1 package Math::BigFloat;
4 # Mike grinned. 'Two down, infinity to go' - Mike Nostrus in 'Before and After'
7 # The following hash values are internally used:
8 # _e : exponent (ref to $CALC object)
9 # _m : mantissa (ref to $CALC object)
11 # sign : +,-,+inf,-inf, or "NaN" if not a number
19 @ISA = qw(Exporter Math::BigInt);
22 # $_trap_inf/$_trap_nan are internal and should never be accessed from outside
23 use vars qw/$AUTOLOAD $accuracy $precision $div_scale $round_mode $rnd_mode
24 $upgrade $downgrade $_trap_nan $_trap_inf/;
25 my $class = "Math::BigFloat";
28 '<=>' => sub { $_[2] ?
29 ref($_[0])->bcmp($_[1],$_[0]) :
30 ref($_[0])->bcmp($_[0],$_[1])},
31 'int' => sub { $_[0]->as_number() }, # 'trunc' to bigint
34 ##############################################################################
35 # global constants, flags and assorted stuff
37 # the following are public, but their usage is not recommended. Use the
38 # accessor methods instead.
40 # class constants, use Class->constant_name() to access
41 $round_mode = 'even'; # one of 'even', 'odd', '+inf', '-inf', 'zero' or 'trunc'
48 # the package we are using for our private parts, defaults to:
49 # Math::BigInt->config()->{lib}
50 my $MBI = 'Math::BigInt::Calc';
52 # are NaNs ok? (otherwise it dies when encountering an NaN) set w/ config()
54 # the same for infinity
57 # constant for easier life
60 my $IMPORT = 0; # was import() called yet? used to make require work
62 # some digits of accuracy for blog(undef,10); which we use in blog() for speed
64 '2.3025850929940456840179914546843642076011014886287729760333279009675726097';
65 my $LOG_10_A = length($LOG_10)-1;
68 '0.6931471805599453094172321214581765680755001343602552541206800094933936220';
69 my $LOG_2_A = length($LOG_2)-1;
70 my $HALF = '0.5'; # made into an object if necc.
72 ##############################################################################
73 # the old code had $rnd_mode, so we need to support it, too
75 sub TIESCALAR { my ($class) = @_; bless \$round_mode, $class; }
76 sub FETCH { return $round_mode; }
77 sub STORE { $rnd_mode = $_[0]->round_mode($_[1]); }
81 # when someone set's $rnd_mode, we catch this and check the value to see
82 # whether it is valid or not.
83 $rnd_mode = 'even'; tie $rnd_mode, 'Math::BigFloat';
86 ##############################################################################
89 # valid method aliases for AUTOLOAD
90 my %methods = map { $_ => 1 }
91 qw / fadd fsub fmul fdiv fround ffround fsqrt fmod fstr fsstr fpow fnorm
92 fint facmp fcmp fzero fnan finf finc fdec flog ffac
93 fceil ffloor frsft flsft fone flog froot
95 # valid method's that can be hand-ed up (for AUTOLOAD)
96 my %hand_ups = map { $_ => 1 }
97 qw / is_nan is_inf is_negative is_positive is_pos is_neg
98 accuracy precision div_scale round_mode fneg fabs fnot
99 objectify upgrade downgrade
103 sub method_alias { exists $methods{$_[0]||''}; }
104 sub method_hand_up { exists $hand_ups{$_[0]||''}; }
107 ##############################################################################
112 # create a new BigFloat object from a string or another bigfloat object.
115 # sign => sign (+/-), or "NaN"
117 my ($class,$wanted,@r) = @_;
119 # avoid numify-calls by not using || on $wanted!
120 return $class->bzero() if !defined $wanted; # default to 0
121 return $wanted->copy() if UNIVERSAL::isa($wanted,'Math::BigFloat');
123 $class->import() if $IMPORT == 0; # make require work
125 my $self = {}; bless $self, $class;
126 # shortcut for bigints and its subclasses
127 if ((ref($wanted)) && (ref($wanted) ne $class))
129 $self->{_m} = $wanted->as_number()->{value}; # get us a bigint copy
130 $self->{_e} = $MBI->_zero();
132 $self->{sign} = $wanted->sign();
133 return $self->bnorm();
136 # handle '+inf', '-inf' first
137 if ($wanted =~ /^[+-]?inf$/)
139 return $downgrade->new($wanted) if $downgrade;
141 $self->{_e} = $MBI->_zero();
143 $self->{_m} = $MBI->_zero();
144 $self->{sign} = $wanted;
145 $self->{sign} = '+inf' if $self->{sign} eq 'inf';
146 return $self->bnorm();
149 my ($mis,$miv,$mfv,$es,$ev) = Math::BigInt::_split($wanted);
155 Carp::croak ("$wanted is not a number initialized to $class");
158 return $downgrade->bnan() if $downgrade;
160 $self->{_e} = $MBI->_zero();
162 $self->{_m} = $MBI->_zero();
163 $self->{sign} = $nan;
167 # make integer from mantissa by adjusting exp, then convert to int
168 $self->{_e} = $MBI->_new($$ev); # exponent
169 $self->{_es} = $$es || '+';
170 my $mantissa = "$$miv$$mfv"; # create mant.
171 $mantissa =~ s/^0+(\d)/$1/; # strip leading zeros
172 $self->{_m} = $MBI->_new($mantissa); # create mant.
174 # 3.123E0 = 3123E-3, and 3.123E-2 => 3123E-5
175 if (CORE::length($$mfv) != 0)
177 my $len = $MBI->_new( CORE::length($$mfv));
178 ($self->{_e}, $self->{_es}) =
179 _e_sub ($self->{_e}, $len, $self->{_es}, '+');
181 $self->{sign} = $$mis;
183 # we can only have trailing zeros on the mantissa of $$mfv eq ''
184 if (CORE::length($$mfv) == 0)
186 my $zeros = $MBI->_zeros($self->{_m}); # correct for trailing zeros
189 my $z = $MBI->_new($zeros);
190 $MBI->_rsft ( $self->{_m}, $z, 10);
191 _e_add ( $self->{_e}, $z, $self->{_es}, '+');
194 # for something like 0Ey, set y to 1, and -0 => +0
195 $self->{sign} = '+', $self->{_e} = $MBI->_one()
196 if $MBI->_is_zero($self->{_m});
197 return $self->round(@r) if !$downgrade;
199 # if downgrade, inf, NaN or integers go down
201 if ($downgrade && $self->{_es} eq '+')
203 if ($MBI->_is_zero( $self->{_e} ))
205 return $downgrade->new($$mis . $MBI->_str( $self->{_m} ));
207 return $downgrade->new($self->bsstr());
209 $self->bnorm()->round(@r); # first normalize, then round
217 # if two arguments, the first one is the class to "swallow" subclasses
225 return unless ref($x); # only for objects
227 my $self = {}; bless $self,$c;
229 $self->{sign} = $x->{sign};
230 $self->{_es} = $x->{_es};
231 $self->{_m} = $MBI->_copy($x->{_m});
232 $self->{_e} = $MBI->_copy($x->{_e});
233 $self->{_a} = $x->{_a} if defined $x->{_a};
234 $self->{_p} = $x->{_p} if defined $x->{_p};
240 # used by parent class bone() to initialize number to NaN
246 my $class = ref($self);
247 Carp::croak ("Tried to set $self to NaN in $class\::_bnan()");
250 $IMPORT=1; # call our import only once
251 $self->{_m} = $MBI->_zero();
252 $self->{_e} = $MBI->_zero();
258 # used by parent class bone() to initialize number to +-inf
264 my $class = ref($self);
265 Carp::croak ("Tried to set $self to +-inf in $class\::_binf()");
268 $IMPORT=1; # call our import only once
269 $self->{_m} = $MBI->_zero();
270 $self->{_e} = $MBI->_zero();
276 # used by parent class bone() to initialize number to 1
278 $IMPORT=1; # call our import only once
279 $self->{_m} = $MBI->_one();
280 $self->{_e} = $MBI->_zero();
286 # used by parent class bone() to initialize number to 0
288 $IMPORT=1; # call our import only once
289 $self->{_m} = $MBI->_zero();
290 $self->{_e} = $MBI->_one();
296 my ($self,$class) = @_;
297 return if $class =~ /^Math::BigInt/; # we aren't one of these
298 UNIVERSAL::isa($self,$class);
303 # return (later set?) configuration data as hash ref
304 my $class = shift || 'Math::BigFloat';
306 my $cfg = $class->SUPER::config(@_);
308 # now we need only to override the ones that are different from our parent
309 $cfg->{class} = $class;
314 ##############################################################################
315 # string conversation
319 # (ref to BFLOAT or num_str ) return num_str
320 # Convert number from internal format to (non-scientific) string format.
