb_equal(step, INT2FIX(0))) {
|
|
inf = 1;
|
|
}
|
|
else if (RB_TYPE_P(to, T_FLOAT)) {
|
|
double f = RFLOAT_VALUE(to);
|
|
inf = isinf(f) && (signbit(f) ? desc : !desc);
|
|
}
|
|
else inf = 0;
|
|
|
|
if (FIXNUM_P(from) && (inf || FIXNUM_P(to)) && FIXNUM_P(step)) {
|
|
long i = FIX2LONG(from);
|
|
long diff = FIX2LONG(step);
|
|
|
|
if (inf) {
|
|
for (;; i += diff)
|
|
rb_yield(LONG2FIX(i));
|
|
}
|
|
else {
|
|
long end = FIX2LONG(to);
|
|
|
|
if (desc) {
|
|
for (; i >= end; i += diff)
|
|
rb_yield(LONG2FIX(i));
|
|
}
|
|
else {
|
|
for (; i <= end; i += diff)
|
|
rb_yield(LONG2FIX(i));
|
|
}
|
|
}
|
|
}
|
|
else if (!ruby_float_step(from, to, step, FALSE)) {
|
|
VALUE i = from;
|
|
|
|
if (inf) {
|
|
for (;; i = rb_funcall(i, '+', 1, step))
|
|
rb_yield(i);
|
|
}
|
|
else {
|
|
ID cmp = desc ? '<' : '>';
|
|
|
|
for (; !RTEST(rb_funcall(i, cmp, 1, to)); i = rb_funcall(i, '+', 1, step))
|
|
rb_yield(i);
|
|
}
|
|
}
|
|
return from;
|
|
}
|
|
|
|
static char *
|
|
out_of_range_float(char (*pbuf)[24], VALUE val)
|
|
{
|
|
char *const buf = *pbuf;
|
|
char *s;
|
|
|
|
snprintf(buf, sizeof(*pbuf), "%-.10g", RFLOAT_VALUE(val));
|
|
if ((s = strchr(buf, ' ')) != 0) *s = '\0';
|
|
return buf;
|
|
}
|
|
|
|
#define FLOAT_OUT_OF_RANGE(val, type) do { \
|
|
char buf[24]; \
|
|
rb_raise(rb_eRangeError, "float %s out of range of "type, \
|
|
out_of_range_float(&buf, (val))); \
|
|
} while (0)
|
|
|
|
#define LONG_MIN_MINUS_ONE ((double)LONG_MIN-1)
|
|
#define LONG_MAX_PLUS_ONE (2*(double)(LONG_MAX/2+1))
|
|
#define ULONG_MAX_PLUS_ONE (2*(double)(ULONG_MAX/2+1))
|
|
#define LONG_MIN_MINUS_ONE_IS_LESS_THAN(n) \
|
|
(LONG_MIN_MINUS_ONE == (double)LONG_MIN ? \
|
|
LONG_MIN <= (n): \
|
|
LONG_MIN_MINUS_ONE < (n))
|
|
|
|
long
|
|
rb_num2long(VALUE val)
|
|
{
|
|
again:
|
|
if (NIL_P(val)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from nil to integer");
|
|
}
|
|
|
|
if (FIXNUM_P(val)) return FIX2LONG(val);
|
|
|
|
else if (RB_TYPE_P(val, T_FLOAT)) {
|
|
if (RFLOAT_VALUE(val) < LONG_MAX_PLUS_ONE
|
|
&& LONG_MIN_MINUS_ONE_IS_LESS_THAN(RFLOAT_VALUE(val))) {
|
|
return (long)RFLOAT_VALUE(val);
|
|
}
|
|
else {
|
|
FLOAT_OUT_OF_RANGE(val, "integer");
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(val, T_BIGNUM)) {
|
|
return rb_big2long(val);
|
|
}
|
|
else {
|
|
val = rb_to_int(val);
|
|
goto again;
|
|
}
|
|
}
|
|
|
|
static unsigned long
|
|
rb_num2ulong_internal(VALUE val, int *wrap_p)
|
|
{
|
|
again:
|
|
if (NIL_P(val)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from nil to integer");
|
|
}
|
|
|
|
if (FIXNUM_P(val)) {
|
|
long l = FIX2LONG(val); /* this is FIX2LONG, intended */
|
|
if (wrap_p)
|
|
*wrap_p = l < 0;
|
|
return (unsigned long)l;
|
|
}
|
|
else if (RB_TYPE_P(val, T_FLOAT)) {
|
|
double d = RFLOAT_VALUE(val);
|
|
if (d < ULONG_MAX_PLUS_ONE && LONG_MIN_MINUS_ONE_IS_LESS_THAN(d)) {
|
|
if (wrap_p)
|
|
*wrap_p = d <= -1.0; /* NUM2ULONG(v) uses v.to_int conceptually. */
|
|
if (0 <= d)
|
|
return (unsigned long)d;
|
|
return (unsigned long)(long)d;
|
|
}
|
|
else {
|
|
FLOAT_OUT_OF_RANGE(val, "integer");
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(val, T_BIGNUM)) {
|
|
{
|
|
unsigned long ul = rb_big2ulong(val);
|
|
if (wrap_p)
|
|
*wrap_p = BIGNUM_NEGATIVE_P(val);
|
|
return ul;
|
|
}
|
|
}
|
|
else {
|
|
val = rb_to_int(val);
|
|
goto again;
|
|
}
|
|
}
|
|
|
|
unsigned long
|
|
rb_num2ulong(VALUE val)
|
|
{
|
|
return rb_num2ulong_internal(val, NULL);
|
|
}
|
|
|
|
#if SIZEOF_INT < SIZEOF_LONG
|
|
void
|
|
rb_out_of_int(SIGNED_VALUE num)
|
|
{
|
|
rb_raise(rb_eRangeError, "integer %"PRIdVALUE " too %s to convert to `int'",
|
|
num, num < 0 ? "small" : "big");
|
|
}
|
|
|
|
static void
|
|
check_int(long num)
|
|
{
|
|
if ((long)(int)num != num) {
|
|
rb_out_of_int(num);
|
|
}
|
|
}
|
|
|
|
static void
|
|
check_uint(unsigned long num, int sign)
|
|
{
|
|
if (sign) {
|
|
/* minus */
|
|
if (num < (unsigned long)INT_MIN)
|
|
rb_raise(rb_eRangeError, "integer %ld too small to convert to `unsigned int'", (long)num);
|
|
}
|
|
else {
|
|
/* plus */
|
|
if (UINT_MAX < num)
|
|
rb_raise(rb_eRangeError, "integer %lu too big to convert to `unsigned int'", num);
|
|
}
|
|
}
|
|
|
|
long
|
|
rb_num2int(VALUE val)
|
|
{
|
|
long num = rb_num2long(val);
|
|
|
|
check_int(num);
|
|
return num;
|
|
}
|
|
|
|
long
|
|
rb_fix2int(VALUE val)
|
|
{
|
|
long num = FIXNUM_P(val)?FIX2LONG(val):rb_num2long(val);
|
|
|
|
check_int(num);
|
|
return num;
|
|
}
|
|
|
|
unsigned long
|
|
rb_num2uint(VALUE val)
|
|
{
|
|
int wrap;
|
|
unsigned long num = rb_num2ulong_internal(val, &wrap);
|
|
|
|
check_uint(num, wrap);
|
|
return num;
|
|
}
|
|
|
|
unsigned long
|
|
rb_fix2uint(VALUE val)
|
|
{
|
|
unsigned long num;
|
|
|
|
if (!FIXNUM_P(val)) {
|
|
return rb_num2uint(val);
|
|
}
|
|
num = FIX2ULONG(val);
|
|
|
|
check_uint(num, rb_num_negative_int_p(val));
|
|
return num;
|
|
}
|
|
#else
|
|
long
|
|
rb_num2int(VALUE val)
|
|
{
|
|
return rb_num2long(val);
|
|
}
|
|
|
|
long
|
|
rb_fix2int(VALUE val)
|
|
{
|
|
return FIX2INT(val);
|
|
}
|
|
#endif
|
|
|
|
NORETURN(static void rb_out_of_short(SIGNED_VALUE num));
|
|
static void
|
|
rb_out_of_short(SIGNED_VALUE num)
|
|
{
|
|
rb_raise(rb_eRangeError, "integer %"PRIdVALUE " too %s to convert to `short'",
|
|
num, num < 0 ? "small" : "big");
|
|
}
|
|
|
|
static void
|
|
check_short(long num)
|
|
{
|
|
if ((long)(short)num != num) {
|
|
rb_out_of_short(num);
|
|
}
|
|
}
|
|
|
|
static void
|
|
check_ushort(unsigned long num, int sign)
|
|
{
|
|
if (sign) {
|
|
/* minus */
|
|
if (num < (unsigned long)SHRT_MIN)
|
|
rb_raise(rb_eRangeError, "integer %ld too small to convert to `unsigned short'", (long)num);
|
|
}
|
|
else {
|
|
/* plus */
|
|
if (USHRT_MAX < num)
|
|
rb_raise(rb_eRangeError, "integer %lu too big to convert to `unsigned short'", num);
|
|
}
|
|
}
|
|
|
|
short
|
|
rb_num2short(VALUE val)
|
|
{
|
|
long num = rb_num2long(val);
|
|
|
|
check_short(num);
|
|
return num;
|
|
}
|
|
|
|
short
|
|
rb_fix2short(VALUE val)
|
|
{
|
|
long num = FIXNUM_P(val)?FIX2LONG(val):rb_num2long(val);
|
|
|
|
check_short(num);
|
|
return num;
|
|
}
|
|
|
|
unsigned short
|
|
rb_num2ushort(VALUE val)
|
|
{
|
|
int wrap;
|
|
unsigned long num = rb_num2ulong_internal(val, &wrap);
|
|
|
|
check_ushort(num, wrap);
|
|
return num;
|
|
}
|
|
|
|
unsigned short
|
|
rb_fix2ushort(VALUE val)
|
|
{
|
|
unsigned long num;
|
|
|
|
if (!FIXNUM_P(val)) {
|
|
return rb_num2ushort(val);
|
|
}
|
|
num = FIX2ULONG(val);
|
|
|
|
check_ushort(num, rb_num_negative_int_p(val));
|
|
return num;
|
|
}
|
|
|
|
VALUE
|
|
rb_num2fix(VALUE val)
|
|
{
|
|
long v;
|
|
|
|
if (FIXNUM_P(val)) return val;
|
|
|
|
v = rb_num2long(val);
|
|
if (!FIXABLE(v))
|
|
rb_raise(rb_eRangeError, "integer %ld out of range of fixnum", v);
|
|
return LONG2FIX(v);
|
|
}
|
|
|
|
#if HAVE_LONG_LONG
|
|
|
|
#define LLONG_MIN_MINUS_ONE ((double)LLONG_MIN-1)
|
|
#define LLONG_MAX_PLUS_ONE (2*(double)(LLONG_MAX/2+1))
|
|
#define ULLONG_MAX_PLUS_ONE (2*(double)(ULLONG_MAX/2+1))
|
|
#ifndef ULLONG_MAX
|
|
#define ULLONG_MAX ((unsigned LONG_LONG)LLONG_MAX*2+1)
|
|
#endif
|
|
#define LLONG_MIN_MINUS_ONE_IS_LESS_THAN(n) \
|
|
(LLONG_MIN_MINUS_ONE == (double)LLONG_MIN ? \
|
|
LLONG_MIN <= (n): \
|
|
LLONG_MIN_MINUS_ONE < (n))
|
|
|
|
LONG_LONG
|
|
rb_num2ll(VALUE val)
|
|
{
|
|
if (NIL_P(val)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from nil");
|
|
}
|
|
|
|
if (FIXNUM_P(val)) return (LONG_LONG)FIX2LONG(val);
|
|
|
|
else if (RB_TYPE_P(val, T_FLOAT)) {
|
|
double d = RFLOAT_VALUE(val);
|
|
if (d < LLONG_MAX_PLUS_ONE && (LLONG_MIN_MINUS_ONE_IS_LESS_THAN(d))) {
|
|
return (LONG_LONG)d;
|
|
}
|
|
else {
|
|
FLOAT_OUT_OF_RANGE(val, "long long");
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(val, T_BIGNUM)) {
|
|
return rb_big2ll(val);
|
|
}
|
|
else if (RB_TYPE_P(val, T_STRING)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from string");
|
|
}
|
|
else if (RB_TYPE_P(val, T_TRUE) || RB_TYPE_P(val, T_FALSE)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from boolean");
|
|
}
|
|
|
|
val = rb_to_int(val);
|
|
return NUM2LL(val);
|
|
}
|
|
|
|
unsigned LONG_LONG
|
|
rb_num2ull(VALUE val)
|
|
{
|
|
if (RB_TYPE_P(val, T_NIL)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from nil");
|
|
}
|
|
else if (RB_TYPE_P(val, T_FIXNUM)) {
|
|
return (LONG_LONG)FIX2LONG(val); /* this is FIX2LONG, intended */
|
|
}
|
|
else if (RB_TYPE_P(val, T_FLOAT)) {
|
|
double d = RFLOAT_VALUE(val);
|
|
if (d < ULLONG_MAX_PLUS_ONE && LLONG_MIN_MINUS_ONE_IS_LESS_THAN(d)) {
|
|
if (0 <= d)
|
|
return (unsigned LONG_LONG)d;
|
|
return (unsigned LONG_LONG)(LONG_LONG)d;
|
|
}
|
|
else {
|
|
FLOAT_OUT_OF_RANGE(val, "unsigned long long");
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(val, T_BIGNUM)) {
|
|
return rb_big2ull(val);
|
|
}
|
|
else if (RB_TYPE_P(val, T_STRING)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from string");
|
|
}
|
|
else if (RB_TYPE_P(val, T_TRUE) || RB_TYPE_P(val, T_FALSE)) {
|
|
rb_raise(rb_eTypeError, "no implicit conversion from boolean");
|
|
}
|
|
|
|
val = rb_to_int(val);
|
|
return NUM2ULL(val);
|
|
}
|
|
|
|
#endif /* HAVE_LONG_LONG */
|
|
|
|
/********************************************************************
|
|
*
|
|
* Document-class: Integer
|
|
*
|
|
* Holds Integer values. You cannot add a singleton method to an
|
|
* Integer object, any attempt to do so will raise a TypeError.
