Numeric
BigDecimal provides arbitrary-precision floating point decimal arithmetic.
Copyright (C) 2002 by Shigeo Kobayashi <shigeo@tinyforest.gr.jp>. You may distribute under the terms of either the GNU General Public License or the Artistic License, as specified in the README file of the BigDecimal distribution.
Documented by mathew <meta@pobox.com>.
Ruby provides built-in support for arbitrary precision integer arithmetic. For example:
42**13 -> 1265437718438866624512
BigDecimal provides similar support for very large or very accurate floating point numbers.
Decimal arithmetic is also useful for general calculation, because it provides the correct answers people expect–whereas normal binary floating point arithmetic often introduces subtle errors because of the conversion between base 10 and base 2. For example, try:
sum = 0 for i in (1..10000) sum = sum + 0.0001 end print sum
and contrast with the output from:
require 'bigdecimal' sum = BigDecimal.new("0") for i in (1..10000) sum = sum + BigDecimal.new("0.0001") end print sum
Similarly:
(BigDecimal.new(“1.2”) - BigDecimal(“1.0”)) == BigDecimal(“0.2”) -> true
(1.2 - 1.0) == 0.2 -> false
Because BigDecimal is more accurate than normal binary floating point arithmetic, it requires some special values.
BigDecimal sometimes needs to return infinity, for example if you divide a value by zero.
::new(“1.0”) / ::new(“0.0”) -> infinity
::new(“-1.0”) / ::new(“0.0”) -> -infinity
You can represent infinite numbers to BigDecimal using the strings 'Infinity', '+Infinity' and '-Infinity' (case-sensitive)
When a computation results in an undefined value, the special value NaN (for 'not a number') is returned.
Example:
::new(“0.0”) / ::new(“0.0”) -> NaN
You can also create undefined values. NaN is never considered to be the same as any other value, even NaN itself:
n = ::new('NaN')
n == 0.0 -> nil
n == n -> nil
If a computation results in a value which is too small to be represented as a BigDecimal within the currently specified limits of precision, zero must be returned.
If the value which is too small to be represented is negative, a BigDecimal value of negative zero is returned. If the value is positive, a value of positive zero is returned.
::new(“1.0”) / ::new(“-Infinity”) -> -0.0
::new(“1.0”) / ::new(“Infinity”) -> 0.0
(See ::mode for how to specify limits of precision.)
Note that -0.0 and 0.0 are considered to be the same for the purposes of comparison.
Note also that in mathematics, there is no particular concept of negative or positive zero; true mathematical zero has no sign.
Base value used in internal calculations. On a 32 bit system, BASE is 10000, indicating that calculation is done in groups of 4 digits.
(If it were larger, BASE**2 wouldn't fit in 32 bits, so you couldn't guarantee that two groups could always be multiplied together without overflow.)
Determines whether overflow, underflow or zero divide result in an exception being thrown. See ::mode.
Determines what happens when the result of a computation is infinity. See ::mode.
Determines what happens when the result of a computation is not a number (NaN). See ::mode.
Determines what happens when the result of a computation is an overflow (a result too large to be represented). See ::mode.
Determines what happens when the result of a computation is an underflow (a result too small to be represented). See ::mode.
Determines what happens when a division by zero is performed. See ::mode.
Round towards +infinity. See ::mode.
Indicates that values should be rounded towards zero. See ::mode.
Round towards -infinity. See ::mode.
Indicates that digits >= 6 should be rounded up, others rounded down. See ::mode.
Round towards the even neighbor. See ::mode.
Indicates that digits >= 5 should be rounded up, others rounded down. See ::mode.
Determines what happens when a result must be rounded in order to fit in the appropriate number of significant digits. See ::mode.
Indicates that values should be rounded away from zero. See ::mode.
Indicates that a value is negative and finite. See #sign.
Indicates that a value is negative and infinite. See #sign.
Indicates that a value is -0. See #sign.
Indicates that a value is not a number. See #sign.
Indicates that a value is positive and finite. See #sign.
Indicates that a value is positive and infinite. See #sign.
Indicates that a value is +0. See #sign.
Internal method used to provide marshalling support. See the Marshal module.
static VALUE BigDecimal_load(VALUE self, VALUE str) { ENTER(2); Real *pv; unsigned char *pch; unsigned char ch; unsigned long m=0; SafeStringValue(str); pch = (unsigned char *)RSTRING_PTR(str); /* First get max prec */ while((*pch)!=(unsigned char)'\0' && (ch=*pch++)!=(unsigned char)':') { if(!ISDIGIT(ch)) { rb_raise(rb_eTypeError, "load failed: invalid character in the marshaled string"); } m = m*10 + (unsigned long)(ch-'0'); } if(m>VpBaseFig()) m -= VpBaseFig(); GUARD_OBJ(pv,VpNewRbClass(m,(char *)pch,self)); m /= VpBaseFig(); if(m && pv->MaxPrec>m) pv->MaxPrec = m+1; return ToValue(pv); }
The ::double_fig class method returns the number of digits a Float number is allowed to have. The result depends upon the CPU and OS in use.
static VALUE BigDecimal_double_fig(VALUE self) { return INT2FIX(VpDblFig()); }
Limit the number of significant digits in newly created BigDecimal numbers to the specified value. Rounding is performed as necessary, as specified by ::mode.