321 # internal format is always normalized (no leading zeros, "-0" => "+0")
322 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
324 if ($x->{sign} !~ /^[+-]$/)
326 return $x->{sign} unless $x->{sign} eq '+inf'; # -inf, NaN
330 my $es = '0'; my $len = 1; my $cad = 0; my $dot = '.';
333 my $not_zero = !($x->{sign} eq '+' && $MBI->_is_zero($x->{_m}));
336 $es = $MBI->_str($x->{_m});
337 $len = CORE::length($es);
338 my $e = $MBI->_num($x->{_e});
339 $e = -$e if $x->{_es} eq '-';
343 # if _e is bigger than a scalar, the following will blow your memory
346 my $r = abs($e) - $len;
347 $es = '0.'. ('0' x $r) . $es; $cad = -($len+$r);
351 substr($es,$e,0) = '.'; $cad = $MBI->_num($x->{_e});
352 $cad = -$cad if $x->{_es} eq '-';
358 $es .= '0' x $e; $len += $e; $cad = 0;
362 $es = '-'.$es if $x->{sign} eq '-';
363 # if set accuracy or precision, pad with zeros on the right side
364 if ((defined $x->{_a}) && ($not_zero))
366 # 123400 => 6, 0.1234 => 4, 0.001234 => 4
367 my $zeros = $x->{_a} - $cad; # cad == 0 => 12340
368 $zeros = $x->{_a} - $len if $cad != $len;
369 $es .= $dot.'0' x $zeros if $zeros > 0;
371 elsif ((($x->{_p} || 0) < 0))
373 # 123400 => 6, 0.1234 => 4, 0.001234 => 6
374 my $zeros = -$x->{_p} + $cad;
375 $es .= $dot.'0' x $zeros if $zeros > 0;
382 # (ref to BFLOAT or num_str ) return num_str
383 # Convert number from internal format to scientific string format.
384 # internal format is always normalized (no leading zeros, "-0E0" => "+0E0")
385 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
387 if ($x->{sign} !~ /^[+-]$/)
389 return $x->{sign} unless $x->{sign} eq '+inf'; # -inf, NaN
392 my $sep = 'e'.$x->{_es};
393 my $sign = $x->{sign}; $sign = '' if $sign eq '+';
394 $sign . $MBI->_str($x->{_m}) . $sep . $MBI->_str($x->{_e});
399 # Make a number from a BigFloat object
400 # simple return a string and let Perl's atoi()/atof() handle the rest
401 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
405 ##############################################################################
406 # public stuff (usually prefixed with "b")
409 # XXX TODO this must be overwritten and return NaN for non-integer values
410 # band(), bior(), bxor(), too
413 # $class->SUPER::bnot($class,@_);
418 # Compares 2 values. Returns one of undef, <0, =0, >0. (suitable for sort)
421 my ($self,$x,$y) = (ref($_[0]),@_);
422 # objectify is costly, so avoid it
423 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
425 ($self,$x,$y) = objectify(2,@_);
428 return $upgrade->bcmp($x,$y) if defined $upgrade &&
429 ((!$x->isa($self)) || (!$y->isa($self)));
431 if (($x->{sign} !~ /^[+-]$/) || ($y->{sign} !~ /^[+-]$/))
433 # handle +-inf and NaN
434 return undef if (($x->{sign} eq $nan) || ($y->{sign} eq $nan));
435 return 0 if ($x->{sign} eq $y->{sign}) && ($x->{sign} =~ /^[+-]inf$/);
436 return +1 if $x->{sign} eq '+inf';
437 return -1 if $x->{sign} eq '-inf';
438 return -1 if $y->{sign} eq '+inf';
442 # check sign for speed first
443 return 1 if $x->{sign} eq '+' && $y->{sign} eq '-'; # does also 0 <=> -y
444 return -1 if $x->{sign} eq '-' && $y->{sign} eq '+'; # does also -x <=> 0
447 my $xz = $x->is_zero();
448 my $yz = $y->is_zero();
449 return 0 if $xz && $yz; # 0 <=> 0
450 return -1 if $xz && $y->{sign} eq '+'; # 0 <=> +y
451 return 1 if $yz && $x->{sign} eq '+'; # +x <=> 0
453 # adjust so that exponents are equal
454 my $lxm = $MBI->_len($x->{_m});
455 my $lym = $MBI->_len($y->{_m});
456 # the numify somewhat limits our length, but makes it much faster
457 my ($xes,$yes) = (1,1);
458 $xes = -1 if $x->{_es} ne '+';
459 $yes = -1 if $y->{_es} ne '+';
460 my $lx = $lxm + $xes * $MBI->_num($x->{_e});
461 my $ly = $lym + $yes * $MBI->_num($y->{_e});
462 my $l = $lx - $ly; $l = -$l if $x->{sign} eq '-';
463 return $l <=> 0 if $l != 0;
465 # lengths (corrected by exponent) are equal
466 # so make mantissa equal length by padding with zero (shift left)
467 my $diff = $lxm - $lym;
468 my $xm = $x->{_m}; # not yet copy it
472 $ym = $MBI->_copy($y->{_m});
473 $ym = $MBI->_lsft($ym, $MBI->_new($diff), 10);
477 $xm = $MBI->_copy($x->{_m});
478 $xm = $MBI->_lsft($xm, $MBI->_new(-$diff), 10);
480 my $rc = $MBI->_acmp($xm,$ym);
481 $rc = -$rc if $x->{sign} eq '-'; # -124 < -123
487 # Compares 2 values, ignoring their signs.