|
|
*
|
|
*/
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.to_i -> integer
|
|
* int.to_int -> integer
|
|
*
|
|
* Since +int+ is already an Integer, returns +self+.
|
|
*
|
|
* #to_int is an alias for #to_i.
|
|
*/
|
|
|
|
static VALUE
|
|
int_to_i(VALUE num)
|
|
{
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.integer? -> true
|
|
*
|
|
* Since +int+ is already an Integer, this always returns +true+.
|
|
*/
|
|
|
|
static VALUE
|
|
int_int_p(VALUE num)
|
|
{
|
|
return Qtrue;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.odd? -> true or false
|
|
*
|
|
* Returns +true+ if +int+ is an odd number.
|
|
*/
|
|
|
|
VALUE
|
|
rb_int_odd_p(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
if (num & 2) {
|
|
return Qtrue;
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_odd_p(num);
|
|
}
|
|
else if (rb_funcall(num, '%', 1, INT2FIX(2)) != INT2FIX(0)) {
|
|
return Qtrue;
|
|
}
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.even? -> true or false
|
|
*
|
|
* Returns +true+ if +int+ is an even number.
|
|
*/
|
|
|
|
static VALUE
|
|
int_even_p(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
if ((num & 2) == 0) {
|
|
return Qtrue;
|
|
}
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_even_p(num);
|
|
}
|
|
else if (rb_funcall(num, '%', 1, INT2FIX(2)) == INT2FIX(0)) {
|
|
return Qtrue;
|
|
}
|
|
return Qfalse;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.allbits?(mask) -> true or false
|
|
*
|
|
* Returns +true+ if all bits of <code>+int+ & +mask+</code> are 1.
|
|
*/
|
|
|
|
static VALUE
|
|
int_allbits_p(VALUE num, VALUE mask)
|
|
{
|
|
mask = rb_to_int(mask);
|
|
return rb_int_equal(rb_int_and(num, mask), mask);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.anybits?(mask) -> true or false
|
|
*
|
|
* Returns +true+ if any bits of <code>+int+ & +mask+</code> are 1.
|
|
*/
|
|
|
|
static VALUE
|
|
int_anybits_p(VALUE num, VALUE mask)
|
|
{
|
|
mask = rb_to_int(mask);
|
|
return num_zero_p(rb_int_and(num, mask)) ? Qfalse : Qtrue;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.nobits?(mask) -> true or false
|
|
*
|
|
* Returns +true+ if no bits of <code>+int+ & +mask+</code> are 1.
|
|
*/
|
|
|
|
static VALUE
|
|
int_nobits_p(VALUE num, VALUE mask)
|
|
{
|
|
mask = rb_to_int(mask);
|
|
return num_zero_p(rb_int_and(num, mask));
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#succ
|
|
* Document-method: Integer#next
|
|
* call-seq:
|
|
* int.next -> integer
|
|
* int.succ -> integer
|
|
*
|
|
* Returns the successor of +int+,
|
|
* i.e. the Integer equal to <code>int+1</code>.
|
|
*
|
|
* 1.next #=> 2
|
|
* (-1).next #=> 0
|
|
* 1.succ #=> 2
|
|
* (-1).succ #=> 0
|
|
*/
|
|
|
|
VALUE
|
|
rb_int_succ(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
long i = FIX2LONG(num) + 1;
|
|
return LONG2NUM(i);
|
|
}
|
|
if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_plus(num, INT2FIX(1));
|
|
}
|
|
return num_funcall1(num, '+', INT2FIX(1));
|
|
}
|
|
|
|
#define int_succ rb_int_succ
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.pred -> integer
|
|
*
|
|
* Returns the predecessor of +int+,
|
|
* i.e. the Integer equal to <code>int-1</code>.
|
|
*
|
|
* 1.pred #=> 0
|
|
* (-1).pred #=> -2
|
|
*/
|
|
|
|
VALUE
|
|
rb_int_pred(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
long i = FIX2LONG(num) - 1;
|
|
return LONG2NUM(i);
|
|
}
|
|
if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_minus(num, INT2FIX(1));
|
|
}
|
|
return num_funcall1(num, '-', INT2FIX(1));
|
|
}
|
|
|
|
#define int_pred rb_int_pred
|
|
|
|
/*
|
|
* Document-method: Integer#chr
|
|
* call-seq:
|
|
* int.chr([encoding]) -> string
|
|
*
|
|
* Returns a string containing the character represented by the +int+'s value
|
|
* according to +encoding+.
|
|
*
|
|
* 65.chr #=> "A"
|
|
* 230.chr #=> "\xE6"
|
|
* 255.chr(Encoding::UTF_8) #=> "\u00FF"
|
|
*/
|
|
|
|
VALUE
|
|
rb_enc_uint_chr(unsigned int code, rb_encoding *enc)
|
|
{
|
|
int n;
|
|
VALUE str;
|
|
switch (n = rb_enc_codelen(code, enc)) {
|
|
case ONIGERR_INVALID_CODE_POINT_VALUE:
|
|
rb_raise(rb_eRangeError, "invalid codepoint 0x%X in %s", code, rb_enc_name(enc));
|
|
break;
|
|
case ONIGERR_TOO_BIG_WIDE_CHAR_VALUE:
|
|
case 0:
|
|
rb_raise(rb_eRangeError, "%u out of char range", code);
|
|
break;
|
|
}
|
|
str = rb_enc_str_new(0, n, enc);
|
|
rb_enc_mbcput(code, RSTRING_PTR(str), enc);
|
|
if (rb_enc_precise_mbclen(RSTRING_PTR(str), RSTRING_END(str), enc) != n) {
|
|
rb_raise(rb_eRangeError, "invalid codepoint 0x%X in %s", code, rb_enc_name(enc));
|
|
}
|
|
return str;
|
|
}
|
|
|
|
static VALUE
|
|
int_chr(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
char c;
|
|
unsigned int i;
|
|
rb_encoding *enc;
|
|
|
|
if (rb_num_to_uint(num, &i) == 0) {
|
|
}
|
|
else if (FIXNUM_P(num)) {
|
|
rb_raise(rb_eRangeError, "%ld out of char range", FIX2LONG(num));
|
|
}
|
|
else {
|
|
rb_raise(rb_eRangeError, "bignum out of char range");
|
|
}
|
|
|
|
switch (argc) {
|
|
case 0:
|
|
if (0xff < i) {
|
|
enc = rb_default_internal_encoding();
|
|
if (!enc) {
|
|
rb_raise(rb_eRangeError, "%d out of char range", i);
|
|
}
|
|
goto decode;
|
|
}
|
|
c = (char)i;
|
|
if (i < 0x80) {
|
|
return rb_usascii_str_new(&c, 1);
|
|
}
|
|
else {
|
|
return rb_str_new(&c, 1);
|
|
}
|
|
case 1:
|
|
break;
|
|
default:
|
|
rb_check_arity(argc, 0, 1);
|
|
break;
|
|
}
|
|
enc = rb_to_encoding(argv[0]);
|
|
if (!enc) enc = rb_ascii8bit_encoding();
|
|
decode:
|
|
return rb_enc_uint_chr(i, enc);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.ord -> self
|
|
*
|
|
* Returns the +int+ itself.
|
|
*
|
|
* 97.ord #=> 97
|
|
*
|
|
* This method is intended for compatibility to character literals
|
|
* in Ruby 1.9.
|
|
*
|
|
* For example, <code>?a.ord</code> returns 97 both in 1.8 and 1.9.
|
|
*/
|
|
|
|
static VALUE
|
|
int_ord(VALUE num)
|
|
{
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* Fixnum
|
|
*/
|
|
|
|
|
|
/*
|
|
* Document-method: Integer#-@
|
|
* call-seq:
|
|
* -int -> integer
|
|
*
|
|
* Returns +int+, negated.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_uminus(VALUE num)
|
|
{
|
|
return LONG2NUM(-FIX2LONG(num));
|
|
}
|
|
|
|
VALUE
|
|
rb_int_uminus(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return fix_uminus(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_uminus(num);
|
|
}
|
|
return num_funcall0(num, idUMinus);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#to_s
|
|
* call-seq:
|
|
* int.to_s(base=10) -> string
|
|
*
|
|
* Returns a string containing the place-value representation of +int+
|
|
* with radix +base+ (between 2 and 36).