A limit of 0, the default, means no upper limit.
The limit specified by this method takes less priority over any limit specified to instance methods such as ceil, floor, truncate, or round.
static VALUE BigDecimal_limit(int argc, VALUE *argv, VALUE self) { VALUE nFig; VALUE nCur = INT2NUM(VpGetPrecLimit()); if(rb_scan_args(argc,argv,"01",&nFig)==1) { int nf; if(nFig==Qnil) return nCur; Check_Type(nFig, T_FIXNUM); nf = FIX2INT(nFig); if(nf<0) { rb_raise(rb_eArgError, "argument must be positive"); } VpSetPrecLimit(nf); } return nCur; }
Controls handling of arithmetic exceptions and rounding. If no value is supplied, the current value is returned.
Six values of the mode parameter control the handling of arithmetic exceptions:
BigDecimal::EXCEPTION_NaN BigDecimal::EXCEPTION_INFINITY BigDecimal::EXCEPTION_UNDERFLOW BigDecimal::EXCEPTION_OVERFLOW BigDecimal::EXCEPTION_ZERODIVIDE BigDecimal::EXCEPTION_ALL
For each mode parameter above, if the value set is false, computation continues after an arithmetic exception of the appropriate type. When computation continues, results are as follows:
NaN
+infinity or -infinity
0
+infinity or -infinity
+infinity or -infinity
One value of the mode parameter controls the rounding of numeric values: BigDecimal::ROUND_MODE. The values it can take are:
round away from zero
round towards zero (truncate)
round up if the appropriate digit >= 5, otherwise truncate (default)
round up if the appropriate digit >= 6, otherwise truncate
round towards the even neighbor (Banker's rounding)
round towards positive infinity (ceil)
round towards negative infinity (floor)
static VALUE BigDecimal_mode(int argc, VALUE *argv, VALUE self) { VALUE which; VALUE val; unsigned long f,fo; if(rb_scan_args(argc,argv,"11",&which,&val)==1) val = Qnil; Check_Type(which, T_FIXNUM); f = (unsigned long)FIX2INT(which); if(f&VP_EXCEPTION_ALL) { /* Exception mode setting */ fo = VpGetException(); if(val==Qnil) return INT2FIX(fo); if(val!=Qfalse && val!=Qtrue) { rb_raise(rb_eTypeError, "second argument must be true or false"); return Qnil; /* Not reached */ } if(f&VP_EXCEPTION_INFINITY) { VpSetException((unsigned short)((val==Qtrue)?(fo|VP_EXCEPTION_INFINITY): (fo&(~VP_EXCEPTION_INFINITY)))); } fo = VpGetException(); if(f&VP_EXCEPTION_NaN) { VpSetException((unsigned short)((val==Qtrue)?(fo|VP_EXCEPTION_NaN): (fo&(~VP_EXCEPTION_NaN)))); } fo = VpGetException(); if(f&VP_EXCEPTION_UNDERFLOW) { VpSetException((unsigned short)((val==Qtrue)?(fo|VP_EXCEPTION_UNDERFLOW): (fo&(~VP_EXCEPTION_UNDERFLOW)))); } fo = VpGetException(); if(f&VP_EXCEPTION_ZERODIVIDE) { VpSetException((unsigned short)((val==Qtrue)?(fo|VP_EXCEPTION_ZERODIVIDE): (fo&(~VP_EXCEPTION_ZERODIVIDE)))); } fo = VpGetException(); return INT2FIX(fo); } if(VP_ROUND_MODE==f) { /* Rounding mode setting */ fo = VpGetRoundMode(); if(val==Qnil) return INT2FIX(fo); Check_Type(val, T_FIXNUM); if(!VpIsRoundMode(FIX2INT(val))) { rb_raise(rb_eTypeError, "invalid rounding mode"); return Qnil; } fo = VpSetRoundMode((unsigned long)FIX2INT(val)); return INT2FIX(fo); } rb_raise(rb_eTypeError, "first argument for BigDecimal#mode invalid"); return Qnil; }
Create a new BigDecimal object.
The initial value, as a String. Spaces are ignored, unrecognized characters terminate the value.
The number of significant digits, as a Fixnum. If omitted or 0, the number of significant digits is determined from the initial value.
The actual number of significant digits used in computation is usually larger than the specified number.
static VALUE BigDecimal_new(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *pv; S_LONG mf; VALUE nFig; VALUE iniValue; if(rb_scan_args(argc,argv,"11",&iniValue,&nFig)==1) { mf = 0; } else { mf = GetPositiveInt(nFig); } SafeStringValue(iniValue); GUARD_OBJ(pv,VpNewRbClass(mf, RSTRING_PTR(iniValue),self)); return ToValue(pv); }
Returns the BigDecimal version number.
Ruby 1.8.0 returns 1.0.0. Ruby 1.8.1 thru 1.8.3 return 1.0.1.
static VALUE BigDecimal_version(VALUE self) { /* * 1.0.0: Ruby 1.8.0 * 1.0.1: Ruby 1.8.1 */ return rb_str_new2("1.0.1"); }
Returns the modulus from dividing by b. See divmod.
static VALUE BigDecimal_mod(VALUE self, VALUE r)
Multiply by the specified value.
e.g.
c = a.mult(b,n) c = a * b
If specified and less than the number of significant digits of the result, the result is rounded to that number of digits, according to ::mode.
static VALUE BigDecimal_mult(VALUE self, VALUE r) { ENTER(5); Real *c, *a, *b; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); b = GetVpValue(r,0); if(!b) return DoSomeOne(self,r,'*'); SAVE(b); mx = a->Prec + b->Prec; GUARD_OBJ(c,VpCreateRbObject(mx *(VpBaseFig() + 1), "0")); VpMult(c, a, b); return ToValue(c); }
Returns the value raised to the power of n. Note that n must be an Integer.