488 # Returns one of undef, <0, =0, >0. (suitable for sort)
491 my ($self,$x,$y) = (ref($_[0]),@_);
492 # objectify is costly, so avoid it
493 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
495 ($self,$x,$y) = objectify(2,@_);
498 return $upgrade->bacmp($x,$y) if defined $upgrade &&
499 ((!$x->isa($self)) || (!$y->isa($self)));
501 # handle +-inf and NaN's
502 if ($x->{sign} !~ /^[+-]$/ || $y->{sign} !~ /^[+-]$/)
504 return undef if (($x->{sign} eq $nan) || ($y->{sign} eq $nan));
505 return 0 if ($x->is_inf() && $y->is_inf());
506 return 1 if ($x->is_inf() && !$y->is_inf());
511 my $xz = $x->is_zero();
512 my $yz = $y->is_zero();
513 return 0 if $xz && $yz; # 0 <=> 0
514 return -1 if $xz && !$yz; # 0 <=> +y
515 return 1 if $yz && !$xz; # +x <=> 0
517 # adjust so that exponents are equal
518 my $lxm = $MBI->_len($x->{_m});
519 my $lym = $MBI->_len($y->{_m});
520 my ($xes,$yes) = (1,1);
521 $xes = -1 if $x->{_es} ne '+';
522 $yes = -1 if $y->{_es} ne '+';
523 # the numify somewhat limits our length, but makes it much faster
524 my $lx = $lxm + $xes * $MBI->_num($x->{_e});
525 my $ly = $lym + $yes * $MBI->_num($y->{_e});
527 return $l <=> 0 if $l != 0;
529 # lengths (corrected by exponent) are equal
530 # so make mantissa equal-length by padding with zero (shift left)
531 my $diff = $lxm - $lym;
532 my $xm = $x->{_m}; # not yet copy it
536 $ym = $MBI->_copy($y->{_m});
537 $ym = $MBI->_lsft($ym, $MBI->_new($diff), 10);
541 $xm = $MBI->_copy($x->{_m});
542 $xm = $MBI->_lsft($xm, $MBI->_new(-$diff), 10);
544 $MBI->_acmp($xm,$ym);
549 # add second arg (BFLOAT or string) to first (BFLOAT) (modifies first)
550 # return result as BFLOAT
553 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
554 # objectify is costly, so avoid it
555 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
557 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
560 # inf and NaN handling
561 if (($x->{sign} !~ /^[+-]$/) || ($y->{sign} !~ /^[+-]$/))
564 return $x->bnan() if (($x->{sign} eq $nan) || ($y->{sign} eq $nan));
566 if (($x->{sign} =~ /^[+-]inf$/) && ($y->{sign} =~ /^[+-]inf$/))
568 # +inf++inf or -inf+-inf => same, rest is NaN
569 return $x if $x->{sign} eq $y->{sign};
572 # +-inf + something => +inf; something +-inf => +-inf
573 $x->{sign} = $y->{sign}, return $x if $y->{sign} =~ /^[+-]inf$/;
577 return $upgrade->badd($x,$y,$a,$p,$r) if defined $upgrade &&
578 ((!$x->isa($self)) || (!$y->isa($self)));
580 # speed: no add for 0+y or x+0
581 return $x->bround($a,$p,$r) if $y->is_zero(); # x+0
582 if ($x->is_zero()) # 0+y
584 # make copy, clobbering up x (modify in place!)
585 $x->{_e} = $MBI->_copy($y->{_e});
586 $x->{_es} = $y->{_es};
587 $x->{_m} = $MBI->_copy($y->{_m});
588 $x->{sign} = $y->{sign} || $nan;
589 return $x->round($a,$p,$r,$y);
592 # take lower of the two e's and adapt m1 to it to match m2
594 $e = $MBI->_zero() if !defined $e; # if no BFLOAT?
595 $e = $MBI->_copy($e); # make copy (didn't do it yet)
599 ($e,$es) = _e_sub($e, $x->{_e}, $y->{_es} || '+', $x->{_es});
601 my $add = $MBI->_copy($y->{_m});
603 if ($es eq '-') # < 0
605 $MBI->_lsft( $x->{_m}, $e, 10);
606 ($x->{_e},$x->{_es}) = _e_add($x->{_e}, $e, $x->{_es}, $es);
608 elsif (!$MBI->_is_zero($e)) # > 0
610 $MBI->_lsft($add, $e, 10);
612 # else: both e are the same, so just leave them
614 if ($x->{sign} eq $y->{sign})
617 $x->{_m} = $MBI->_add($x->{_m}, $add);
621 ($x->{_m}, $x->{sign}) =
622 _e_add($x->{_m}, $add, $x->{sign}, $y->{sign});
625 # delete trailing zeros, then round
626 $x->bnorm()->round($a,$p,$r,$y);
629 # sub bsub is inherited from Math::BigInt!
633 # increment arg by one
634 my ($self,$x,@r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
636 if ($x->{_es} eq '-')
638 return $x->badd($self->bone(),@r); # digits after dot
641 if (!$MBI->_is_zero($x->{_e})) # _e == 0 for NaN, inf, -inf
643 # 1e2 => 100, so after the shift below _m has a '0' as last digit
644 $x->{_m} = $MBI->_lsft($x->{_m}, $x->{_e},10); # 1e2 => 100
645 $x->{_e} = $MBI->_zero(); # normalize
647 # we know that the last digit of $x will be '1' or '9', depending on the
651 if ($x->{sign} eq '+')
653 $MBI->_inc($x->{_m});
654 return $x->bnorm()->bround(@r);
656 elsif ($x->{sign} eq '-')
658 $MBI->_dec($x->{_m});
659 $x->{sign} = '+' if $MBI->_is_zero($x->{_m}); # -1 +1 => -0 => +0
660 return $x->bnorm()->bround(@r);
662 # inf, nan handling etc
663 $x->badd($self->bone(),@r); # badd() does round
668 # decrement arg by one
669 my ($self,$x,@r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
671 if ($x->{_es} eq '-')
673 return $x->badd($self->bone('-'),@r); # digits after dot
676 if (!$MBI->_is_zero($x->{_e}))
678 $x->{_m} = $MBI->_lsft($x->{_m}, $x->{_e},10); # 1e2 => 100
679 $x->{_e} = $MBI->_zero(); # normalize
683 my $zero = $x->is_zero();
685 if (($x->{sign} eq '-') || $zero)
687 $MBI->_inc($x->{_m});
688 $x->{sign} = '-' if $zero; # 0 => 1 => -1
689 $x->{sign} = '+' if $MBI->_is_zero($x->{_m}); # -1 +1 => -0 => +0
690 return $x->bnorm()->round(@r);
693 elsif ($x->{sign} eq '+')
695 $MBI->_dec($x->{_m});
696 return $x->bnorm()->round(@r);
698 # inf, nan handling etc
699 $x->badd($self->bone('-'),@r); # does round
706 my ($self,$x,$base,$a,$p,$r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
708 # $base > 0, $base != 1; if $base == undef default to $base == e
711 # we need to limit the accuracy to protect against overflow
714 ($x,@params) = $x->_find_round_parameters($a,$p,$r);
716 # also takes care of the "error in _find_round_parameters?" case
717 return $x->bnan() if $x->{sign} ne '+' || $x->is_zero();
720 # no rounding at all, so must use fallback
721 if (scalar @params == 0)
723 # simulate old behaviour
724 $params[0] = $self->div_scale(); # and round to it as accuracy
725 $params[1] = undef; # P = undef
726 $scale = $params[0]+4; # at least four more for proper round
727 $params[2] = $r; # round mode by caller or undef
728 $fallback = 1; # to clear a/p afterwards
732 # the 4 below is empirical, and there might be cases where it is not
734 $scale = abs($params[0] || $params[1]) + 4; # take whatever is defined
737 return $x->bzero(@params) if $x->is_one();
738 # base not defined => base == Euler's constant e
741 # make object, since we don't feed it through objectify() to still get the
742 # case of $base == undef
743 $base = $self->new($base) unless ref($base);
744 # $base > 0; $base != 1
745 return $x->bnan() if $base->is_zero() || $base->is_one() ||
746 $base->{sign} ne '+';
747 # if $x == $base, we know the result must be 1.0
748 return $x->bone('+',@params) if $x->bcmp($base) == 0;
751 # when user set globals, they would interfere with our calculation, so
752 # disable them and later re-enable them
754 my $abr = "$self\::accuracy"; my $ab = $$abr; $$abr = undef;
755 my $pbr = "$self\::precision"; my $pb = $$pbr; $$pbr = undef;
756 # we also need to disable any set A or P on $x (_find_round_parameters took
757 # them already into account), since these would interfere, too
758 delete $x->{_a}; delete $x->{_p};
759 # need to disable $upgrade in BigInt, to avoid deep recursion
760 local $Math::BigInt::upgrade = undef;
761 local $Math::BigFloat::downgrade = undef;
763 # upgrade $x if $x is not a BigFloat (handle BigInt input)
764 if (!$x->isa('Math::BigFloat'))
766 $x = Math::BigFloat->new($x);
772 # If the base is defined and an integer, try to calculate integer result
773 # first. This is very fast, and in case the real result was found, we can
775 if (defined $base && $base->is_int() && $x->is_int())
777 my $i = $MBI->_copy( $x->{_m} );
778 $MBI->_lsft( $i, $x->{_e}, 10 ) unless $MBI->_is_zero($x->{_e});
779 my $int = Math::BigInt->bzero();
781 $int->blog($base->as_number());
783 if ($base->as_number()->bpow($int) == $x)
785 # found result, return it
786 $x->{_m} = $int->{value};
787 $x->{_e} = $MBI->_zero();
796 # first calculate the log to base e (using reduction by 10 (and probably 2))
797 $self->_log_10($x,$scale);
799 # and if a different base was requested, convert it
802 $base = Math::BigFloat->new($base) unless $base->isa('Math::BigFloat');
803 # not ln, but some other base (don't modify $base)
804 $x->bdiv( $base->copy()->blog(undef,$scale), $scale );
808 # shortcut to not run through _find_round_parameters again
809 if (defined $params[0])
811 $x->bround($params[0],$params[2]); # then round accordingly
815 $x->bfround($params[1],$params[2]); # then round accordingly
819 # clear a/p after round, since user did not request it
820 delete $x->{_a}; delete $x->{_p};
823 $$abr = $ab; $$pbr = $pb;
830 # internal log function to calculate ln() based on Taylor series.