|
|
*
|
|
* 12345.to_s #=> "12345"
|
|
* 12345.to_s(2) #=> "11000000111001"
|
|
* 12345.to_s(8) #=> "30071"
|
|
* 12345.to_s(10) #=> "12345"
|
|
* 12345.to_s(16) #=> "3039"
|
|
* 12345.to_s(36) #=> "9ix"
|
|
* 78546939656932.to_s(36) #=> "rubyrules"
|
|
*/
|
|
|
|
VALUE
|
|
rb_fix2str(VALUE x, int base)
|
|
{
|
|
char buf[SIZEOF_VALUE*CHAR_BIT + 1], *const e = buf + sizeof buf, *b = e;
|
|
long val = FIX2LONG(x);
|
|
unsigned long u;
|
|
int neg = 0;
|
|
|
|
if (base < 2 || 36 < base) {
|
|
rb_raise(rb_eArgError, "invalid radix %d", base);
|
|
}
|
|
#if SIZEOF_LONG < SIZEOF_VOIDP
|
|
# if SIZEOF_VOIDP == SIZEOF_LONG_LONG
|
|
if ((val >= 0 && (x & 0xFFFFFFFF00000000ull)) ||
|
|
(val < 0 && (x & 0xFFFFFFFF00000000ull) != 0xFFFFFFFF00000000ull)) {
|
|
rb_bug("Unnormalized Fixnum value %p", (void *)x);
|
|
}
|
|
# else
|
|
/* should do something like above code, but currently ruby does not know */
|
|
/* such platforms */
|
|
# endif
|
|
#endif
|
|
if (val == 0) {
|
|
return rb_usascii_str_new2("0");
|
|
}
|
|
if (val < 0) {
|
|
u = 1 + (unsigned long)(-(val + 1)); /* u = -val avoiding overflow */
|
|
neg = 1;
|
|
}
|
|
else {
|
|
u = val;
|
|
}
|
|
do {
|
|
*--b = ruby_digitmap[(int)(u % base)];
|
|
} while (u /= base);
|
|
if (neg) {
|
|
*--b = '-';
|
|
}
|
|
|
|
return rb_usascii_str_new(b, e - b);
|
|
}
|
|
|
|
static VALUE
|
|
int_to_s(int argc, VALUE *argv, VALUE x)
|
|
{
|
|
int base;
|
|
|
|
if (rb_check_arity(argc, 0, 1))
|
|
base = NUM2INT(argv[0]);
|
|
else
|
|
base = 10;
|
|
return rb_int2str(x, base);
|
|
}
|
|
|
|
VALUE
|
|
rb_int2str(VALUE x, int base)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return rb_fix2str(x, base);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big2str(x, base);
|
|
}
|
|
|
|
return rb_any_to_s(x);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#+
|
|
* call-seq:
|
|
* int + numeric -> numeric_result
|
|
*
|
|
* Performs addition: the class of the resulting object depends on
|
|
* the class of +numeric+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_plus(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
return rb_fix_plus_fix(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_plus(y, x);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM((double)FIX2LONG(x) + RFLOAT_VALUE(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_COMPLEX)) {
|
|
return rb_complex_plus(y, x);
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '+');
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_fix_plus(VALUE x, VALUE y)
|
|
{
|
|
return fix_plus(x, y);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_plus(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_plus(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_plus(x, y);
|
|
}
|
|
return rb_num_coerce_bin(x, y, '+');
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#-
|
|
* call-seq:
|
|
* int - numeric -> numeric_result
|
|
*
|
|
* Performs subtraction: the class of the resulting object depends on
|
|
* the class of +numeric+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_minus(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
return rb_fix_minus_fix(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
x = rb_int2big(FIX2LONG(x));
|
|
return rb_big_minus(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM((double)FIX2LONG(x) - RFLOAT_VALUE(y));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '-');
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_minus(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_minus(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_minus(x, y);
|
|
}
|
|
return rb_num_coerce_bin(x, y, '-');
|
|
}
|
|
|
|
|
|
#define SQRT_LONG_MAX HALF_LONG_MSB
|
|
/*tests if N*N would overflow*/
|
|
#define FIT_SQRT_LONG(n) (((n)<SQRT_LONG_MAX)&&((n)>=-SQRT_LONG_MAX))
|
|
|
|
/*
|
|
* Document-method: Integer#*
|
|
* call-seq:
|
|
* int * numeric -> numeric_result
|
|
*
|
|
* Performs multiplication: the class of the resulting object depends on
|
|
* the class of +numeric+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_mul(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
return rb_fix_mul_fix(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
switch (x) {
|
|
case INT2FIX(0): return x;
|
|
case INT2FIX(1): return y;
|
|
}
|
|
return rb_big_mul(y, x);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM((double)FIX2LONG(x) * RFLOAT_VALUE(y));
|
|
}
|
|
else if (RB_TYPE_P(y, T_COMPLEX)) {
|
|
return rb_complex_mul(y, x);
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '*');
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_mul(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_mul(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_mul(x, y);
|
|
}
|
|
return rb_num_coerce_bin(x, y, '*');
|
|
}
|
|
|
|
static double
|
|
fix_fdiv_double(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
return (double)FIX2LONG(x) / (double)FIX2LONG(y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_fdiv_double(rb_int2big(FIX2LONG(x)), y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return (double)FIX2LONG(x) / RFLOAT_VALUE(y);
|
|
}
|
|
else {
|
|
return NUM2DBL(rb_num_coerce_bin(x, y, rb_intern("fdiv")));
|
|
}
|
|
}
|
|
|
|
double
|
|
rb_int_fdiv_double(VALUE x, VALUE y)
|
|
{
|
|
if (RB_INTEGER_TYPE_P(y) && !FIXNUM_ZERO_P(y)) {
|
|
VALUE gcd = rb_gcd(x, y);
|
|
if (!FIXNUM_ZERO_P(gcd)) {
|
|
x = rb_int_idiv(x, gcd);
|
|
y = rb_int_idiv(y, gcd);
|
|
}
|
|
}
|
|
if (FIXNUM_P(x)) {
|
|
return fix_fdiv_double(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_fdiv_double(x, y);
|
|
}
|
|
return NAN;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#fdiv
|
|
* call-seq:
|
|
* int.fdiv(numeric) -> float
|
|
*
|
|
* Returns the floating point result of dividing +int+ by +numeric+.
|
|
*
|
|
* 654321.fdiv(13731) #=> 47.652829364212366
|
|
* 654321.fdiv(13731.24) #=> 47.65199646936475
|
|
* -654321.fdiv(13731) #=> -47.652829364212366
|
|
*/
|
|
|
|
VALUE
|
|
rb_int_fdiv(VALUE x, VALUE y)
|
|
{
|
|
if (RB_INTEGER_TYPE_P(x)) {
|
|
return DBL2NUM(rb_int_fdiv_double(x, y));
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#/
|
|
* call-seq:
|
|
* int / numeric -> numeric_result
|
|
*
|
|
* Performs division: the class of the resulting object depends on
|
|
* the class of +numeric+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_divide(VALUE x, VALUE y, ID op)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIXNUM_ZERO_P(y)) rb_num_zerodiv();
|
|
return rb_fix_div_fix(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
x = rb_int2big(FIX2LONG(x));
|
|
return rb_big_div(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
{
|
|
double div;
|
|
|
|
if (op == '/') {
|
|
div = (double)FIX2LONG(x) / RFLOAT_VALUE(y);
|
|
return DBL2NUM(div);
|
|
}
|
|
else {
|
|
if (RFLOAT_VALUE(y) == 0) rb_num_zerodiv();
|
|
div = (double)FIX2LONG(x) / RFLOAT_VALUE(y);
|
|
return rb_dbl2big(floor(div));
|
|
}
|
|
}
|
|
}
|
|
else {
|
|
if (RB_TYPE_P(y, T_RATIONAL) &&
|
|
op == '/' && FIX2LONG(x) == 1)
|
|
return rb_rational_reciprocal(y);
|
|
return rb_num_coerce_bin(x, y, op);
|
|
}
|
|
}
|
|
|
|
static VALUE
|
|
fix_div(VALUE x, VALUE y)
|
|
{
|
|
return fix_divide(x, y, '/');
|
|
}
|
|
|
|
VALUE
|
|
rb_int_div(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_div(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_div(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#div
|
|
* call-seq:
|
|
* int.div(numeric) -> integer
|
|
*
|
|
* Performs integer division: returns the integer result of dividing +int+
|
|
* by +numeric+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_idiv(VALUE x, VALUE y)
|
|
{
|
|
return fix_divide(x, y, id_div);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_idiv(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_idiv(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_idiv(x, y);
|
|
}
|
|
return num_div(x, y);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#%
|
|
* Document-method: Integer#modulo
|
|
* call-seq:
|
|
* int % other -> real
|
|
* int.modulo(other) -> real
|
|
*
|
|
* Returns +int+ modulo +other+.
|
|
*
|
|
* See Numeric#divmod for more information.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_mod(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIXNUM_ZERO_P(y)) rb_num_zerodiv();
|
|
return rb_fix_mod_fix(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
x = rb_int2big(FIX2LONG(x));
|
|
return rb_big_modulo(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return DBL2NUM(ruby_float_mod((double)FIX2LONG(x), RFLOAT_VALUE(y)));
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, '%');
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_modulo(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_mod(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_modulo(x, y);
|
|
}
|
|
return num_modulo(x, y);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* int.remainder(numeric) -> real
|
|
*
|
|
* Returns the remainder after dividing +int+ by +numeric+.
|
|
*
|
|
* <code>x.remainder(y)</code> means <code>x-y*(x/y).truncate</code>.
|
|
*
|
|
* 5.remainder(3) #=> 2
|
|
* -5.remainder(3) #=> -2
|
|
* 5.remainder(-3) #=> 2
|
|
* -5.remainder(-3) #=> -2
|
|
* 5.remainder(1.5) #=> 0.5
|
|
*
|
|
* See Numeric#divmod.
|
|
*/
|
|
|
|
VALUE
|
|
int_remainder(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return num_remainder(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_remainder(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#divmod
|
|
* call-seq:
|
|
* int.divmod(numeric) -> array
|
|
*
|
|
* See Numeric#divmod.
|
|
*/
|
|
static VALUE
|
|
fix_divmod(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
VALUE div, mod;
|
|
if (FIXNUM_ZERO_P(y)) rb_num_zerodiv();
|
|
rb_fix_divmod_fix(x, y, &div, &mod);
|
|
return rb_assoc_new(div, mod);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
x = rb_int2big(FIX2LONG(x));
|
|
return rb_big_divmod(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
{
|
|
double div, mod;
|
|
volatile VALUE a, b;
|
|
|
|
flodivmod((double)FIX2LONG(x), RFLOAT_VALUE(y), &div, &mod);
|
|
a = dbl2ival(div);
|
|
b = DBL2NUM(mod);
|
|
return rb_assoc_new(a, b);
|
|
}
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, id_divmod);
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_divmod(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_divmod(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_divmod(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#**
|
|
* call-seq:
|
|
* int ** numeric -> numeric_result
|
|
*
|
|
* Raises +int+ to the power of +numeric+, which may be negative or
|
|
* fractional.
|
|
* The result may be an Integer, a Float, a Rational, or a complex number.