Also available as the operator **
static VALUE BigDecimal_power(VALUE self, VALUE p) { ENTER(5); Real *x, *y; S_LONG mp, ma, n; Check_Type(p, T_FIXNUM); n = FIX2INT(p); ma = n; if(ma < 0) ma = -ma; if(ma == 0) ma = 1; GUARD_OBJ(x,GetVpValue(self,1)); if(VpIsDef(x)) { mp = x->Prec *(VpBaseFig() + 1); GUARD_OBJ(y,VpCreateRbObject(mp *(ma + 1), "0")); } else { GUARD_OBJ(y,VpCreateRbObject(1, "0")); } VpPower(y, x, n); return ToValue(y); }
Add the specified value.
e.g.
c = a.add(b,n) c = a + b
If specified and less than the number of significant digits of the result, the result is rounded to that number of digits, according to ::mode.
static VALUE BigDecimal_add(VALUE self, VALUE r) { ENTER(5); Real *c, *a, *b; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); b = GetVpValue(r,0); if(!b) return DoSomeOne(self,r,'+'); SAVE(b); if(VpIsNaN(b)) return b->obj; if(VpIsNaN(a)) return a->obj; mx = GetAddSubPrec(a,b); if(mx==(-1L)) { GUARD_OBJ(c,VpCreateRbObject(VpBaseFig() + 1, "0")); VpAddSub(c, a, b, 1); } else { GUARD_OBJ(c,VpCreateRbObject(mx *(VpBaseFig() + 1), "0")); if(!mx) { VpSetInf(c,VpGetSign(a)); } else { VpAddSub(c, a, b, 1); } } return ToValue(c); }
Subtract the specified value.
e.g.
c = a.sub(b,n) c = a - b
If specified and less than the number of significant digits of the result, the result is rounded to that number of digits, according to ::mode.
static VALUE BigDecimal_sub(VALUE self, VALUE r) { ENTER(5); Real *c, *a, *b; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); b = GetVpValue(r,0); if(!b) return DoSomeOne(self,r,'-'); SAVE(b); if(VpIsNaN(b)) return b->obj; if(VpIsNaN(a)) return a->obj; mx = GetAddSubPrec(a,b); if(mx==(-1L)) { GUARD_OBJ(c,VpCreateRbObject(VpBaseFig() + 1, "0")); VpAddSub(c, a, b, -1); } else { GUARD_OBJ(c,VpCreateRbObject(mx *(VpBaseFig() + 1), "0")); if(!mx) { VpSetInf(c,VpGetSign(a)); } else { VpAddSub(c, a, b, -1); } } return ToValue(c); }
static VALUE BigDecimal_neg(VALUE self) { ENTER(5); Real *c, *a; GUARD_OBJ(a,GetVpValue(self,1)); GUARD_OBJ(c,VpCreateRbObject(a->Prec *(VpBaseFig() + 1), "0")); VpAsgn(c, a, -1); return ToValue(c); }
Divide by the specified value.
e.g.
c = a.div(b,n)
If specified and less than the number of significant digits of the result, the result is rounded to that number of digits, according to ::mode.
If digits is 0, the result is the same as the / operator. If not, the result is an integer BigDecimal, by analogy with Float#div.
The alias quo is provided since div(value, 0) is the same as computing the quotient; see divmod.
static VALUE BigDecimal_div(VALUE self, VALUE r) /* For c = self/r: with round operation */ { ENTER(5); Real *c=NULL, *res=NULL, *div = NULL; r = BigDecimal_divide(&c, &res, &div, self, r); if(r!=(VALUE)0) return r; /* coerced by other */ SAVE(c);SAVE(res);SAVE(div); /* a/b = c + r/b */ /* c xxxxx r 00000yyyyy ==> (y/b)*BASE >= HALF_BASE */ /* Round */ if(VpHasVal(div)) { /* frac[0] must be zero for NaN,INF,Zero */ VpInternalRound(c,0,c->frac[c->Prec-1],(VpBaseVal()*res->frac[0])/div->frac[0]); } return ToValue(c); }
Returns true if a is less than b. Values may be coerced to perform the comparison (see ==, coerce).
static VALUE BigDecimal_lt(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '<'); }
Returns true if a is less than or equal to b. Values may be coerced to perform the comparison (see ==, coerce).
static VALUE BigDecimal_le(VALUE self, VALUE r) { return BigDecimalCmp(self, r, 'L'); }
The comparison operator. a <=> b is 0 if a == b, 1 if a > b, -1 if a < b.
static VALUE BigDecimal_comp(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '*'); }
Tests for value equality; returns true if the values are equal.
The == and === operators and the eql? method have the same implementation for BigDecimal.
Values may be coerced to perform the comparison:
::new('1.0') == 1.0 -> true
static VALUE BigDecimal_eq(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '='); }
Tests for value equality; returns true if the values are equal.