831 # Modifies $x in place.
832 my ($self,$x,$scale) = @_;
834 # in case of $x == 1, result is 0
835 return $x->bzero() if $x->is_one();
837 # http://www.efunda.com/math/taylor_series/logarithmic.cfm?search_string=log
841 # Taylor: | u 1 u^3 1 u^5 |
842 # ln (x) = 2 | --- + - * --- + - * --- + ... | x > 0
843 # |_ v 3 v^3 5 v^5 _|
845 # This takes much more steps to calculate the result and is thus not used
848 # Taylor: | u 1 u^2 1 u^3 |
849 # ln (x) = 2 | --- + - * --- + - * --- + ... | x > 1/2
850 # |_ x 2 x^2 3 x^3 _|
852 my ($limit,$v,$u,$below,$factor,$two,$next,$over,$f);
854 $v = $x->copy(); $v->binc(); # v = x+1
855 $x->bdec(); $u = $x->copy(); # u = x-1; x = x-1
856 $x->bdiv($v,$scale); # first term: u/v
859 $u *= $u; $v *= $v; # u^2, v^2
860 $below->bmul($v); # u^3, v^3
862 $factor = $self->new(3); $f = $self->new(2);
864 my $steps = 0 if DEBUG;
865 $limit = $self->new("1E-". ($scale-1));
868 # we calculate the next term, and add it to the last
869 # when the next term is below our limit, it won't affect the outcome
870 # anymore, so we stop
872 # calculating the next term simple from over/below will result in quite
873 # a time hog if the input has many digits, since over and below will
874 # accumulate more and more digits, and the result will also have many
875 # digits, but in the end it is rounded to $scale digits anyway. So if we
876 # round $over and $below first, we save a lot of time for the division
877 # (not with log(1.2345), but try log (123**123) to see what I mean. This
878 # can introduce a rounding error if the division result would be f.i.
879 # 0.1234500000001 and we round it to 5 digits it would become 0.12346, but
880 # if we truncated $over and $below we might get 0.12345. Does this matter
881 # for the end result? So we give $over and $below 4 more digits to be
882 # on the safe side (unscientific error handling as usual... :+D
884 $next = $over->copy->bround($scale+4)->bdiv(
885 $below->copy->bmul($factor)->bround($scale+4),
889 ## $next = $over->copy()->bdiv($below->copy()->bmul($factor),$scale);
891 last if $next->bacmp($limit) <= 0;
893 delete $next->{_a}; delete $next->{_p};
895 # calculate things for the next term
896 $over *= $u; $below *= $v; $factor->badd($f);
899 $steps++; print "step $steps = $x\n" if $steps % 10 == 0;
902 $x->bmul($f); # $x *= 2
903 print "took $steps steps\n" if DEBUG;
908 # Internal log function based on reducing input to the range of 0.1 .. 9.99
909 # and then "correcting" the result to the proper one. Modifies $x in place.
910 my ($self,$x,$scale) = @_;
912 # taking blog() from numbers greater than 10 takes a *very long* time, so we
913 # break the computation down into parts based on the observation that:
914 # blog(x*y) = blog(x) + blog(y)
915 # We set $y here to multiples of 10 so that $x is below 1 (the smaller $x is
916 # the faster it get's, especially because 2*$x takes about 10 times as long,
917 # so by dividing $x by 10 we make it at least factor 100 faster...)
919 # The same observation is valid for numbers smaller than 0.1 (e.g. computing
920 # log(1) is fastest, and the farther away we get from 1, the longer it takes)
921 # so we also 'break' this down by multiplying $x with 10 and subtract the
922 # log(10) afterwards to get the correct result.
924 # calculate nr of digits before dot
925 my $dbd = $MBI->_num($x->{_e});
926 $dbd = -$dbd if $x->{_es} eq '-';
927 $dbd += $MBI->_len($x->{_m});
929 # more than one digit (e.g. at least 10), but *not* exactly 10 to avoid
932 my $calc = 1; # do some calculation?