|
|
*
|
|
* 2 ** 3 #=> 8
|
|
* 2 ** -1 #=> (1/2)
|
|
* 2 ** 0.5 #=> 1.4142135623730951
|
|
* (-1) ** 0.5 #=> (0.0+1.0i)
|
|
*
|
|
* 123456789 ** 2 #=> 15241578750190521
|
|
* 123456789 ** 1.2 #=> 5126464716.0993185
|
|
* 123456789 ** -2 #=> (1/15241578750190521)
|
|
*/
|
|
|
|
static VALUE
|
|
int_pow(long x, unsigned long y)
|
|
{
|
|
int neg = x < 0;
|
|
long z = 1;
|
|
|
|
if (y == 0) return INT2FIX(1);
|
|
if (neg) x = -x;
|
|
if (y & 1)
|
|
z = x;
|
|
else
|
|
neg = 0;
|
|
y &= ~1;
|
|
do {
|
|
while (y % 2 == 0) {
|
|
if (!FIT_SQRT_LONG(x)) {
|
|
VALUE v;
|
|
bignum:
|
|
v = rb_big_pow(rb_int2big(x), LONG2NUM(y));
|
|
if (RB_FLOAT_TYPE_P(v)) /* infinity due to overflow */
|
|
return v;
|
|
if (z != 1) v = rb_big_mul(rb_int2big(neg ? -z : z), v);
|
|
return v;
|
|
}
|
|
x = x * x;
|
|
y >>= 1;
|
|
}
|
|
{
|
|
if (MUL_OVERFLOW_FIXNUM_P(x, z)) {
|
|
goto bignum;
|
|
}
|
|
z = x * z;
|
|
}
|
|
} while (--y);
|
|
if (neg) z = -z;
|
|
return LONG2NUM(z);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_positive_pow(long x, unsigned long y)
|
|
{
|
|
return int_pow(x, y);
|
|
}
|
|
|
|
static VALUE
|
|
fix_pow(VALUE x, VALUE y)
|
|
{
|
|
long a = FIX2LONG(x);
|
|
|
|
if (FIXNUM_P(y)) {
|
|
long b = FIX2LONG(y);
|
|
|
|
if (a == 1) return INT2FIX(1);
|
|
if (a == -1) {
|
|
if (b % 2 == 0)
|
|
return INT2FIX(1);
|
|
else
|
|
return INT2FIX(-1);
|
|
}
|
|
if (b < 0)
|
|
return num_funcall1(rb_rational_raw1(x), idPow, y);
|
|
|
|
if (b == 0) return INT2FIX(1);
|
|
if (b == 1) return x;
|
|
if (a == 0) {
|
|
if (b > 0) return INT2FIX(0);
|
|
return DBL2NUM(INFINITY);
|
|
}
|
|
return int_pow(a, b);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
if (a == 1) return INT2FIX(1);
|
|
if (a == -1) {
|
|
if (int_even_p(y)) return INT2FIX(1);
|
|
else return INT2FIX(-1);
|
|
}
|
|
if (rb_num_negative_int_p(y))
|
|
return num_funcall1(rb_rational_raw1(x), idPow, y);
|
|
if (a == 0) return INT2FIX(0);
|
|
x = rb_int2big(FIX2LONG(x));
|
|
return rb_big_pow(x, y);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
double dy = RFLOAT_VALUE(y);
|
|
if (dy == 0.0) return DBL2NUM(1.0);
|
|
if (a == 0) {
|
|
return DBL2NUM(dy < 0 ? INFINITY : 0.0);
|
|
}
|
|
if (a == 1) return DBL2NUM(1.0);
|
|
{
|
|
if (a < 0 && dy != round(dy))
|
|
return num_funcall1(rb_complex_raw1(x), idPow, y);
|
|
return DBL2NUM(pow((double)a, dy));
|
|
}
|
|
}
|
|
else {
|
|
return rb_num_coerce_bin(x, y, idPow);
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_pow(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_pow(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_pow(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#==
|
|
* Document-method: Integer#===
|
|
* call-seq:
|
|
* int == other -> true or false
|
|
*
|
|
* Returns +true+ if +int+ equals +other+ numerically.
|
|
* Contrast this with Integer#eql?, which requires +other+ to be an Integer.
|
|
*
|
|
* 1 == 2 #=> false
|
|
* 1 == 1.0 #=> true
|
|
*/
|
|
|
|
static VALUE
|
|
fix_equal(VALUE x, VALUE y)
|
|
{
|
|
if (x == y) return Qtrue;
|
|
if (FIXNUM_P(y)) return Qfalse;
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_eq(y, x);
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return rb_integer_float_eq(x, y);
|
|
}
|
|
else {
|
|
return num_equal(x, y);
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_equal(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_equal(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_eq(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#<=>
|
|
* call-seq:
|
|
* int <=> numeric -> -1, 0, +1, or nil
|
|
*
|
|
* Comparison---Returns -1, 0, or +1 depending on whether +int+ is
|
|
* less than, equal to, or greater than +numeric+.
|
|
*
|
|
* This is the basis for the tests in the Comparable module.
|
|
*
|
|
* +nil+ is returned if the two values are incomparable.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_cmp(VALUE x, VALUE y)
|
|
{
|
|
if (x == y) return INT2FIX(0);
|
|
if (FIXNUM_P(y)) {
|
|
if (FIX2LONG(x) > FIX2LONG(y)) return INT2FIX(1);
|
|
return INT2FIX(-1);
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
VALUE cmp = rb_big_cmp(y, x);
|
|
switch (cmp) {
|
|
case INT2FIX(+1): return INT2FIX(-1);
|
|
case INT2FIX(-1): return INT2FIX(+1);
|
|
}
|
|
return cmp;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return rb_integer_float_cmp(x, y);
|
|
}
|
|
else {
|
|
return rb_num_coerce_cmp(x, y, id_cmp);
|
|
}
|
|
return rb_num_coerce_cmp(x, y, id_cmp);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_cmp(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_cmp(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_cmp(x, y);
|
|
}
|
|
else {
|
|
rb_raise(rb_eNotImpError, "need to define `<=>' in %s", rb_obj_classname(x));
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#>
|
|
* call-seq:
|
|
* int > real -> true or false
|
|
*
|
|
* Returns +true+ if the value of +int+ is greater than that of +real+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_gt(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIX2LONG(x) > FIX2LONG(y)) return Qtrue;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_cmp(y, x) == INT2FIX(-1) ? Qtrue : Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return rb_integer_float_cmp(x, y) == INT2FIX(1) ? Qtrue : Qfalse;
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, '>');
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_gt(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_gt(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_gt(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#>=
|
|
* call-seq:
|
|
* int >= real -> true or false
|
|
*
|
|
* Returns +true+ if the value of +int+ is greater than or equal to that of
|
|
* +real+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_ge(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIX2LONG(x) >= FIX2LONG(y)) return Qtrue;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_cmp(y, x) != INT2FIX(+1) ? Qtrue : Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
VALUE rel = rb_integer_float_cmp(x, y);
|
|
return rel == INT2FIX(1) || rel == INT2FIX(0) ? Qtrue : Qfalse;
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, idGE);
|
|
}
|
|
}
|
|
|
|
VALUE
|
|
rb_int_ge(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_ge(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_ge(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#<
|
|
* call-seq:
|
|
* int < real -> true or false
|
|
*
|
|
* Returns +true+ if the value of +int+ is less than that of +real+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_lt(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIX2LONG(x) < FIX2LONG(y)) return Qtrue;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_cmp(y, x) == INT2FIX(+1) ? Qtrue : Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
return rb_integer_float_cmp(x, y) == INT2FIX(-1) ? Qtrue : Qfalse;
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, '<');
|
|
}
|
|
}
|
|
|
|
static VALUE
|
|
int_lt(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_lt(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_lt(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#<=
|
|
* call-seq:
|
|
* int <= real -> true or false
|
|
*
|
|
* Returns +true+ if the value of +int+ is less than or equal to that of
|
|
* +real+.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_le(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
if (FIX2LONG(x) <= FIX2LONG(y)) return Qtrue;
|
|
return Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_cmp(y, x) != INT2FIX(-1) ? Qtrue : Qfalse;
|
|
}
|
|
else if (RB_TYPE_P(y, T_FLOAT)) {
|
|
VALUE rel = rb_integer_float_cmp(x, y);
|
|
return rel == INT2FIX(-1) || rel == INT2FIX(0) ? Qtrue : Qfalse;
|
|
}
|
|
else {
|
|
return rb_num_coerce_relop(x, y, idLE);
|
|
}
|
|
}
|
|
|
|
static VALUE
|
|
int_le(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_le(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_le(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#~
|
|
* call-seq:
|
|
* ~int -> integer
|
|
*
|
|
* One's complement: returns a number where each bit is flipped.
|
|
*
|
|
* Inverts the bits in an Integer. As integers are conceptually of
|
|
* infinite length, the result acts as if it had an infinite number of
|
|
* one bits to the left. In hex representations, this is displayed
|
|
* as two periods to the left of the digits.
|
|
*
|
|
* sprintf("%X", ~0x1122334455) #=> "..FEEDDCCBBAA"
|
|
*/
|
|
|
|
static VALUE
|
|
fix_comp(VALUE num)
|
|
{
|
|
return ~num | FIXNUM_FLAG;
|
|
}
|
|
|
|
static VALUE
|
|
int_comp(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return fix_comp(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_comp(num);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
static VALUE
|
|
num_funcall_bit_1(VALUE y, VALUE arg, int recursive)
|
|
{
|
|
ID func = (ID)((VALUE *)arg)[0];
|
|
VALUE x = ((VALUE *)arg)[1];
|
|
if (recursive) {
|
|
num_funcall_op_1_recursion(x, func, y);
|
|
}
|
|
return rb_check_funcall(x, func, 1, &y);
|
|
}
|
|
|
|
VALUE
|
|
rb_num_coerce_bit(VALUE x, VALUE y, ID func)
|
|
{
|
|
VALUE ret, args[3];
|
|
|
|
args[0] = (VALUE)func;
|
|
args[1] = x;
|
|
args[2] = y;
|
|
do_coerce(&args[1], &args[2], TRUE);
|
|
ret = rb_exec_recursive_paired(num_funcall_bit_1,
|
|
args[2], args[1], (VALUE)args);
|
|
if (ret == Qundef) {
|
|
/* show the original object, not coerced object */
|
|
coerce_failed(x, y);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#&
|
|
* call-seq:
|
|
* int & other_int -> integer
|
|
*
|
|
* Bitwise AND.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_and(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
long val = FIX2LONG(x) & FIX2LONG(y);
|
|
return LONG2NUM(val);
|
|
}
|
|
|
|
if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_and(y, x);
|
|
}
|
|
|
|
return rb_num_coerce_bit(x, y, '&');
|
|
}
|
|
|
|
VALUE
|
|
rb_int_and(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_and(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_and(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#|
|
|
* call-seq:
|
|
* int | other_int -> integer
|
|
*
|
|
* Bitwise OR.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_or(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
long val = FIX2LONG(x) | FIX2LONG(y);
|
|
return LONG2NUM(val);
|
|
}
|
|
|
|
if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_or(y, x);
|
|
}
|
|
|
|
return rb_num_coerce_bit(x, y, '|');
|
|
}
|
|
|
|
static VALUE
|
|
int_or(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_or(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_or(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#^
|
|
* call-seq:
|
|
* int ^ other_int -> integer
|
|
*
|
|
* Bitwise EXCLUSIVE OR.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_xor(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(y)) {
|
|
long val = FIX2LONG(x) ^ FIX2LONG(y);
|
|
return LONG2NUM(val);
|
|
}
|
|
|
|
if (RB_TYPE_P(y, T_BIGNUM)) {
|
|
return rb_big_xor(y, x);
|
|
}
|
|
|
|
return rb_num_coerce_bit(x, y, '^');
|
|
}
|
|
|
|
static VALUE
|
|
int_xor(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return fix_xor(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_xor(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#<<
|
|
* call-seq:
|
|
* int << count -> integer
|
|
*
|
|
* Returns +int+ shifted left +count+ positions, or right if +count+
|
|
* is negative.
|
|
*/
|
|
|
|
static VALUE
|
|
rb_fix_lshift(VALUE x, VALUE y)
|
|
{
|
|
long val, width;
|
|
|
|
val = NUM2LONG(x);
|
|
if (!FIXNUM_P(y))
|
|
return rb_big_lshift(rb_int2big(val), y);
|
|
width = FIX2LONG(y);
|
|
if (width < 0)
|
|
return fix_rshift(val, (unsigned long)-width);
|
|
return fix_lshift(val, width);
|
|
}
|
|
|
|
static VALUE
|
|
fix_lshift(long val, unsigned long width)
|
|
{
|
|
if (width > (SIZEOF_LONG*CHAR_BIT-1)
|
|
|| ((unsigned long)val)>>(SIZEOF_LONG*CHAR_BIT-1-width) > 0) {
|
|
return rb_big_lshift(rb_int2big(val), ULONG2NUM(width));
|
|
}
|
|
val = val << width;
|
|
return LONG2NUM(val);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_lshift(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return rb_fix_lshift(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_lshift(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#>>
|
|
* call-seq:
|
|
* int >> count -> integer
|
|
*
|
|
* Returns +int+ shifted right +count+ positions, or left if +count+
|
|
* is negative.