The == and === operators and the eql? method have the same implementation for BigDecimal.
Values may be coerced to perform the comparison:
::new('1.0') == 1.0 -> true
static VALUE BigDecimal_eq(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '='); }
Returns true if a is greater than b. Values may be coerced to perform the comparison (see ==, coerce).
static VALUE BigDecimal_gt(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '>'); }
Returns true if a is greater than or equal to b. Values may be coerced to perform the comparison (see ==, coerce)
static VALUE BigDecimal_ge(VALUE self, VALUE r) { return BigDecimalCmp(self, r, 'G'); }
static VALUE BigDecimal_dump(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *vp; char *psz; VALUE dummy; volatile VALUE dump; rb_scan_args(argc, argv, "01", &dummy); GUARD_OBJ(vp,GetVpValue(self,1)); dump = rb_str_new(0,VpNumOfChars(vp,"E")+50); psz = RSTRING_PTR(dump); sprintf(psz,"%lu:",VpMaxPrec(vp)*VpBaseFig()); VpToString(vp, psz+strlen(psz), 0, 0); rb_str_resize(dump, strlen(psz)); return dump; }
Returns the absolute value.
BigDecimal('5').abs -> 5
BigDecimal('-3').abs -> 3
static VALUE BigDecimal_abs(VALUE self) { ENTER(5); Real *c, *a; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpAsgn(c, a, 1); VpChangeSign(c,(S_INT)1); return ToValue(c); }
static VALUE BigDecimal_add2(VALUE self, VALUE b, VALUE n) { ENTER(2); Real *cv; U_LONG mx = (U_LONG)GetPositiveInt(n); if(mx==0) return BigDecimal_add(self,b); else { U_LONG pl = VpSetPrecLimit(0); VALUE c = BigDecimal_add(self,b); VpSetPrecLimit(pl); GUARD_OBJ(cv,GetVpValue(c,1)); VpLeftRound(cv,VpGetRoundMode(),mx); return ToValue(cv); } }
Return the smallest integer greater than or equal to the value, as a BigDecimal.
BigDecimal('3.14159').ceil -> 4
BigDecimal('-9.1').ceil -> -9
If n is specified and positive, the fractional part of the result has no more than that many digits.
If n is specified and negative, at least that many digits to the left of the decimal point will be 0 in the result.
BigDecimal('3.14159').ceil(3) -> 3.142
BigDecimal('13345.234').ceil(-2) -> 13400.0
static VALUE BigDecimal_ceil(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *c, *a; U_LONG mx; int iLoc; VALUE vLoc; U_LONG pl = VpSetPrecLimit(0); if(rb_scan_args(argc,argv,"01",&vLoc)==0) { iLoc = 0; } else { Check_Type(vLoc, T_FIXNUM); iLoc = FIX2INT(vLoc); } GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpSetPrecLimit(pl); VpActiveRound(c,a,VP_ROUND_CEIL,iLoc); return ToValue(c); }
The coerce method provides support for Ruby type coercion. It is not enabled by default.
This means that binary operations like + * / or - can often be performed on a BigDecimal and an object of another type, if the other object can be coerced into a BigDecimal value.
e.g. a = ::new(“1.0”) b = a / 2.0 -> 0.5
Note that coercing a String to a BigDecimal is not supported by default; it requires a special compile-time option when building Ruby.
static VALUE BigDecimal_coerce(VALUE self, VALUE other) { ENTER(2); VALUE obj; Real *b; if(TYPE(other) == T_FLOAT) { obj = rb_assoc_new(other, BigDecimal_to_f(self)); } else { GUARD_OBJ(b,GetVpValue(other,1)); obj = rb_assoc_new(b->obj, self); } return obj; }
static VALUE BigDecimal_div2(int argc, VALUE *argv, VALUE self) { ENTER(5); VALUE b,n; int na = rb_scan_args(argc,argv,"11",&b,&n); if(na==1) { /* div in Float sense */ VALUE obj; Real *div=NULL; Real *mod; obj = BigDecimal_DoDivmod(self,b,&div,&mod); if(obj!=(VALUE)0) return obj; if(VpIsNaN(div) && rb_equal(b, INT2FIX(0))) { rb_raise(rb_eZeroDivError, "divided by 0"); } return BigDecimal_to_i(ToValue(div)); } else { /* div in BigDecimal sense */ U_LONG ix = (U_LONG)GetPositiveInt(n); if(ix==0) return BigDecimal_div(self,b); else { Real *res=NULL; Real *av=NULL, *bv=NULL, *cv=NULL; U_LONG mx = (ix+VpBaseFig()*2); U_LONG pl = VpSetPrecLimit(0); GUARD_OBJ(cv,VpCreateRbObject(mx,"0")); GUARD_OBJ(av,GetVpValue(self,1)); GUARD_OBJ(bv,GetVpValue(b,1)); mx = av->Prec + bv->Prec + 2; if(mx <= cv->MaxPrec) mx = cv->MaxPrec+1; GUARD_OBJ(res,VpCreateRbObject((mx * 2 + 2)*VpBaseFig(), "#0")); VpDivd(cv,res,av,bv); VpSetPrecLimit(pl); VpLeftRound(cv,VpGetRoundMode(),ix); return ToValue(cv); } } }
Divides by the specified value, and returns the quotient and modulus as BigDecimal numbers. The quotient is rounded towards negative infinity.