934 # disable the shortcut for 10, since we need log(10) and this would recurse
936 if ($x->{_es} eq '+' && $MBI->_is_one($x->{_e}) && $MBI->_is_one($x->{_m}))
938 $dbd = 0; # disable shortcut
939 # we can use the cached value in these cases
940 if ($scale <= $LOG_10_A)
942 $x->bzero(); $x->badd($LOG_10);
943 $calc = 0; # no need to calc, but round
948 # disable the shortcut for 2, since we maybe have it cached
949 if (($MBI->_is_zero($x->{_e}) && $MBI->_is_two($x->{_m})))
951 $dbd = 0; # disable shortcut
952 # we can use the cached value in these cases
953 if ($scale <= $LOG_2_A)
955 $x->bzero(); $x->badd($LOG_2);
956 $calc = 0; # no need to calc, but round
961 # if $x = 0.1, we know the result must be 0-log(10)
962 if ($calc != 0 && $x->{_es} eq '-' && $MBI->_is_one($x->{_e}) &&
963 $MBI->_is_one($x->{_m}))
965 $dbd = 0; # disable shortcut
966 # we can use the cached value in these cases
967 if ($scale <= $LOG_10_A)
969 $x->bzero(); $x->bsub($LOG_10);
970 $calc = 0; # no need to calc, but round
974 return if $calc == 0; # already have the result
976 # default: these correction factors are undef and thus not used
977 my $l_10; # value of ln(10) to A of $scale
978 my $l_2; # value of ln(2) to A of $scale
980 # $x == 2 => 1, $x == 13 => 2, $x == 0.1 => 0, $x == 0.01 => -1
981 # so don't do this shortcut for 1 or 0
982 if (($dbd > 1) || ($dbd < 0))
984 # convert our cached value to an object if not already (avoid doing this
985 # at import() time, since not everybody needs this)
986 $LOG_10 = $self->new($LOG_10,undef,undef) unless ref $LOG_10;
988 #print "x = $x, dbd = $dbd, calc = $calc\n";
989 # got more than one digit before the dot, or more than one zero after the
991 # log(123) == log(1.23) + log(10) * 2
992 # log(0.0123) == log(1.23) - log(10) * 2
994 if ($scale <= $LOG_10_A)
997 $l_10 = $LOG_10->copy(); # copy for mul
1001 # else: slower, compute it (but don't cache it, because it could be big)
1002 # also disable downgrade for this code path
1003 local $Math::BigFloat::downgrade = undef;
1004 $l_10 = $self->new(10)->blog(undef,$scale); # scale+4, actually
1006 $dbd-- if ($dbd > 1); # 20 => dbd=2, so make it dbd=1
1007 $l_10->bmul( $self->new($dbd)); # log(10) * (digits_before_dot-1)
1014 ($x->{_e}, $x->{_es}) =
1015 _e_sub( $x->{_e}, $MBI->_new($dbd), $x->{_es}, $dbd_sign); # 123 => 1.23
1019 # Now: 0.1 <= $x < 10 (and possible correction in l_10)
1021 ### Since $x in the range 0.5 .. 1.5 is MUCH faster, we do a repeated div
1022 ### or mul by 2 (maximum times 3, since x < 10 and x > 0.1)
1024 $HALF = $self->new($HALF) unless ref($HALF);
1026 my $twos = 0; # default: none (0 times)
1027 my $two = $self->new(2);
1028 while ($x->bacmp($HALF) <= 0)
1030 $twos--; $x->bmul($two);
1032 while ($x->bacmp($two) >= 0)
1034 $twos++; $x->bdiv($two,$scale+4); # keep all digits
1036 # $twos > 0 => did mul 2, < 0 => did div 2 (never both)
1037 # calculate correction factor based on ln(2)
1040 $LOG_2 = $self->new($LOG_2,undef,undef) unless ref $LOG_2;
1041 if ($scale <= $LOG_2_A)
1044 $l_2 = $LOG_2->copy(); # copy for mul
1048 # else: slower, compute it (but don't cache it, because it could be big)
1049 # also disable downgrade for this code path
1050 local $Math::BigFloat::downgrade = undef;
1051 $l_2 = $two->blog(undef,$scale); # scale+4, actually
1053 $l_2->bmul($twos); # * -2 => subtract, * 2 => add
1056 $self->_log($x,$scale); # need to do the "normal" way
1057 $x->badd($l_10) if defined $l_10; # correct it by ln(10)
1058 $x->badd($l_2) if defined $l_2; # and maybe by ln(2)
1059 # all done, $x contains now the result
1064 # (BFLOAT or num_str, BFLOAT or num_str) return BFLOAT
1065 # does not modify arguments, but returns new object
1066 # Lowest Common Multiplicator
1068 my ($self,@arg) = objectify(0,@_);
1069 my $x = $self->new(shift @arg);
1070 while (@arg) { $x = _lcm($x,shift @arg); }
1076 # (BFLOAT or num_str, BFLOAT or num_str) return BINT
1077 # does not modify arguments, but returns new object
1078 # GCD -- Euclids algorithm Knuth Vol 2 pg 296
1080 my ($self,@arg) = objectify(0,@_);
1081 my $x = $self->new(shift @arg);
1082 while (@arg) { $x = _gcd($x,shift @arg); }
1086 ##############################################################################
1090 # Internal helper sub to take two positive integers and their signs and
1091 # then add them. Input ($CALC,$CALC,('+'|'-'),('+'|'-')),
1092 # output ($CALC,('+'|'-'))
1093 my ($x,$y,$xs,$ys) = @_;
1095 # if the signs are equal we can add them (-5 + -3 => -(5 + 3) => -8)
1098 $x = $MBI->_add ($x, $y ); # a+b
1099 # the sign follows $xs
1103 my $a = $MBI->_acmp($x,$y);
1106 $x = $MBI->_sub ($x , $y); # abs sub
1110 $x = $MBI->_zero(); # result is 0
1115 $x = $MBI->_sub ( $y, $x, 1 ); # abs sub
1123 # Internal helper sub to take two positive integers and their signs and
1124 # then subtract them. Input ($CALC,$CALC,('+'|'-'),('+'|'-')),
1125 # output ($CALC,('+'|'-'))
1126 my ($x,$y,$xs,$ys) = @_;
1130 _e_add($x,$y,$xs,$ys); # call add (does subtract now)
1133 ###############################################################################
1134 # is_foo methods (is_negative, is_positive are inherited from BigInt)
1138 # return true if arg (BFLOAT or num_str) is an integer
1139 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
1141 return 1 if ($x->{sign} =~ /^[+-]$/) && # NaN and +-inf aren't
1142 $x->{_es} eq '+'; # 1e-1 => no integer
1148 # return true if arg (BFLOAT or num_str) is zero
1149 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
1151 return 1 if $x->{sign} eq '+' && $MBI->_is_zero($x->{_m});
1157 # return true if arg (BFLOAT or num_str) is +1 or -1 if signis given
1158 my ($self,$x,$sign) = ref($_[0]) ? (undef,@_) : objectify(1,@_);
1160 $sign = '+' if !defined $sign || $sign ne '-';
1162 if ($x->{sign} eq $sign &&
1163 $MBI->_is_zero($x->{_e}) && $MBI->_is_one($x->{_m}));
1169 # return true if arg (BFLOAT or num_str) is odd or false if even
1170 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
1172 return 1 if ($x->{sign} =~ /^[+-]$/) && # NaN & +-inf aren't
1173 ($MBI->_is_zero($x->{_e}) && $MBI->_is_odd($x->{_m}));
1179 # return true if arg (BINT or num_str) is even or false if odd
1180 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
1182 return 0 if $x->{sign} !~ /^[+-]$/; # NaN & +-inf aren't
1183 return 1 if ($x->{_es} eq '+' # 123.45 is never
1184 && $MBI->_is_even($x->{_m})); # but 1200 is
1190 # multiply two numbers -- stolen from Knuth Vol 2 pg 233
1191 # (BINT or num_str, BINT or num_str) return BINT
1194 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
1195 # objectify is costly, so avoid it
1196 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
1198 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
1201 return $x->bnan() if (($x->{sign} eq $nan) || ($y->{sign} eq $nan));
1204 if (($x->{sign} =~ /^[+-]inf$/) || ($y->{sign} =~ /^[+-]inf$/))
1206 return $x->bnan() if $x->is_zero() || $y->is_zero();
1207 # result will always be +-inf:
1208 # +inf * +/+inf => +inf, -inf * -/-inf => +inf
1209 # +inf * -/-inf => -inf, -inf * +/+inf => -inf
1210 return $x->binf() if ($x->{sign} =~ /^\+/ && $y->{sign} =~ /^\+/);
1211 return $x->binf() if ($x->{sign} =~ /^-/ && $y->{sign} =~ /^-/);
1212 return $x->binf('-');
1215 return $x->bzero() if $x->is_zero() || $y->is_zero();
1217 return $upgrade->bmul($x,$y,$a,$p,$r) if defined $upgrade &&
1218 ((!$x->isa($self)) || (!$y->isa($self)));
1220 # aEb * cEd = (a*c)E(b+d)
1221 $MBI->_mul($x->{_m},$y->{_m});
1222 ($x->{_e}, $x->{_es}) = _e_add($x->{_e}, $y->{_e}, $x->{_es}, $y->{_es});
1225 $x->{sign} = $x->{sign} ne $y->{sign} ? '-' : '+';
1226 return $x->bnorm()->round($a,$p,$r,$y);
1231 # (dividend: BFLOAT or num_str, divisor: BFLOAT or num_str) return
1232 # (BFLOAT,BFLOAT) (quo,rem) or BFLOAT (only rem)
1235 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
1236 # objectify is costly, so avoid it
1237 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
1239 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
1242 return $self->_div_inf($x,$y)
1243 if (($x->{sign} !~ /^[+-]$/) || ($y->{sign} !~ /^[+-]$/) || $y->is_zero());
1245 # x== 0 # also: or y == 1 or y == -1
1246 return wantarray ? ($x,$self->bzero()) : $x if $x->is_zero();
1249 return $upgrade->bdiv($upgrade->new($x),$y,$a,$p,$r) if defined $upgrade;
1251 # we need to limit the accuracy to protect against overflow
1253 my (@params,$scale);
1254 ($x,@params) = $x->_find_round_parameters($a,$p,$r,$y);
1256 return $x if $x->is_nan(); # error in _find_round_parameters?