|
|
*/
|
|
|
|
static VALUE
|
|
rb_fix_rshift(VALUE x, VALUE y)
|
|
{
|
|
long i, val;
|
|
|
|
val = FIX2LONG(x);
|
|
if (!FIXNUM_P(y))
|
|
return rb_big_rshift(rb_int2big(val), y);
|
|
i = FIX2LONG(y);
|
|
if (i == 0) return x;
|
|
if (i < 0)
|
|
return fix_lshift(val, (unsigned long)-i);
|
|
return fix_rshift(val, i);
|
|
}
|
|
|
|
static VALUE
|
|
fix_rshift(long val, unsigned long i)
|
|
{
|
|
if (i >= sizeof(long)*CHAR_BIT-1) {
|
|
if (val < 0) return INT2FIX(-1);
|
|
return INT2FIX(0);
|
|
}
|
|
val = RSHIFT(val, i);
|
|
return LONG2FIX(val);
|
|
}
|
|
|
|
static VALUE
|
|
rb_int_rshift(VALUE x, VALUE y)
|
|
{
|
|
if (FIXNUM_P(x)) {
|
|
return rb_fix_rshift(x, y);
|
|
}
|
|
else if (RB_TYPE_P(x, T_BIGNUM)) {
|
|
return rb_big_rshift(x, y);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#[]
|
|
* call-seq:
|
|
* int[n] -> 0, 1
|
|
*
|
|
* Bit Reference---Returns the <code>n</code>th bit in the
|
|
* binary representation of +int+, where <code>int[0]</code>
|
|
* is the least significant bit.
|
|
*
|
|
* a = 0b11001100101010
|
|
* 30.downto(0) {|n| print a[n] }
|
|
* #=> 0000000000000000011001100101010
|
|
*
|
|
* a = 9**15
|
|
* 50.downto(0) {|n| print a[n] }
|
|
* #=> 000101110110100000111000011110010100111100010111001
|
|
*/
|
|
|
|
static VALUE
|
|
fix_aref(VALUE fix, VALUE idx)
|
|
{
|
|
long val = FIX2LONG(fix);
|
|
long i;
|
|
|
|
idx = rb_to_int(idx);
|
|
if (!FIXNUM_P(idx)) {
|
|
idx = rb_big_norm(idx);
|
|
if (!FIXNUM_P(idx)) {
|
|
if (!BIGNUM_SIGN(idx) || val >= 0)
|
|
return INT2FIX(0);
|
|
return INT2FIX(1);
|
|
}
|
|
}
|
|
i = FIX2LONG(idx);
|
|
|
|
if (i < 0) return INT2FIX(0);
|
|
if (SIZEOF_LONG*CHAR_BIT-1 <= i) {
|
|
if (val < 0) return INT2FIX(1);
|
|
return INT2FIX(0);
|
|
}
|
|
if (val & (1L<<i))
|
|
return INT2FIX(1);
|
|
return INT2FIX(0);
|
|
}
|
|
|
|
static VALUE
|
|
int_aref(VALUE num, VALUE idx)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return fix_aref(num, idx);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_aref(num, idx);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#to_f
|
|
* call-seq:
|
|
* int.to_f -> float
|
|
*
|
|
* Converts +int+ to a Float. If +int+ doesn't fit in a Float,
|
|
* the result is infinity.
|
|
*/
|
|
|
|
static VALUE
|
|
int_to_f(VALUE num)
|
|
{
|
|
double val;
|
|
|
|
if (FIXNUM_P(num)) {
|
|
val = (double)FIX2LONG(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
val = rb_big2dbl(num);
|
|
}
|
|
else {
|
|
rb_raise(rb_eNotImpError, "Unknown subclass for to_f: %s", rb_obj_classname(num));
|
|
}
|
|
|
|
return DBL2NUM(val);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#abs
|
|
* Document-method: Integer#magnitude
|
|
* call-seq:
|
|
* int.abs -> integer
|
|
* int.magnitude -> integer
|
|
*
|
|
* Returns the absolute value of +int+.
|
|
*
|
|
* (-12345).abs #=> 12345
|
|
* -12345.abs #=> 12345
|
|
* 12345.abs #=> 12345
|
|
*
|
|
* Integer#magnitude is an alias for Integer#abs.
|
|
*/
|
|
|
|
static VALUE
|
|
fix_abs(VALUE fix)
|
|
{
|
|
long i = FIX2LONG(fix);
|
|
|
|
if (i < 0) i = -i;
|
|
|
|
return LONG2NUM(i);
|
|
}
|
|
|
|
VALUE
|
|
rb_int_abs(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return fix_abs(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_abs(num);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#size
|
|
* call-seq:
|
|
* int.size -> int
|
|
*
|
|
* Returns the number of bytes in the machine representation of +int+
|
|
* (machine dependent).
|
|
*
|
|
* 1.size #=> 8
|
|
* -1.size #=> 8
|
|
* 2147483647.size #=> 8
|
|
* (256**10 - 1).size #=> 10
|
|
* (256**20 - 1).size #=> 20
|
|
* (256**40 - 1).size #=> 40
|
|
*/
|
|
|
|
static VALUE
|
|
fix_size(VALUE fix)
|
|
{
|
|
return INT2FIX(sizeof(long));
|
|
}
|
|
|
|
static VALUE
|
|
int_size(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return fix_size(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_size_m(num);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#bit_length
|
|
* call-seq:
|
|
* int.bit_length -> integer
|
|
*
|
|
* Returns the number of bits of the value of +int+.
|
|
*
|
|
* "Number of bits" means the bit position of the highest bit
|
|
* which is different from the sign bit
|
|
* (where the least significant bit has bit position 1).
|
|
* If there is no such bit (zero or minus one), zero is returned.
|
|
*
|
|
* I.e. this method returns <i>ceil(log2(int < 0 ? -int : int+1))</i>.
|
|
*
|
|
* (-2**1000-1).bit_length #=> 1001
|
|
* (-2**1000).bit_length #=> 1000
|
|
* (-2**1000+1).bit_length #=> 1000
|
|
* (-2**12-1).bit_length #=> 13
|
|
* (-2**12).bit_length #=> 12
|
|
* (-2**12+1).bit_length #=> 12
|
|
* -0x101.bit_length #=> 9
|
|
* -0x100.bit_length #=> 8
|
|
* -0xff.bit_length #=> 8
|
|
* -2.bit_length #=> 1
|
|
* -1.bit_length #=> 0
|
|
* 0.bit_length #=> 0
|
|
* 1.bit_length #=> 1
|
|
* 0xff.bit_length #=> 8
|
|
* 0x100.bit_length #=> 9
|
|
* (2**12-1).bit_length #=> 12
|
|
* (2**12).bit_length #=> 13
|
|
* (2**12+1).bit_length #=> 13
|
|
* (2**1000-1).bit_length #=> 1000
|
|
* (2**1000).bit_length #=> 1001
|
|
* (2**1000+1).bit_length #=> 1001
|
|
*
|
|
* This method can be used to detect overflow in Array#pack as follows:
|
|
*
|
|
* if n.bit_length < 32
|
|
* [n].pack("l") # no overflow
|
|
* else
|
|
* raise "overflow"
|
|
* end
|
|
*/
|
|
|
|
static VALUE
|
|
rb_fix_bit_length(VALUE fix)
|
|
{
|
|
long v = FIX2LONG(fix);
|
|
if (v < 0)
|
|
v = ~v;
|
|
return LONG2FIX(bit_length(v));
|
|
}
|
|
|
|
static VALUE
|
|
rb_int_bit_length(VALUE num)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
return rb_fix_bit_length(num);
|
|
}
|
|
else if (RB_TYPE_P(num, T_BIGNUM)) {
|
|
return rb_big_bit_length(num);
|
|
}
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#digits
|
|
* call-seq:
|
|
* int.digits -> array
|
|
* int.digits(base) -> array
|
|
*
|
|
* Returns the digits of +int+'s place-value representation
|
|
* with radix +base+ (default: 10).
|
|
* The digits are returned as an array with the least significant digit
|
|
* as the first array element.
|
|
*
|
|
* +base+ must be greater than or equal to 2.
|
|
*
|
|
* 12345.digits #=> [5, 4, 3, 2, 1]
|
|
* 12345.digits(7) #=> [4, 6, 6, 0, 5]
|
|
* 12345.digits(100) #=> [45, 23, 1]
|
|
*
|
|
* -12345.digits(7) #=> Math::DomainError
|
|
*/
|
|
|
|
static VALUE
|
|
rb_fix_digits(VALUE fix, long base)
|
|
{
|
|
VALUE digits;
|
|
long x = FIX2LONG(fix);
|
|
|
|
assert(x >= 0);
|
|
|
|
if (base < 2)
|
|
rb_raise(rb_eArgError, "invalid radix %ld", base);
|
|
|
|
if (x == 0)
|
|
return rb_ary_new_from_args(1, INT2FIX(0));
|
|
|
|
digits = rb_ary_new();
|
|
while (x > 0) {
|
|
long q = x % base;
|
|
rb_ary_push(digits, LONG2NUM(q));
|
|
x /= base;
|
|
}
|
|
|
|
return digits;
|
|
}
|
|
|
|
static VALUE
|
|
rb_int_digits_bigbase(VALUE num, VALUE base)
|
|
{
|
|
VALUE digits;
|
|
|
|
assert(!rb_num_negative_p(num));
|
|
|
|
if (RB_TYPE_P(base, T_BIGNUM))
|
|
base = rb_big_norm(base);
|
|
|
|
if (FIXNUM_P(base) && FIX2LONG(base) < 2)
|
|
rb_raise(rb_eArgError, "invalid radix %ld", FIX2LONG(base));
|
|
else if (RB_TYPE_P(base, T_BIGNUM) && BIGNUM_NEGATIVE_P(base))
|
|
rb_raise(rb_eArgError, "negative radix");
|
|
|
|
if (FIXNUM_P(base) && FIXNUM_P(num))
|
|
return rb_fix_digits(num, FIX2LONG(base));
|
|
|
|
if (FIXNUM_P(num))
|
|
return rb_ary_new_from_args(1, num);
|
|
|
|
digits = rb_ary_new();
|
|
while (!FIXNUM_P(num) || FIX2LONG(num) > 0) {
|
|
VALUE qr = rb_int_divmod(num, base);
|
|
rb_ary_push(digits, RARRAY_AREF(qr, 1));
|
|
num = RARRAY_AREF(qr, 0);
|
|
}
|
|
|
|
return digits;
|
|
}
|
|
|
|
static VALUE
|
|
rb_int_digits(int argc, VALUE *argv, VALUE num)
|
|
{
|
|
VALUE base_value;
|
|
long base;
|
|
|
|
if (rb_num_negative_p(num))
|
|
rb_raise(rb_eMathDomainError, "out of domain");
|
|
|
|
if (rb_check_arity(argc, 0, 1)) {
|
|
base_value = rb_to_int(argv[0]);
|
|
if (!RB_INTEGER_TYPE_P(base_value))
|
|
rb_raise(rb_eTypeError, "wrong argument type %s (expected Integer)",
|
|
rb_obj_classname(argv[0]));
|
|
if (RB_TYPE_P(base_value, T_BIGNUM))
|
|
return rb_int_digits_bigbase(num, base_value);
|
|
|
|
base = FIX2LONG(base_value);
|
|
if (base < 0)
|
|
rb_raise(rb_eArgError, "negative radix");
|
|
else if (base < 2)
|
|
rb_raise(rb_eArgError, "invalid radix %ld", base);
|
|
}
|
|
else
|
|
base = 10;
|
|
|
|
if (FIXNUM_P(num))
|
|
return rb_fix_digits(num, base);
|
|
else if (RB_TYPE_P(num, T_BIGNUM))
|
|
return rb_int_digits_bigbase(num, LONG2FIX(base));
|
|
|
|
return Qnil;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#upto
|
|
* call-seq:
|
|
* int.upto(limit) {|i| block } -> self
|
|
* int.upto(limit) -> an_enumerator
|
|
*
|
|
* Iterates the given block, passing in integer values from +int+ up to and
|
|
* including +limit+.