For example:
require 'bigdecimal'
a = ::new(“42”) b = ::new(“9”)
q,m = a.divmod(b)
c = q * b + m
a == c -> true
The quotient q is (a/b).floor, and the modulus is the amount that must be added to q * b to get a.
static VALUE BigDecimal_divmod(VALUE self, VALUE r) { ENTER(5); VALUE obj; Real *div=NULL, *mod=NULL; obj = BigDecimal_DoDivmod(self,r,&div,&mod); if(obj!=(VALUE)0) return obj; SAVE(div);SAVE(mod); obj = rb_assoc_new(ToValue(div), ToValue(mod)); return obj; }
Tests for value equality; returns true if the values are equal.
The == and === operators and the eql? method have the same implementation for BigDecimal.
Values may be coerced to perform the comparison:
::new('1.0') == 1.0 -> true
static VALUE BigDecimal_eq(VALUE self, VALUE r) { return BigDecimalCmp(self, r, '='); }
Returns the exponent of the BigDecimal number, as an Integer.
If the number can be represented as 0.xxxxxx*10**n where xxxxxx is a string of digits with no leading zeros, then n is the exponent.
static VALUE BigDecimal_exponent(VALUE self) { S_LONG e = VpExponent10(GetVpValue(self,1)); return INT2NUM(e); }
Returns True if the value is finite (not NaN or infinite)
static VALUE BigDecimal_IsFinite(VALUE self) { Real *p = GetVpValue(self,1); if(VpIsNaN(p)) return Qfalse; if(VpIsInf(p)) return Qfalse; return Qtrue; }
Return the integer part of the number.
static VALUE BigDecimal_fix(VALUE self) { ENTER(5); Real *c, *a; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpActiveRound(c,a,VP_ROUND_DOWN,0); /* 0: round off */ return ToValue(c); }
Return the largest integer less than or equal to the value, as a BigDecimal.
BigDecimal('3.14159').floor -> 3
BigDecimal('-9.1').floor -> -10
If n is specified and positive, the fractional part of the result has no more than that many digits.
If n is specified and negative, at least that many digits to the left of the decimal point will be 0 in the result.
BigDecimal('3.14159').floor(3) -> 3.141
BigDecimal('13345.234').floor(-2) -> 13300.0
static VALUE BigDecimal_floor(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *c, *a; U_LONG mx; int iLoc; VALUE vLoc; U_LONG pl = VpSetPrecLimit(0); if(rb_scan_args(argc,argv,"01",&vLoc)==0) { iLoc = 0; } else { Check_Type(vLoc, T_FIXNUM); iLoc = FIX2INT(vLoc); } GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpSetPrecLimit(pl); VpActiveRound(c,a,VP_ROUND_FLOOR,iLoc); return ToValue(c); }
Return the fractional part of the number.
static VALUE BigDecimal_frac(VALUE self) { ENTER(5); Real *c, *a; U_LONG mx; GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpFrac(c, a); return ToValue(c); }
static VALUE BigDecimal_hash(VALUE self) { ENTER(1); Real *p; U_LONG hash,i; GUARD_OBJ(p,GetVpValue(self,1)); hash = (U_LONG)p->sign; /* hash!=2: the case for 0(1),NaN(0) or +-Infinity(3) is sign itself */ if(hash==2) { for(i = 0; i < p->Prec;i++) { hash = 31 * hash + p->frac[i]; hash ^= p->frac[i]; } hash += p->exponent; } return INT2FIX(hash); }
Returns True if the value is infinite
static VALUE BigDecimal_IsInfinite(VALUE self) { Real *p = GetVpValue(self,1); if(VpIsPosInf(p)) return INT2FIX(1); if(VpIsNegInf(p)) return INT2FIX(-1); return Qnil; }
Returns debugging information about the value as a string of comma-separated values in angle brackets with a leading #:
::new(“1234.5678”).inspect -> “#<BigDecimal:b7ea1130,'0.12345678E4',8(12)>”
The first part is the address, the second is the value as a string, and the final part ss(mm) is the current number of significant digits and the maximum number of significant digits, respectively.
static VALUE BigDecimal_inspect(VALUE self) { ENTER(5); Real *vp; volatile VALUE obj; unsigned int nc; char *psz, *tmp; GUARD_OBJ(vp,GetVpValue(self,1)); nc = VpNumOfChars(vp,"E"); nc +=(nc + 9) / 10; obj = rb_str_new(0, nc+256); psz = RSTRING_PTR(obj); sprintf(psz,"#<BigDecimal:%lx,'",self); tmp = psz + strlen(psz); VpToString(vp, tmp, 10, 0); tmp += strlen(tmp); sprintf(tmp,"',%lu(%lu)>",VpPrec(vp)*VpBaseFig(),VpMaxPrec(vp)*VpBaseFig()); rb_str_resize(obj, strlen(psz)); return obj; }
Returns the modulus from dividing by b. See divmod.
static VALUE BigDecimal_mod(VALUE self, VALUE r)
static VALUE BigDecimal_mult2(VALUE self, VALUE b, VALUE n) { ENTER(2); Real *cv; U_LONG mx = (U_LONG)GetPositiveInt(n); if(mx==0) return BigDecimal_mult(self,b); else { U_LONG pl = VpSetPrecLimit(0); VALUE c = BigDecimal_mult(self,b); VpSetPrecLimit(pl); GUARD_OBJ(cv,GetVpValue(c,1)); VpLeftRound(cv,VpGetRoundMode(),mx); return ToValue(cv); } }
Returns True if the value is Not a Number
static VALUE BigDecimal_IsNaN(VALUE self) { Real *p = GetVpValue(self,1); if(VpIsNaN(p)) return Qtrue; return Qfalse; }
Returns True if the value is non-zero.
static VALUE BigDecimal_nonzero(VALUE self) { Real *a = GetVpValue(self,1); return VpIsZero(a) ? Qnil : self; }
Returns the value raised to the power of n. Note that n must be an Integer.