1258 # no rounding at all, so must use fallback
1259 if (scalar @params == 0)
1261 # simulate old behaviour
1262 $params[0] = $self->div_scale(); # and round to it as accuracy
1263 $scale = $params[0]+4; # at least four more for proper round
1264 $params[2] = $r; # round mode by caller or undef
1265 $fallback = 1; # to clear a/p afterwards
1269 # the 4 below is empirical, and there might be cases where it is not
1271 $scale = abs($params[0] || $params[1]) + 4; # take whatever is defined
1274 my $rem; $rem = $self->bzero() if wantarray;
1276 $y = $self->new($y) unless $y->isa('Math::BigFloat');
1278 my $lx = $MBI->_len($x->{_m}); my $ly = $MBI->_len($y->{_m});
1279 $scale = $lx if $lx > $scale;
1280 $scale = $ly if $ly > $scale;
1281 my $diff = $ly - $lx;
1282 $scale += $diff if $diff > 0; # if lx << ly, but not if ly << lx!
1284 # cases like $x /= $x (but not $x /= $y!) were wrong due to modifying $x
1286 if (overload::StrVal($x) eq overload::StrVal($y))
1288 $x->bone(); # x/x => 1, rem 0
1293 # make copy of $x in case of list context for later reminder calculation
1294 if (wantarray && !$y->is_one())
1299 $x->{sign} = $x->{sign} ne $y->sign() ? '-' : '+';
1301 # check for / +-1 ( +/- 1E0)
1304 # promote BigInts and it's subclasses (except when already a BigFloat)
1305 $y = $self->new($y) unless $y->isa('Math::BigFloat');
1307 # calculate the result to $scale digits and then round it
1308 # a * 10 ** b / c * 10 ** d => a/c * 10 ** (b-d)
1309 $MBI->_lsft($x->{_m},$MBI->_new($scale),10);
1310 $MBI->_div ($x->{_m},$y->{_m}); # a/c
1312 # correct exponent of $x
1313 ($x->{_e},$x->{_es}) = _e_sub($x->{_e}, $y->{_e}, $x->{_es}, $y->{_es});
1314 # correct for 10**scale
1315 ($x->{_e},$x->{_es}) = _e_sub($x->{_e}, $MBI->_new($scale), $x->{_es}, '+');
1316 $x->bnorm(); # remove trailing 0's
1318 } # ende else $x != $y
1320 # shortcut to not run through _find_round_parameters again
1321 if (defined $params[0])
1323 delete $x->{_a}; # clear before round
1324 $x->bround($params[0],$params[2]); # then round accordingly
1328 delete $x->{_p}; # clear before round
1329 $x->bfround($params[1],$params[2]); # then round accordingly
1333 # clear a/p after round, since user did not request it
1334 delete $x->{_a}; delete $x->{_p};
1341 $rem->bmod($y,@params); # copy already done
1345 # clear a/p after round, since user did not request it
1346 delete $rem->{_a}; delete $rem->{_p};
1355 # (dividend: BFLOAT or num_str, divisor: BFLOAT or num_str) return reminder
1358 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
1359 # objectify is costly, so avoid it
1360 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
1362 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
1365 # handle NaN, inf, -inf
1366 if (($x->{sign} !~ /^[+-]$/) || ($y->{sign} !~ /^[+-]$/))
1368 my ($d,$re) = $self->SUPER::_div_inf($x,$y);
1369 $x->{sign} = $re->{sign};
1370 $x->{_e} = $re->{_e};
1371 $x->{_m} = $re->{_m};
1372 return $x->round($a,$p,$r,$y);
1376 return $x->bnan() if $x->is_zero();
1379 return $x->bzero() if $y->is_one() || $x->is_zero();
1381 my $cmp = $x->bacmp($y); # equal or $x < $y?
1382 return $x->bzero($a,$p) if $cmp == 0; # $x == $y => result 0
1384 # only $y of the operands negative?
1385 my $neg = 0; $neg = 1 if $x->{sign} ne $y->{sign};
1387 $x->{sign} = $y->{sign}; # calc sign first
1388 return $x->round($a,$p,$r) if $cmp < 0 && $neg == 0; # $x < $y => result $x
1390 my $ym = $MBI->_copy($y->{_m});
1393 $MBI->_lsft( $ym, $y->{_e}, 10)
1394 if $y->{_es} eq '+' && !$MBI->_is_zero($y->{_e});
1396 # if $y has digits after dot
1397 my $shifty = 0; # correct _e of $x by this
1398 if ($y->{_es} eq '-') # has digits after dot
1400 # 123 % 2.5 => 1230 % 25 => 5 => 0.5
1401 $shifty = $MBI->_num($y->{_e}); # no more digits after dot
1402 $MBI->_lsft($x->{_m}, $y->{_e}, 10);# 123 => 1230, $y->{_m} is already 25
1404 # $ym is now mantissa of $y based on exponent 0
1406 my $shiftx = 0; # correct _e of $x by this
1407 if ($x->{_es} eq '-') # has digits after dot
1409 # 123.4 % 20 => 1234 % 200
1410 $shiftx = $MBI->_num($x->{_e}); # no more digits after dot
1411 $MBI->_lsft($ym, $x->{_e}, 10); # 123 => 1230
1413 # 123e1 % 20 => 1230 % 20
1414 if ($x->{_es} eq '+' && !$MBI->_is_zero($x->{_e}))
1416 $MBI->_lsft( $x->{_m}, $x->{_e},10); # es => '+' here
1419 $x->{_e} = $MBI->_new($shiftx);
1421 $x->{_es} = '-' if $shiftx != 0 || $shifty != 0;
1422 $MBI->_add( $x->{_e}, $MBI->_new($shifty)) if $shifty != 0;
1424 # now mantissas are equalized, exponent of $x is adjusted, so calc result
1426 $x->{_m} = $MBI->_mod( $x->{_m}, $ym);
1428 $x->{sign} = '+' if $MBI->_is_zero($x->{_m}); # fix sign for -0
1431 if ($neg != 0) # one of them negative => correct in place
1434 $x->{_m} = $r->{_m};
1435 $x->{_e} = $r->{_e};
1436 $x->{_es} = $r->{_es};
1437 $x->{sign} = '+' if $MBI->_is_zero($x->{_m}); # fix sign for -0
1441 $x->round($a,$p,$r,$y); # round and return
1446 # calculate $y'th root of $x
1449 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
1450 # objectify is costly, so avoid it
1451 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
1453 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
1456 # NaN handling: $x ** 1/0, x or y NaN, or y inf/-inf or y == 0
1457 return $x->bnan() if $x->{sign} !~ /^\+/ || $y->is_zero() ||
1458 $y->{sign} !~ /^\+$/;
1460 return $x if $x->is_zero() || $x->is_one() || $x->is_inf() || $y->is_one();
1462 # we need to limit the accuracy to protect against overflow
1464 my (@params,$scale);
1465 ($x,@params) = $x->_find_round_parameters($a,$p,$r);
1467 return $x if $x->is_nan(); # error in _find_round_parameters?