|
|
*
|
|
* If no block is given, an Enumerator is returned instead.
|
|
*
|
|
* 5.upto(10) {|i| print i, " " } #=> 5 6 7 8 9 10
|
|
*/
|
|
|
|
static VALUE
|
|
int_upto_size(VALUE from, VALUE args, VALUE eobj)
|
|
{
|
|
return ruby_num_interval_step_size(from, RARRAY_AREF(args, 0), INT2FIX(1), FALSE);
|
|
}
|
|
|
|
static VALUE
|
|
int_upto(VALUE from, VALUE to)
|
|
{
|
|
RETURN_SIZED_ENUMERATOR(from, 1, &to, int_upto_size);
|
|
if (FIXNUM_P(from) && FIXNUM_P(to)) {
|
|
long i, end;
|
|
|
|
end = FIX2LONG(to);
|
|
for (i = FIX2LONG(from); i <= end; i++) {
|
|
rb_yield(LONG2FIX(i));
|
|
}
|
|
}
|
|
else {
|
|
VALUE i = from, c;
|
|
|
|
while (!(c = rb_funcall(i, '>', 1, to))) {
|
|
rb_yield(i);
|
|
i = rb_funcall(i, '+', 1, INT2FIX(1));
|
|
}
|
|
if (NIL_P(c)) rb_cmperr(i, to);
|
|
}
|
|
return from;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#downto
|
|
* call-seq:
|
|
* int.downto(limit) {|i| block } -> self
|
|
* int.downto(limit) -> an_enumerator
|
|
*
|
|
* Iterates the given block, passing in decreasing values from +int+ down to
|
|
* and including +limit+.
|
|
*
|
|
* If no block is given, an Enumerator is returned instead.
|
|
*
|
|
* 5.downto(1) { |n| print n, ".. " }
|
|
* puts "Liftoff!"
|
|
* #=> "5.. 4.. 3.. 2.. 1.. Liftoff!"
|
|
*/
|
|
|
|
static VALUE
|
|
int_downto_size(VALUE from, VALUE args, VALUE eobj)
|
|
{
|
|
return ruby_num_interval_step_size(from, RARRAY_AREF(args, 0), INT2FIX(-1), FALSE);
|
|
}
|
|
|
|
static VALUE
|
|
int_downto(VALUE from, VALUE to)
|
|
{
|
|
RETURN_SIZED_ENUMERATOR(from, 1, &to, int_downto_size);
|
|
if (FIXNUM_P(from) && FIXNUM_P(to)) {
|
|
long i, end;
|
|
|
|
end = FIX2LONG(to);
|
|
for (i=FIX2LONG(from); i >= end; i--) {
|
|
rb_yield(LONG2FIX(i));
|
|
}
|
|
}
|
|
else {
|
|
VALUE i = from, c;
|
|
|
|
while (!(c = rb_funcall(i, '<', 1, to))) {
|
|
rb_yield(i);
|
|
i = rb_funcall(i, '-', 1, INT2FIX(1));
|
|
}
|
|
if (NIL_P(c)) rb_cmperr(i, to);
|
|
}
|
|
return from;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#times
|
|
* call-seq:
|
|
* int.times {|i| block } -> self
|
|
* int.times -> an_enumerator
|
|
*
|
|
* Iterates the given block +int+ times, passing in values from zero to
|
|
* <code>int - 1</code>.
|
|
*
|
|
* If no block is given, an Enumerator is returned instead.
|
|
*
|
|
* 5.times {|i| print i, " " } #=> 0 1 2 3 4
|
|
*/
|
|
|
|
static VALUE
|
|
int_dotimes_size(VALUE num, VALUE args, VALUE eobj)
|
|
{
|
|
if (FIXNUM_P(num)) {
|
|
if (NUM2LONG(num) <= 0) return INT2FIX(0);
|
|
}
|
|
else {
|
|
if (RTEST(rb_funcall(num, '<', 1, INT2FIX(0)))) return INT2FIX(0);
|
|
}
|
|
return num;
|
|
}
|
|
|
|
static VALUE
|
|
int_dotimes(VALUE num)
|
|
{
|
|
RETURN_SIZED_ENUMERATOR(num, 0, 0, int_dotimes_size);
|
|
|
|
if (FIXNUM_P(num)) {
|
|
long i, end;
|
|
|
|
end = FIX2LONG(num);
|
|
for (i=0; i<end; i++) {
|
|
rb_yield_1(LONG2FIX(i));
|
|
}
|
|
}
|
|
else {
|
|
VALUE i = INT2FIX(0);
|
|
|
|
for (;;) {
|
|
if (!RTEST(rb_funcall(i, '<', 1, num))) break;
|
|
rb_yield(i);
|
|
i = rb_funcall(i, '+', 1, INT2FIX(1));
|
|
}
|
|
}
|
|
return num;
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#round
|
|
* call-seq:
|
|
* int.round([ndigits] [, half: mode]) -> integer or float
|
|
*
|
|
* Returns +int+ rounded to the nearest value with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns +self+ when +ndigits+ is zero or positive.
|
|
*
|
|
* 1.round #=> 1
|
|
* 1.round(2) #=> 1
|
|
* 15.round(-1) #=> 20
|
|
* (-15).round(-1) #=> -20
|
|
*
|
|
* The optional +half+ keyword argument is available
|
|
* similar to Float#round.
|
|
*
|
|
* 25.round(-1, half: :up) #=> 30
|
|
* 25.round(-1, half: :down) #=> 20
|
|
* 25.round(-1, half: :even) #=> 20
|
|
* 35.round(-1, half: :up) #=> 40
|
|
* 35.round(-1, half: :down) #=> 30
|
|
* 35.round(-1, half: :even) #=> 40
|
|
* (-25).round(-1, half: :up) #=> -30
|
|
* (-25).round(-1, half: :down) #=> -20
|
|
* (-25).round(-1, half: :even) #=> -20
|
|
*/
|
|
|
|
static VALUE
|
|
int_round(int argc, VALUE* argv, VALUE num)
|
|
{
|
|
int ndigits;
|
|
int mode;
|
|
VALUE nd, opt;
|
|
|
|
if (!rb_scan_args(argc, argv, "01:", &nd, &opt)) return num;
|
|
ndigits = NUM2INT(nd);
|
|
mode = rb_num_get_rounding_option(opt);
|
|
if (ndigits >= 0) {
|
|
return num;
|
|
}
|
|
return rb_int_round(num, ndigits, mode);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#floor
|
|
* call-seq:
|
|
* int.floor([ndigits]) -> integer or float
|
|
*
|
|
* Returns the largest number less than or equal to +int+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns +self+ when +ndigits+ is zero or positive.
|
|
*
|
|
* 1.floor #=> 1
|
|
* 1.floor(2) #=> 1
|
|
* 18.floor(-1) #=> 10
|
|
* (-18).floor(-1) #=> -20
|
|
*/
|
|
|
|
static VALUE
|
|
int_floor(int argc, VALUE* argv, VALUE num)
|
|
{
|
|
int ndigits;
|
|
|
|
if (!rb_check_arity(argc, 0, 1)) return num;
|
|
ndigits = NUM2INT(argv[0]);
|
|
if (ndigits >= 0) {
|
|
return num;
|
|
}
|
|
return rb_int_floor(num, ndigits);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#ceil
|
|
* call-seq:
|
|
* int.ceil([ndigits]) -> integer or float
|
|
*
|
|
* Returns the smallest number greater than or equal to +int+ with
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns +self+ when +ndigits+ is zero or positive.
|
|
*
|
|
* 1.ceil #=> 1
|
|
* 1.ceil(2) #=> 1
|
|
* 18.ceil(-1) #=> 20
|
|
* (-18).ceil(-1) #=> -10
|
|
*/
|
|
|
|
static VALUE
|
|
int_ceil(int argc, VALUE* argv, VALUE num)
|
|
{
|
|
int ndigits;
|
|
|
|
if (!rb_check_arity(argc, 0, 1)) return num;
|
|
ndigits = NUM2INT(argv[0]);
|
|
if (ndigits >= 0) {
|
|
return num;
|
|
}
|
|
return rb_int_ceil(num, ndigits);
|
|
}
|
|
|
|
/*
|
|
* Document-method: Integer#truncate
|
|
* call-seq:
|
|
* int.truncate([ndigits]) -> integer or float
|
|
*
|
|
* Returns +int+ truncated (toward zero) to
|
|
* a precision of +ndigits+ decimal digits (default: 0).
|
|
*
|
|
* When the precision is negative, the returned value is an integer
|
|
* with at least <code>ndigits.abs</code> trailing zeros.
|
|
*
|
|
* Returns +self+ when +ndigits+ is zero or positive.
|
|
*
|
|
* 1.truncate #=> 1
|
|
* 1.truncate(2) #=> 1
|
|
* 18.truncate(-1) #=> 10
|
|
* (-18).truncate(-1) #=> -10
|
|
*/
|
|
|
|
static VALUE
|
|
int_truncate(int argc, VALUE* argv, VALUE num)
|
|
{
|
|
int ndigits;
|
|
|
|
if (!rb_check_arity(argc, 0, 1)) return num;
|
|
ndigits = NUM2INT(argv[0]);
|
|
if (ndigits >= 0) {
|
|
return num;
|
|
}
|
|
return rb_int_truncate(num, ndigits);
|
|
}
|
|
|
|
#define DEFINE_INT_SQRT(rettype, prefix, argtype) \
|
|
rettype \
|
|
prefix##_isqrt(argtype n) \
|
|
{ \
|
|
if (!argtype##_IN_DOUBLE_P(n)) { \
|
|
unsigned int b = bit_length(n); \
|
|
argtype t; \
|
|
rettype x = (rettype)(n >> (b/2+1)); \
|
|
x |= ((rettype)1LU << (b-1)/2); \
|
|
while ((t = n/x) < (argtype)x) x = (rettype)((x + t) >> 1); \
|
|
return x; \
|
|
} \
|
|
return (rettype)sqrt(argtype##_TO_DOUBLE(n)); \
|
|
}
|
|
|
|
#if SIZEOF_LONG*CHAR_BIT > DBL_MANT_DIG
|
|
# define RB_ULONG_IN_DOUBLE_P(n) ((n) < (1UL << DBL_MANT_DIG))
|
|
#else
|
|
# define RB_ULONG_IN_DOUBLE_P(n) 1
|
|
#endif
|
|
#define RB_ULONG_TO_DOUBLE(n) (double)(n)
|
|
#define RB_ULONG unsigned long
|
|
DEFINE_INT_SQRT(unsigned long, rb_ulong, RB_ULONG)
|
|
|
|
#if 2*SIZEOF_BDIGIT > SIZEOF_LONG
|
|
# if 2*SIZEOF_BDIGIT*CHAR_BIT > DBL_MANT_DIG
|
|
# define BDIGIT_DBL_IN_DOUBLE_P(n) ((n) < ((BDIGIT_DBL)1UL << DBL_MANT_DIG))
|
|
# else
|
|
# define BDIGIT_DBL_IN_DOUBLE_P(n) 1
|
|
# endif
|
|
# ifdef ULL_TO_DOUBLE
|
|
# define BDIGIT_DBL_TO_DOUBLE(n) ULL_TO_DOUBLE(n)
|
|
# else
|
|
# define BDIGIT_DBL_TO_DOUBLE(n) (double)(n)
|
|
# endif
|
|
DEFINE_INT_SQRT(BDIGIT, rb_bdigit_dbl, BDIGIT_DBL)
|
|
#endif
|
|
|
|
#define domain_error(msg) \
|
|
rb_raise(rb_eMathDomainError, "Numerical argument is out of domain - " #msg)
|
|
|
|
VALUE rb_big_isqrt(VALUE);
|
|
|
|
/*
|
|
* Document-method: Integer::sqrt
|
|
* call-seq:
|
|
* Integer.sqrt(n) -> integer
|
|
*
|
|
* Returns the integer square root of the non-negative integer +n+,
|
|
* i.e. the largest non-negative integer less than or equal to the
|
|
* square root of +n+.