Also available as the operator **
static VALUE BigDecimal_power(VALUE self, VALUE p) { ENTER(5); Real *x, *y; S_LONG mp, ma, n; Check_Type(p, T_FIXNUM); n = FIX2INT(p); ma = n; if(ma < 0) ma = -ma; if(ma == 0) ma = 1; GUARD_OBJ(x,GetVpValue(self,1)); if(VpIsDef(x)) { mp = x->Prec *(VpBaseFig() + 1); GUARD_OBJ(y,VpCreateRbObject(mp *(ma + 1), "0")); } else { GUARD_OBJ(y,VpCreateRbObject(1, "0")); } VpPower(y, x, n); return ToValue(y); }
Returns an Array of two Integer values.
The first value is the current number of significant digits in the BigDecimal. The second value is the maximum number of significant digits for the BigDecimal.
static VALUE BigDecimal_prec(VALUE self) { ENTER(1); Real *p; VALUE obj; GUARD_OBJ(p,GetVpValue(self,1)); obj = rb_assoc_new(INT2NUM(p->Prec*VpBaseFig()), INT2NUM(p->MaxPrec*VpBaseFig())); return obj; }
Divide by the specified value.
e.g.
c = a.div(b,n)
If specified and less than the number of significant digits of the result, the result is rounded to that number of digits, according to ::mode.
If digits is 0, the result is the same as the / operator. If not, the result is an integer BigDecimal, by analogy with Float#div.
The alias quo is provided since div(value, 0) is the same as computing the quotient; see divmod.
static VALUE BigDecimal_div(VALUE self, VALUE r) /* For c = self/r: with round operation */ { ENTER(5); Real *c=NULL, *res=NULL, *div = NULL; r = BigDecimal_divide(&c, &res, &div, self, r); if(r!=(VALUE)0) return r; /* coerced by other */ SAVE(c);SAVE(res);SAVE(div); /* a/b = c + r/b */ /* c xxxxx r 00000yyyyy ==> (y/b)*BASE >= HALF_BASE */ /* Round */ if(VpHasVal(div)) { /* frac[0] must be zero for NaN,INF,Zero */ VpInternalRound(c,0,c->frac[c->Prec-1],(VpBaseVal()*res->frac[0])/div->frac[0]); } return ToValue(c); }
Returns the remainder from dividing by the value.
If the values divided are of the same sign, the remainder is the same as the modulus (see divmod).
Otherwise, the remainder is the modulus minus the value divided by.
static VALUE BigDecimal_remainder(VALUE self, VALUE r) /* remainder */ { VALUE f; Real *d,*rv=0; f = BigDecimal_divremain(self,r,&d,&rv); if(f!=(VALUE)0) return f; return ToValue(rv); }
Round to the nearest 1 (by default), returning the result as a BigDecimal.
BigDecimal('3.14159').round -> 3
BigDecimal('8.7').round -> 9
If n is specified and positive, the fractional part of the result has no more than that many digits.
If n is specified and negative, at least that many digits to the left of the decimal point will be 0 in the result.
BigDecimal('3.14159').round(3) -> 3.142
BigDecimal('13345.234').round(-2) -> 13300.0
The value of the optional mode argument can be used to determine how rounding is performed; see ::mode.
static VALUE BigDecimal_round(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *c, *a; int iLoc = 0; U_LONG mx; VALUE vLoc; VALUE vRound; U_LONG pl; int sw = VpGetRoundMode(); int na = rb_scan_args(argc,argv,"02",&vLoc,&vRound); switch(na) { case 0: iLoc = 0; break; case 1: Check_Type(vLoc, T_FIXNUM); iLoc = FIX2INT(vLoc); break; case 2: Check_Type(vLoc, T_FIXNUM); iLoc = FIX2INT(vLoc); Check_Type(vRound, T_FIXNUM); sw = FIX2INT(vRound); if(!VpIsRoundMode(sw)) { rb_raise(rb_eTypeError, "invalid rounding mode"); return Qnil; } break; } pl = VpSetPrecLimit(0); GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpSetPrecLimit(pl); VpActiveRound(c,a,sw,iLoc); return ToValue(c); }
Returns the sign of the value.
Returns a positive value if > 0, a negative value if < 0, and a zero if == 0.
The specific value returned indicates the type and sign of the BigDecimal, as follows:
value is Not a Number
value is +0
value is -0
value is +infinity
value is -infinity
value is positive
value is negative
static VALUE BigDecimal_sign(VALUE self) { /* sign */ int s = GetVpValue(self,1)->sign; return INT2FIX(s); }
Splits a BigDecimal number into four parts, returned as an array of values.