1469 # no rounding at all, so must use fallback
1470 if (scalar @params == 0)
1472 # simulate old behaviour
1473 $params[0] = $self->div_scale(); # and round to it as accuracy
1474 $scale = $params[0]+4; # at least four more for proper round
1475 $params[2] = $r; # iound mode by caller or undef
1476 $fallback = 1; # to clear a/p afterwards
1480 # the 4 below is empirical, and there might be cases where it is not
1482 $scale = abs($params[0] || $params[1]) + 4; # take whatever is defined
1485 # when user set globals, they would interfere with our calculation, so
1486 # disable them and later re-enable them
1488 my $abr = "$self\::accuracy"; my $ab = $$abr; $$abr = undef;
1489 my $pbr = "$self\::precision"; my $pb = $$pbr; $$pbr = undef;
1490 # we also need to disable any set A or P on $x (_find_round_parameters took
1491 # them already into account), since these would interfere, too
1492 delete $x->{_a}; delete $x->{_p};
1493 # need to disable $upgrade in BigInt, to avoid deep recursion
1494 local $Math::BigInt::upgrade = undef; # should be really parent class vs MBI
1496 # remember sign and make $x positive, since -4 ** (1/2) => -2
1497 my $sign = 0; $sign = 1 if $x->{sign} eq '-'; $x->{sign} = '+';
1500 if ($y->isa('Math::BigFloat'))
1502 $is_two = ($y->{sign} eq '+' && $MBI->_is_two($y->{_m}) && $MBI->_is_zero($y->{_e}));
1506 $is_two = ($y == 2);
1509 # normal square root if $y == 2:
1512 $x->bsqrt($scale+4);
1514 elsif ($y->is_one('-'))
1517 my $u = $self->bone()->bdiv($x,$scale);
1518 # copy private parts over
1519 $x->{_m} = $u->{_m};
1520 $x->{_e} = $u->{_e};
1521 $x->{_es} = $u->{_es};
1525 # calculate the broot() as integer result first, and if it fits, return
1526 # it rightaway (but only if $x and $y are integer):
1528 my $done = 0; # not yet
1529 if ($y->is_int() && $x->is_int())
1531 my $i = $MBI->_copy( $x->{_m} );
1532 $MBI->_lsft( $i, $x->{_e}, 10 ) unless $MBI->_is_zero($x->{_e});
1533 my $int = Math::BigInt->bzero();
1535 $int->broot($y->as_number());
1537 if ($int->copy()->bpow($y) == $x)
1539 # found result, return it
1540 $x->{_m} = $int->{value};
1541 $x->{_e} = $MBI->_zero();
1549 my $u = $self->bone()->bdiv($y,$scale+4);
1550 delete $u->{_a}; delete $u->{_p}; # otherwise it conflicts
1551 $x->bpow($u,$scale+4); # el cheapo
1554 $x->bneg() if $sign == 1;
1556 # shortcut to not run through _find_round_parameters again
1557 if (defined $params[0])
1559 $x->bround($params[0],$params[2]); # then round accordingly
1563 $x->bfround($params[1],$params[2]); # then round accordingly
1567 # clear a/p after round, since user did not request it
1568 delete $x->{_a}; delete $x->{_p};
1571 $$abr = $ab; $$pbr = $pb;
1577 # calculate square root
1578 my ($self,$x,$a,$p,$r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
1580 return $x->bnan() if $x->{sign} !~ /^[+]/; # NaN, -inf or < 0
1581 return $x if $x->{sign} eq '+inf'; # sqrt(inf) == inf
1582 return $x->round($a,$p,$r) if $x->is_zero() || $x->is_one();
1584 # we need to limit the accuracy to protect against overflow
1586 my (@params,$scale);
1587 ($x,@params) = $x->_find_round_parameters($a,$p,$r);
1589 return $x if $x->is_nan(); # error in _find_round_parameters?
1591 # no rounding at all, so must use fallback
1592 if (scalar @params == 0)
1594 # simulate old behaviour
1595 $params[0] = $self->div_scale(); # and round to it as accuracy
1596 $scale = $params[0]+4; # at least four more for proper round
1597 $params[2] = $r; # round mode by caller or undef
1598 $fallback = 1; # to clear a/p afterwards
1602 # the 4 below is empirical, and there might be cases where it is not
1604 $scale = abs($params[0] || $params[1]) + 4; # take whatever is defined
1607 # when user set globals, they would interfere with our calculation, so
1608 # disable them and later re-enable them
1610 my $abr = "$self\::accuracy"; my $ab = $$abr; $$abr = undef;
1611 my $pbr = "$self\::precision"; my $pb = $$pbr; $$pbr = undef;
1612 # we also need to disable any set A or P on $x (_find_round_parameters took
1613 # them already into account), since these would interfere, too
1614 delete $x->{_a}; delete $x->{_p};
1615 # need to disable $upgrade in BigInt, to avoid deep recursion
1616 local $Math::BigInt::upgrade = undef; # should be really parent class vs MBI
1618 my $i = $MBI->_copy( $x->{_m} );
1619 $MBI->_lsft( $i, $x->{_e}, 10 ) unless $MBI->_is_zero($x->{_e});
1620 my $xas = Math::BigInt->bzero();
1623 my $gs = $xas->copy()->bsqrt(); # some guess
1625 if (($x->{_es} ne '-') # guess can't be accurate if there are
1626 # digits after the dot
1627 && ($xas->bacmp($gs * $gs) == 0)) # guess hit the nail on the head?
1629 # exact result, copy result over to keep $x
1630 $x->{_m} = $gs->{value}; $x->{_e} = $MBI->_zero(); $x->{_es} = '+';
1632 # shortcut to not run through _find_round_parameters again
1633 if (defined $params[0])
1635 $x->bround($params[0],$params[2]); # then round accordingly
1639 $x->bfround($params[1],$params[2]); # then round accordingly
1643 # clear a/p after round, since user did not request it
1644 delete $x->{_a}; delete $x->{_p};
1646 # re-enable A and P, upgrade is taken care of by "local"
1647 ${"$self\::accuracy"} = $ab; ${"$self\::precision"} = $pb;
1651 # sqrt(2) = 1.4 because sqrt(2*100) = 1.4*10; so we can increase the accuracy
1652 # of the result by multipyling the input by 100 and then divide the integer
1653 # result of sqrt(input) by 10. Rounding afterwards returns the real result.
1655 # The following steps will transform 123.456 (in $x) into 123456 (in $y1)
1656 my $y1 = $MBI->_copy($x->{_m});
1658 my $length = $MBI->_len($y1);
1660 # Now calculate how many digits the result of sqrt(y1) would have
1661 my $digits = int($length / 2);
1663 # But we need at least $scale digits, so calculate how many are missing
1664 my $shift = $scale - $digits;
1666 # That should never happen (we take care of integer guesses above)
1667 # $shift = 0 if $shift < 0;
1669 # Multiply in steps of 100, by shifting left two times the "missing" digits
1670 my $s2 = $shift * 2;
1672 # We now make sure that $y1 has the same odd or even number of digits than
1673 # $x had. So when _e of $x is odd, we must shift $y1 by one digit left,
1674 # because we always must multiply by steps of 100 (sqrt(100) is 10) and not
1675 # steps of 10. The length of $x does not count, since an even or odd number
1676 # of digits before the dot is not changed by adding an even number of digits
1677 # after the dot (the result is still odd or even digits long).