|
|
*
|
|
* Integer.sqrt(0) #=> 0
|
|
* Integer.sqrt(1) #=> 1
|
|
* Integer.sqrt(24) #=> 4
|
|
* Integer.sqrt(25) #=> 5
|
|
* Integer.sqrt(10**400) #=> 10**200
|
|
*
|
|
* Equivalent to <code>Math.sqrt(n).floor</code>, except that
|
|
* the result of the latter code may differ from the true value
|
|
* due to the limited precision of floating point arithmetic.
|
|
*
|
|
* Integer.sqrt(10**46) #=> 100000000000000000000000
|
|
* Math.sqrt(10**46).floor #=> 99999999999999991611392 (!)
|
|
*
|
|
* If +n+ is not an Integer, it is converted to an Integer first.
|
|
* If +n+ is negative, a Math::DomainError is raised.
|
|
*/
|
|
|
|
static VALUE
|
|
rb_int_s_isqrt(VALUE self, VALUE num)
|
|
{
|
|
unsigned long n, sq;
|
|
num = rb_to_int(num);
|
|
if (FIXNUM_P(num)) {
|
|
if (FIXNUM_NEGATIVE_P(num)) {
|
|
domain_error("isqrt");
|
|
}
|
|
n = FIX2ULONG(num);
|
|
sq = rb_ulong_isqrt(n);
|
|
return LONG2FIX(sq);
|
|
}
|
|
else {
|
|
size_t biglen;
|
|
if (RBIGNUM_NEGATIVE_P(num)) {
|
|
domain_error("isqrt");
|
|
}
|
|
biglen = BIGNUM_LEN(num);
|
|
if (biglen == 0) return INT2FIX(0);
|
|
#if SIZEOF_BDIGIT <= SIZEOF_LONG
|
|
/* short-circuit */
|
|
if (biglen == 1) {
|
|
n = BIGNUM_DIGITS(num)[0];
|
|
sq = rb_ulong_isqrt(n);
|
|
return ULONG2NUM(sq);
|
|
}
|
|
#endif
|
|
return rb_big_isqrt(num);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Document-class: ZeroDivisionError
|
|
*
|
|
* Raised when attempting to divide an integer by 0.
|
|
*
|
|
* 42 / 0 #=> ZeroDivisionError: divided by 0
|
|
*
|
|
* Note that only division by an exact 0 will raise the exception:
|
|
*
|
|
* 42 / 0.0 #=> Float::INFINITY
|
|
* 42 / -0.0 #=> -Float::INFINITY
|
|
* 0 / 0.0 #=> NaN
|
|
*/
|
|
|
|
/*
|
|
* Document-class: FloatDomainError
|
|
*
|
|
* Raised when attempting to convert special float values (in particular
|
|
* +Infinity+ or +NaN+) to numerical classes which don't support them.
|
|
*
|
|
* Float::INFINITY.to_r #=> FloatDomainError: Infinity
|
|
*/
|
|
|
|
/*
|
|
* Document-class: Numeric
|
|
*
|
|
* Numeric is the class from which all higher-level numeric classes should inherit.
|
|
*
|
|
* Numeric allows instantiation of heap-allocated objects. Other core numeric classes such as
|
|
* Integer are implemented as immediates, which means that each Integer is a single immutable
|
|
* object which is always passed by value.
|
|
*
|
|
* a = 1
|
|
* 1.object_id == a.object_id #=> true
|
|
*
|
|
* There can only ever be one instance of the integer +1+, for example. Ruby ensures this
|
|
* by preventing instantiation and duplication.
|
|
*
|
|
* Integer.new(1) #=> NoMethodError: undefined method `new' for Integer:Class
|
|
* 1.dup #=> TypeError: can't dup Integer
|
|
*
|
|
* For this reason, Numeric should be used when defining other numeric classes.
|
|
*
|
|
* Classes which inherit from Numeric must implement +coerce+, which returns a two-member
|
|
* Array containing an object that has been coerced into an instance of the new class
|
|
* and +self+ (see #coerce).
|
|
*
|
|
* Inheriting classes should also implement arithmetic operator methods (<code>+</code>,
|
|
* <code>-</code>, <code>*</code> and <code>/</code>) and the <code><=></code> operator (see
|
|
* Comparable). These methods may rely on +coerce+ to ensure interoperability with
|
|
* instances of other numeric classes.
|
|
*
|
|
* class Tally < Numeric
|
|
* def initialize(string)
|
|
* @string = string
|
|
* end
|
|
*
|
|
* def to_s
|
|
* @string
|
|
* end
|
|
*
|
|
* def to_i
|
|
* @string.size
|
|
* end
|
|
*
|
|
* def coerce(other)
|
|
* [self.class.new('|' * other.to_i), self]
|
|
* end
|
|
*
|
|
* def <=>(other)
|
|
* to_i <=> other.to_i
|
|
* end
|
|
*
|
|
* def +(other)
|
|
* self.class.new('|' * (to_i + other.to_i))
|
|
* end
|
|
*
|
|
* def -(other)
|
|
* self.class.new('|' * (to_i - other.to_i))
|
|
* end
|
|
*
|
|
* def *(other)
|
|
* self.class.new('|' * (to_i * other.to_i))
|
|
* end
|
|
*
|
|
* def /(other)
|
|
* self.class.new('|' * (to_i / other.to_i))
|
|
* end
|
|
* end
|
|
*
|
|
* tally = Tally.new('||')
|
|
* puts tally * 2 #=> "||||"
|
|
* puts tally > 1 #=> true
|
|
*/
|
|
void
|
|
Init_Numeric(void)
|
|
{
|
|
#undef rb_intern
|
|
#define rb_intern(str) rb_intern_const(str)
|
|
|
|
#ifdef _UNICOSMP
|
|
/* Turn off floating point exceptions for divide by zero, etc. */
|
|
_set_Creg(0, 0);
|
|
#endif
|
|
id_coerce = rb_intern("coerce");
|
|
id_div = rb_intern("div");
|
|
id_divmod = rb_intern("divmod");
|
|
|
|
rb_eZeroDivError = rb_define_class("ZeroDivisionError", rb_eStandardError);
|
|
rb_eFloatDomainError = rb_define_class("FloatDomainError", rb_eRangeError);
|
|
rb_cNumeric = rb_define_class("Numeric", rb_cObject);
|
|
|
|
rb_define_method(rb_cNumeric, "singleton_method_added", num_sadded, 1);
|
|
rb_include_module(rb_cNumeric, rb_mComparable);
|
|
rb_define_method(rb_cNumeric, "coerce", num_coerce, 1);
|
|
rb_define_method(rb_cNumeric, "clone", num_clone, -1);
|
|
rb_define_method(rb_cNumeric, "dup", num_dup, 0);
|
|
|
|
rb_define_method(rb_cNumeric, "i", num_imaginary, 0);
|
|
rb_define_method(rb_cNumeric, "+@", num_uplus, 0);
|
|
rb_define_method(rb_cNumeric, "-@", num_uminus, 0);
|
|
rb_define_method(rb_cNumeric, "<=>", num_cmp, 1);
|
|
rb_define_method(rb_cNumeric, "eql?", num_eql, 1);
|
|
rb_define_method(rb_cNumeric, "fdiv", num_fdiv, 1);
|
|
rb_define_method(rb_cNumeric, "div", num_div, 1);
|
|
rb_define_method(rb_cNumeric, "divmod", num_divmod, 1);
|
|
rb_define_method(rb_cNumeric, "%", num_modulo, 1);
|
|
rb_define_method(rb_cNumeric, "modulo", num_modulo, 1);
|
|
rb_define_method(rb_cNumeric, "remainder", num_remainder, 1);
|
|
rb_define_method(rb_cNumeric, "abs", num_abs, 0);
|
|
rb_define_method(rb_cNumeric, "magnitude", num_abs, 0);
|
|
rb_define_method(rb_cNumeric, "to_int", num_to_int, 0);
|
|
|
|
rb_define_method(rb_cNumeric, "real?", num_real_p, 0);
|
|
rb_define_method(rb_cNumeric, "integer?", num_int_p, 0);
|
|
rb_define_method(rb_cNumeric, "zero?", num_zero_p, 0);
|
|
rb_define_method(rb_cNumeric, "nonzero?", num_nonzero_p, 0);
|
|
rb_define_method(rb_cNumeric, "finite?", num_finite_p, 0);
|
|
rb_define_method(rb_cNumeric, "infinite?", num_infinite_p, 0);
|
|
|
|
rb_define_method(rb_cNumeric, "floor", num_floor, -1);
|
|
rb_define_method(rb_cNumeric, "ceil", num_ceil, -1);
|
|
rb_define_method(rb_cNumeric, "round", num_round, -1);
|
|
rb_define_method(rb_cNumeric, "truncate", num_truncate, -1);
|
|
rb_define_method(rb_cNumeric, "step", num_step, -1);
|
|
rb_define_method(rb_cNumeric, "positive?", num_positive_p, 0);
|
|
rb_define_method(rb_cNumeric, "negative?", num_negative_p, 0);
|
|
|
|
rb_cInteger = rb_define_class("Integer", rb_cNumeric);
|
|
rb_undef_alloc_func(rb_cInteger);
|
|
rb_undef_method(CLASS_OF(rb_cInteger), "new");
|
|
rb_define_singleton_method(rb_cInteger, "sqrt", rb_int_s_isqrt, 1);
|
|
|
|
rb_define_method(rb_cInteger, "to_s", int_to_s, -1);
|
|
rb_define_alias(rb_cInteger, "inspect", "to_s");
|
|
rb_define_method(rb_cInteger, "integer?", int_int_p, 0);
|
|
rb_define_method(rb_cInteger, "odd?", rb_int_odd_p, 0);
|
|
rb_define_method(rb_cInteger, "even?", int_even_p, 0);
|
|
rb_define_method(rb_cInteger, "allbits?", int_allbits_p, 1);
|
|
rb_define_method(rb_cInteger, "anybits?", int_anybits_p, 1);
|
|
rb_define_method(rb_cInteger, "nobits?", int_nobits_p, 1);
|
|
rb_define_method(rb_cInteger, "upto", int_upto, 1);
|
|
rb_define_method(rb_cInteger, "downto", int_downto, 1);
|
|
rb_define_method(rb_cInteger, "times", int_dotimes, 0);
|
|
rb_define_method(rb_cInteger, "succ", int_succ, 0);
|
|
rb_define_method(rb_cInteger, "next", int_succ, 0);
|
|
rb_define_method(rb_cInteger, "pred", int_pred, 0);
|
|
rb_define_method(rb_cInteger, "chr", int_chr, -1);
|
|
rb_define_method(rb_cInteger, "ord", int_ord, 0);
|
|
rb_define_method(rb_cInteger, "to_i", int_to_i, 0);
|
|
rb_define_method(rb_cInteger, "to_int", int_to_i, 0);
|
|
rb_define_method(rb_cInteger, "to_f", int_to_f, 0);
|
|
rb_define_method(rb_cInteger, "floor", int_floor, -1);
|
|
rb_define_method(rb_cInteger, "ceil", int_ceil, -1);
|
|