The first value represents the sign of the BigDecimal, and is -1 or 1, or 0 if the BigDecimal is Not a Number.
The second value is a string representing the significant digits of the BigDecimal, with no leading zeros.
The third value is the base used for arithmetic (currently always 10) as an Integer.
The fourth value is an Integer exponent.
If the BigDecimal can be represented as 0.xxxxxx*10**n, then xxxxxx is the string of significant digits with no leading zeros, and n is the exponent.
From these values, you can translate a BigDecimal to a float as follows:
sign, significant_digits, base, exponent = a.split f = sign * "0.#{significant_digits}".to_f * (base ** exponent)
(Note that the #to_f method is provided as a more convenient way to translate a BigDecimal to a Float.)
static VALUE BigDecimal_split(VALUE self) { ENTER(5); Real *vp; VALUE obj,str; S_LONG e; S_LONG s; char *psz1; GUARD_OBJ(vp,GetVpValue(self,1)); str = rb_str_new(0, VpNumOfChars(vp,"E")); psz1 = RSTRING_PTR(str); VpSzMantissa(vp,psz1); s = 1; if(psz1[0]=='-') { int len = strlen(psz1+1); memmove(psz1, psz1+1, len); psz1[len] = '\0'; s = -1; } if(psz1[0]=='N') s=0; /* NaN */ e = VpExponent10(vp); obj = rb_ary_new2(4); rb_ary_push(obj, INT2FIX(s)); rb_ary_push(obj, str); rb_str_resize(str, strlen(psz1)); rb_ary_push(obj, INT2FIX(10)); rb_ary_push(obj, INT2NUM(e)); return obj; }
Returns the square root of the value.
If n is specified, returns at least that many significant digits.
static VALUE BigDecimal_sqrt(VALUE self, VALUE nFig) { ENTER(5); Real *c, *a; S_INT mx, n; GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); n = GetPositiveInt(nFig) + VpDblFig() + 1; if(mx <= n) mx = n; GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpSqrt(c, a); return ToValue(c); }
static VALUE BigDecimal_sub2(VALUE self, VALUE b, VALUE n) { ENTER(2); Real *cv; U_LONG mx = (U_LONG)GetPositiveInt(n); if(mx==0) return BigDecimal_sub(self,b); else { U_LONG pl = VpSetPrecLimit(0); VALUE c = BigDecimal_sub(self,b); VpSetPrecLimit(pl); GUARD_OBJ(cv,GetVpValue(c,1)); VpLeftRound(cv,VpGetRoundMode(),mx); return ToValue(cv); } }
Converts a BigDecimal to a String of the form “nnnnnn.mmm”. This method is deprecated; use #to_s(“F”) instead.
# File bigdecimal/lib/bigdecimal/util.rb, line 33 def to_digits if self.nan? || self.infinite? || self.zero? self.to_s else i = self.to_i.to_s s,f,y,z = self.frac.split i + "." + ("0"*(-z)) + f end end
Returns a new Float object having approximately the same value as the BigDecimal number. Normal accuracy limits and built-in errors of binary Float arithmetic apply.
static VALUE BigDecimal_to_f(VALUE self) { ENTER(1); Real *p; double d; S_LONG e; char *buf; volatile VALUE str; GUARD_OBJ(p,GetVpValue(self,1)); if (VpVtoD(&d, &e, p)!=1) return rb_float_new(d); if (e > DBL_MAX_10_EXP+BASE_FIG) goto erange; str = rb_str_new(0, VpNumOfChars(p,"E")); buf = RSTRING_PTR(str); VpToString(p, buf, 0, 0); errno = 0; d = strtod(buf, 0); if(errno == ERANGE) { erange: VpException(VP_EXCEPTION_OVERFLOW,"BigDecimal to Float conversion",0); if(d>0.0) d = VpGetDoublePosInf(); else d = VpGetDoubleNegInf(); } return rb_float_new(d); }
Returns the value as an integer (Fixnum or Bignum).
If the BigNumber is infinity or NaN, returns nil.
static VALUE BigDecimal_to_i(VALUE self) { ENTER(5); int e,n,i,nf; U_LONG v,b,j; volatile VALUE str; char *psz,*pch; Real *p; GUARD_OBJ(p,GetVpValue(self,1)); BigDecimal_check_num(p); e = VpExponent10(p); if(e<=0) return INT2FIX(0); nf = VpBaseFig(); if(e<=nf) { e = VpGetSign(p)*p->frac[0]; return INT2FIX(e); } str = rb_str_new(0, e+nf+2); psz = RSTRING_PTR(str); n = (e+nf-1)/nf; pch = psz; if(VpGetSign(p)<0) *pch++ = '-'; for(i=0;i<n;++i) { b = VpBaseVal()/10; if(i>=(int)p->Prec) { while(b) { *pch++ = '0'; b /= 10; } continue; } v = p->frac[i]; while(b) { j = v/b; *pch++ = (char)(j + '0'); v -= j*b; b /= 10; } } *pch++ = 0; return rb_cstr2inum(psz,10); }
Returns the value as an integer (Fixnum or Bignum).
If the BigNumber is infinity or NaN, returns nil.
static VALUE BigDecimal_to_i(VALUE self) { ENTER(5); int e,n,i,nf; U_LONG v,b,j; volatile VALUE str; char *psz,*pch; Real *p; GUARD_OBJ(p,GetVpValue(self,1)); BigDecimal_check_num(p); e = VpExponent10(p); if(e<=0) return INT2FIX(0); nf = VpBaseFig(); if(e<=nf) { e = VpGetSign(p)*p->frac[0]; return INT2FIX(e); } str = rb_str_new(0, e+nf+2); psz = RSTRING_PTR(str); n = (e+nf-1)/nf; pch = psz; if(VpGetSign(p)<0) *pch++ = '-'; for(i=0;i<n;++i) { b = VpBaseVal()/10; if(i>=(int)p->Prec) { while(b) { *pch++ = '0'; b /= 10; } continue; } v = p->frac[i]; while(b) { j = v/b; *pch++ = (char)(j + '0'); v -= j*b; b /= 10; } } *pch++ = 0; return rb_cstr2inum(psz,10); }
Converts a BigDecimal to a Rational.
static VALUE BigDecimal_to_r(VALUE self) { Real *p; S_LONG sign, power, denomi_power; VALUE a, digits, numerator; p = GetVpValue(self,1); BigDecimal_check_num(p); sign = VpGetSign(p); power = VpExponent10(p); a = BigDecimal_split(self); digits = RARRAY_PTR(a)[1]; denomi_power = power - RSTRING_LEN(digits); numerator = rb_funcall(digits, rb_intern("to_i"), 0); if (sign < 0) { numerator = rb_funcall(numerator, '*', 1, INT2FIX(-1)); } if (denomi_power < 0) { return rb_Rational(numerator, rb_funcall(INT2FIX(10), rb_intern("**"), 1, INT2FIX(-denomi_power))); } else { return rb_Rational1(rb_funcall(numerator, '*', 1, rb_funcall(INT2FIX(10), rb_intern("**"), 1, INT2FIX(denomi_power)))); } }
Converts the value to a string.
The default format looks like 0.xxxxEnn.
The optional parameter s consists of either an integer; or an optional '+' or ' ', followed by an optional number, followed by an optional 'E' or 'F'.
If there is a '+' at the start of s, positive values are returned with a leading '+'.
A space at the start of s returns positive values with a leading space.
If s contains a number, a space is inserted after each group of that many fractional digits.
If s ends with an 'E', engineering notation (0.xxxxEnn) is used.
If s ends with an 'F', conventional floating point notation is used.
Examples:
::new('-123.45678901234567890').to_s('5F') -> '-123.45678 90123 45678 9'
::new('123.45678901234567890').to_s('+8F') -> '+123.45678901 23456789'
::new('123.45678901234567890').to_s(' F') -> ' 123.4567890123456789'
static VALUE BigDecimal_to_s(int argc, VALUE *argv, VALUE self) { ENTER(5); int fmt=0; /* 0:E format */ int fPlus=0; /* =0:default,=1: set ' ' before digits ,set '+' before digits. */ Real *vp; volatile VALUE str; char *psz; char ch; U_LONG nc; S_INT mc = 0; VALUE f; GUARD_OBJ(vp,GetVpValue(self,1)); if(rb_scan_args(argc,argv,"01",&f)==1) { if(TYPE(f)==T_STRING) { SafeStringValue(f); psz = RSTRING_PTR(f); if(*psz==' ') { fPlus = 1; psz++; } else if(*psz=='+') { fPlus = 2; psz++; } while((ch=*psz++)!=0) { if(ISSPACE(ch)) continue; if(!ISDIGIT(ch)) { if(ch=='F' || ch=='f') fmt = 1; /* F format */ break; } mc = mc * 10 + ch - '0'; } } else { mc = GetPositiveInt(f); } } if(fmt) { nc = VpNumOfChars(vp,"F"); } else { nc = VpNumOfChars(vp,"E"); } if(mc>0) nc += (nc + mc - 1) / mc + 1; str = rb_str_new(0, nc); psz = RSTRING_PTR(str); if(fmt) { VpToFString(vp, psz, mc, fPlus); } else { VpToString (vp, psz, mc, fPlus); } rb_str_resize(str, strlen(psz)); return str; }
Truncate to the nearest 1, returning the result as a BigDecimal.
BigDecimal('3.14159').truncate -> 3
BigDecimal('8.7').truncate -> 8
If n is specified and positive, the fractional part of the result has no more than that many digits.
If n is specified and negative, at least that many digits to the left of the decimal point will be 0 in the result.
BigDecimal('3.14159').truncate(3) -> 3.141
BigDecimal('13345.234').truncate(-2) -> 13300.0
static VALUE BigDecimal_truncate(int argc, VALUE *argv, VALUE self) { ENTER(5); Real *c, *a; int iLoc; U_LONG mx; VALUE vLoc; U_LONG pl = VpSetPrecLimit(0); if(rb_scan_args(argc,argv,"01",&vLoc)==0) { iLoc = 0; } else { Check_Type(vLoc, T_FIXNUM); iLoc = FIX2INT(vLoc); } GUARD_OBJ(a,GetVpValue(self,1)); mx = a->Prec *(VpBaseFig() + 1); GUARD_OBJ(c,VpCreateRbObject(mx, "0")); VpSetPrecLimit(pl); VpActiveRound(c,a,VP_ROUND_DOWN,iLoc); /* 0: truncate */ return ToValue(c); }