rb_define_method(rb_cInteger, "truncate", int_truncate, -1);
|
|
rb_define_method(rb_cInteger, "round", int_round, -1);
|
|
rb_define_method(rb_cInteger, "<=>", rb_int_cmp, 1);
|
|
|
|
rb_define_method(rb_cInteger, "-@", rb_int_uminus, 0);
|
|
rb_define_method(rb_cInteger, "+", rb_int_plus, 1);
|
|
rb_define_method(rb_cInteger, "-", rb_int_minus, 1);
|
|
rb_define_method(rb_cInteger, "*", rb_int_mul, 1);
|
|
rb_define_method(rb_cInteger, "/", rb_int_div, 1);
|
|
rb_define_method(rb_cInteger, "div", rb_int_idiv, 1);
|
|
rb_define_method(rb_cInteger, "%", rb_int_modulo, 1);
|
|
rb_define_method(rb_cInteger, "modulo", rb_int_modulo, 1);
|
|
rb_define_method(rb_cInteger, "remainder", int_remainder, 1);
|
|
rb_define_method(rb_cInteger, "divmod", rb_int_divmod, 1);
|
|
rb_define_method(rb_cInteger, "fdiv", rb_int_fdiv, 1);
|
|
rb_define_method(rb_cInteger, "**", rb_int_pow, 1);
|
|
|
|
rb_define_method(rb_cInteger, "pow", rb_int_powm, -1); /* in bignum.c */
|
|
|
|
rb_define_method(rb_cInteger, "abs", rb_int_abs, 0);
|
|
rb_define_method(rb_cInteger, "magnitude", rb_int_abs, 0);
|
|
|
|
rb_define_method(rb_cInteger, "===", rb_int_equal, 1);
|
|
rb_define_method(rb_cInteger, "==", rb_int_equal, 1);
|
|
rb_define_method(rb_cInteger, ">", rb_int_gt, 1);
|
|
rb_define_method(rb_cInteger, ">=", rb_int_ge, 1);
|
|
rb_define_method(rb_cInteger, "<", int_lt, 1);
|
|
rb_define_method(rb_cInteger, "<=", int_le, 1);
|
|
|
|
rb_define_method(rb_cInteger, "~", int_comp, 0);
|
|
rb_define_method(rb_cInteger, "&", rb_int_and, 1);
|
|
rb_define_method(rb_cInteger, "|", int_or, 1);
|
|
rb_define_method(rb_cInteger, "^", int_xor, 1);
|
|
rb_define_method(rb_cInteger, "[]", int_aref, 1);
|
|
|
|
rb_define_method(rb_cInteger, "<<", rb_int_lshift, 1);
|
|
rb_define_method(rb_cInteger, ">>", rb_int_rshift, 1);
|
|
|
|
rb_define_method(rb_cInteger, "size", int_size, 0);
|
|
rb_define_method(rb_cInteger, "bit_length", rb_int_bit_length, 0);
|
|
rb_define_method(rb_cInteger, "digits", rb_int_digits, -1);
|
|
|
|
#ifndef RUBY_INTEGER_UNIFICATION
|
|
rb_cFixnum = rb_cInteger;
|
|
#endif
|
|
rb_define_const(rb_cObject, "Fixnum", rb_cInteger);
|
|
rb_deprecate_constant(rb_cObject, "Fixnum");
|
|
|
|
rb_cFloat = rb_define_class("Float", rb_cNumeric);
|
|
|
|
rb_undef_alloc_func(rb_cFloat);
|
|
rb_undef_method(CLASS_OF(rb_cFloat), "new");
|
|
|
|
/*
|
|
* Represents the rounding mode for floating point addition.
|
|
*
|
|
* Usually defaults to 1, rounding to the nearest number.
|
|
*
|
|
* Other modes include:
|
|
*
|
|
* -1:: Indeterminable
|
|
* 0:: Rounding towards zero
|
|
* 1:: Rounding to the nearest number
|
|
* 2:: Rounding towards positive infinity
|
|
* 3:: Rounding towards negative infinity
|
|
*/
|
|
rb_define_const(rb_cFloat, "ROUNDS", INT2FIX(FLT_ROUNDS));
|
|
/*
|
|
* The base of the floating point, or number of unique digits used to
|
|
* represent the number.
|
|
*
|
|
* Usually defaults to 2 on most systems, which would represent a base-10 decimal.
|
|
*/
|
|
rb_define_const(rb_cFloat, "RADIX", INT2FIX(FLT_RADIX));
|
|
/*
|
|
* The number of base digits for the +double+ data type.
|
|
*
|
|
* Usually defaults to 53.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MANT_DIG", INT2FIX(DBL_MANT_DIG));
|
|
/*
|
|
* The minimum number of significant decimal digits in a double-precision
|
|
* floating point.
|
|
*
|
|
* Usually defaults to 15.
|
|
*/
|
|
rb_define_const(rb_cFloat, "DIG", INT2FIX(DBL_DIG));
|
|
/*
|
|
* The smallest possible exponent value in a double-precision floating
|
|
* point.
|
|
*
|
|
* Usually defaults to -1021.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MIN_EXP", INT2FIX(DBL_MIN_EXP));
|
|
/*
|
|
* The largest possible exponent value in a double-precision floating
|
|
* point.
|
|
*
|
|
* Usually defaults to 1024.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MAX_EXP", INT2FIX(DBL_MAX_EXP));
|
|
/*
|
|
* The smallest negative exponent in a double-precision floating point
|
|
* where 10 raised to this power minus 1.
|
|
*
|
|
* Usually defaults to -307.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MIN_10_EXP", INT2FIX(DBL_MIN_10_EXP));
|
|
/*
|
|
* The largest positive exponent in a double-precision floating point where
|
|
* 10 raised to this power minus 1.
|
|
*
|
|
* Usually defaults to 308.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MAX_10_EXP", INT2FIX(DBL_MAX_10_EXP));
|
|
/*
|
|
* The smallest positive normalized number in a double-precision floating point.
|
|
*
|
|
* Usually defaults to 2.2250738585072014e-308.
|
|
*
|
|
* If the platform supports denormalized numbers,
|
|
* there are numbers between zero and Float::MIN.
|
|
* 0.0.next_float returns the smallest positive floating point number
|
|
* including denormalized numbers.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MIN", DBL2NUM(DBL_MIN));
|
|
/*
|
|
* The largest possible integer in a double-precision floating point number.
|
|
*
|
|
* Usually defaults to 1.7976931348623157e+308.
|
|
*/
|
|
rb_define_const(rb_cFloat, "MAX", DBL2NUM(DBL_MAX));
|
|
/*
|
|
* The difference between 1 and the smallest double-precision floating
|
|
* point number greater than 1.
|
|
*
|
|
* Usually defaults to 2.2204460492503131e-16.
|
|
*/
|
|
rb_define_const(rb_cFloat, "EPSILON", DBL2NUM(DBL_EPSILON));
|
|
/*
|
|
* An expression representing positive infinity.
|
|
*/
|
|
rb_define_const(rb_cFloat, "INFINITY", DBL2NUM(INFINITY));
|
|
/*
|
|
* An expression representing a value which is "not a number".
|
|
*/
|
|
rb_define_const(rb_cFloat, "NAN", DBL2NUM(NAN));
|
|
|
|
rb_define_method(rb_cFloat, "to_s", flo_to_s, 0);
|
|
rb_define_alias(rb_cFloat, "inspect", "to_s");
|
|
rb_define_method(rb_cFloat, "coerce", flo_coerce, 1);
|
|
rb_define_method(rb_cFloat, "-@", rb_float_uminus, 0);
|
|
rb_define_method(rb_cFloat, "+", flo_plus, 1);
|
|
rb_define_method(rb_cFloat, "-", flo_minus, 1);
|
|
rb_define_method(rb_cFloat, "*", flo_mul, 1);
|
|
rb_define_method(rb_cFloat, "/", flo_div, 1);
|
|
rb_define_method(rb_cFloat, "quo", flo_quo, 1);
|
|
rb_define_method(rb_cFloat, "fdiv", flo_quo, 1);
|
|
rb_define_method(rb_cFloat, "%", flo_mod, 1);
|
|
rb_define_method(rb_cFloat, "modulo", flo_mod, 1);
|
|
rb_define_method(rb_cFloat, "divmod", flo_divmod, 1);
|
|
rb_define_method(rb_cFloat, "**", rb_float_pow, 1);
|
|
rb_define_method(rb_cFloat, "==", flo_eq, 1);
|
|
rb_define_method(rb_cFloat, "===", flo_eq, 1);
|
|
rb_define_method(rb_cFloat, "<=>", flo_cmp, 1);
|
|
rb_define_method(rb_cFloat, ">", rb_float_gt, 1);
|
|
rb_define_method(rb_cFloat, ">=", flo_ge, 1);
|
|
rb_define_method(rb_cFloat, "<", flo_lt, 1);
|
|
rb_define_method(rb_cFloat, "<=", flo_le, 1);
|
|
rb_define_method(rb_cFloat, "eql?", flo_eql, 1);
|
|
rb_define_method(rb_cFloat, "hash", flo_hash, 0);
|
|
rb_define_method(rb_cFloat, "to_f", flo_to_f, 0);
|
|
rb_define_method(rb_cFloat, "abs", rb_float_abs, 0);
|
|
rb_define_method(rb_cFloat, "magnitude", rb_float_abs, 0);
|
|
rb_define_method(rb_cFloat, "zero?", flo_zero_p, 0);
|
|
|
|
rb_define_method(rb_cFloat, "to_i", flo_to_i, 0);
|
|
rb_define_method(rb_cFloat, "to_int", flo_to_i, 0);
|
|
rb_define_method(rb_cFloat, "floor", flo_floor, -1);
|
|
rb_define_method(rb_cFloat, "ceil", flo_ceil, -1);
|
|
rb_define_method(rb_cFloat, "round", flo_round, -1);
|
|
rb_define_method(rb_cFloat, "truncate", flo_truncate, -1);
|
|
|
|
rb_define_method(rb_cFloat, "nan?", flo_is_nan_p, 0);
|
|
rb_define_method(rb_cFloat, "infinite?", rb_flo_is_infinite_p, 0);
|
|
rb_define_method(rb_cFloat, "finite?", rb_flo_is_finite_p, 0);
|
|
rb_define_method(rb_cFloat, "next_float", flo_next_float, 0);
|
|
rb_define_method(rb_cFloat, "prev_float", flo_prev_float, 0);
|
|
rb_define_method(rb_cFloat, "positive?", flo_positive_p, 0);
|
|
rb_define_method(rb_cFloat, "negative?", flo_negative_p, 0);
|
|
|
|
id_to = rb_intern("to");
|
|
id_by = rb_intern("by");
|
|
}
|
|
|
|
#undef rb_float_value
|
|
double
|
|
rb_float_value(VALUE v)
|
|
{
|
|
return rb_float_value_inline(v);
|
|
}
|
|
|
|
#undef rb_float_new
|
|
VALUE
|
|
rb_float_new(double d)
|
|
{
|
|
return rb_float_new_inline(d);
|
|
}
|