1678 $s2++ if $MBI->_is_odd($x->{_e});
1680 $MBI->_lsft( $y1, $MBI->_new($s2), 10);
1682 # now take the square root and truncate to integer
1683 $y1 = $MBI->_sqrt($y1);
1685 # By "shifting" $y1 right (by creating a negative _e) we calculate the final
1686 # result, which is than later rounded to the desired scale.
1688 # calculate how many zeros $x had after the '.' (or before it, depending
1689 # on sign of $dat, the result should have half as many:
1690 my $dat = $MBI->_num($x->{_e});
1691 $dat = -$dat if $x->{_es} eq '-';
1696 # no zeros after the dot (e.g. 1.23, 0.49 etc)
1697 # preserve half as many digits before the dot than the input had
1698 # (but round this "up")
1699 $dat = int(($dat+1)/2);
1703 $dat = int(($dat)/2);
1705 $dat -= $MBI->_len($y1);
1709 $x->{_e} = $MBI->_new( $dat );
1714 $x->{_e} = $MBI->_new( $dat );
1720 # shortcut to not run through _find_round_parameters again
1721 if (defined $params[0])
1723 $x->bround($params[0],$params[2]); # then round accordingly
1727 $x->bfround($params[1],$params[2]); # then round accordingly
1731 # clear a/p after round, since user did not request it
1732 delete $x->{_a}; delete $x->{_p};
1735 $$abr = $ab; $$pbr = $pb;
1741 # (BFLOAT or num_str, BFLOAT or num_str) return BFLOAT
1742 # compute factorial number, modifies first argument
1745 my ($self,$x,@r) = (ref($_[0]),@_);
1746 # objectify is costly, so avoid it
1747 ($self,$x,@r) = objectify(1,@_) if !ref($x);
1749 return $x if $x->{sign} eq '+inf'; # inf => inf
1751 if (($x->{sign} ne '+') || # inf, NaN, <0 etc => NaN
1752 ($x->{_es} ne '+')); # digits after dot?
1754 # use BigInt's bfac() for faster calc
1755 if (! $MBI->_is_zero($x->{_e}))
1757 $MBI->_lsft($x->{_m}, $x->{_e},10); # change 12e1 to 120e0
1758 $x->{_e} = $MBI->_zero(); # normalize
1761 $MBI->_fac($x->{_m}); # calculate factorial
1762 $x->bnorm()->round(@r); # norm again and round result
1767 # Calculate a power where $y is a non-integer, like 2 ** 0.5
1768 my ($x,$y,$a,$p,$r) = @_;
1771 # if $y == 0.5, it is sqrt($x)
1772 $HALF = $self->new($HALF) unless ref($HALF);
1773 return $x->bsqrt($a,$p,$r,$y) if $y->bcmp($HALF) == 0;
1776 # a ** x == e ** (x * ln a)
1780 # Taylor: | u u^2 u^3 |
1781 # x ** y = 1 + | --- + --- + ----- + ... |
1784 # we need to limit the accuracy to protect against overflow
1786 my ($scale,@params);
1787 ($x,@params) = $x->_find_round_parameters($a,$p,$r);
1789 return $x if $x->is_nan(); # error in _find_round_parameters?
1791 # no rounding at all, so must use fallback
1792 if (scalar @params == 0)
1794 # simulate old behaviour
1795 $params[0] = $self->div_scale(); # and round to it as accuracy
1796 $params[1] = undef; # disable P
1797 $scale = $params[0]+4; # at least four more for proper round
1798 $params[2] = $r; # round mode by caller or undef
1799 $fallback = 1; # to clear a/p afterwards
1803 # the 4 below is empirical, and there might be cases where it is not
1805 $scale = abs($params[0] || $params[1]) + 4; # take whatever is defined
1808 # when user set globals, they would interfere with our calculation, so
1809 # disable them and later re-enable them
1811 my $abr = "$self\::accuracy"; my $ab = $$abr; $$abr = undef;
1812 my $pbr = "$self\::precision"; my $pb = $$pbr; $$pbr = undef;
1813 # we also need to disable any set A or P on $x (_find_round_parameters took
1814 # them already into account), since these would interfere, too
1815 delete $x->{_a}; delete $x->{_p};
1816 # need to disable $upgrade in BigInt, to avoid deep recursion
1817 local $Math::BigInt::upgrade = undef;
1819 my ($limit,$v,$u,$below,$factor,$next,$over);
1821 $u = $x->copy()->blog(undef,$scale)->bmul($y);
1822 $v = $self->bone(); # 1
1823 $factor = $self->new(2); # 2
1824 $x->bone(); # first term: 1
1826 $below = $v->copy();
1829 $limit = $self->new("1E-". ($scale-1));
1833 # we calculate the next term, and add it to the last
1834 # when the next term is below our limit, it won't affect the outcome
1835 # anymore, so we stop
1836 $next = $over->copy()->bdiv($below,$scale);
1837 last if $next->bacmp($limit) <= 0;
1839 # calculate things for the next term
1840 $over *= $u; $below *= $factor; $factor->binc();
1842 last if $x->{sign} !~ /^[-+]$/;
1847 # shortcut to not run through _find_round_parameters again
1848 if (defined $params[0])
1850 $x->bround($params[0],$params[2]); # then round accordingly
1854 $x->bfround($params[1],$params[2]); # then round accordingly
1858 # clear a/p after round, since user did not request it
1859 delete $x->{_a}; delete $x->{_p};
1862 $$abr = $ab; $$pbr = $pb;
1868 # (BFLOAT or num_str, BFLOAT or num_str) return BFLOAT
1869 # compute power of two numbers, second arg is used as integer
1870 # modifies first argument
1873 my ($self,$x,$y,$a,$p,$r) = (ref($_[0]),@_);
1874 # objectify is costly, so avoid it
1875 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
1877 ($self,$x,$y,$a,$p,$r) = objectify(2,@_);
1880 return $x if $x->{sign} =~ /^[+-]inf$/;
1881 return $x->bnan() if $x->{sign} eq $nan || $y->{sign} eq $nan;
1882 return $x->bone() if $y->is_zero();
1883 return $x if $x->is_one() || $y->is_one();
1885 return $x->_pow($y,$a,$p,$r) if !$y->is_int(); # non-integer power
1887 my $y1 = $y->as_number()->{value}; # make CALC
1890 if ($x->{sign} eq '-' && $MBI->_is_one($x->{_m}) && $MBI->_is_zero($x->{_e}))
1892 # if $x == -1 and odd/even y => +1/-1 because +-1 ^ (+-1) => +-1
1893 return $MBI->_is_odd($y1) ? $x : $x->babs(1);
1897 return $x->bone() if $y->is_zero();
1898 return $x if $y->{sign} eq '+'; # 0**y => 0 (if not y <= 0)
1899 # 0 ** -y => 1 / (0 ** y) => 1 / 0! (1 / 0 => +inf)
1904 $new_sign = $y->is_odd() ? '-' : '+' if ($x->{sign} ne '+');
1906 # calculate $x->{_m} ** $y and $x->{_e} * $y separately (faster)
1907 $x->{_m} = $MBI->_pow( $x->{_m}, $y1);
1908 $MBI->_mul ($x->{_e}, $y1);
1910 $x->{sign} = $new_sign;
1912 if ($y->{sign} eq '-')
1914 # modify $x in place!
1915 my $z = $x->copy(); $x->bzero()->binc();
1916 return $x->bdiv($z,$a,$p,$r); # round in one go (might ignore y's A!)
1918 $x->round($a,$p,$r,$y);
1921 ###############################################################################
1922 # rounding functions
1926 # precision: round to the $Nth digit left (+$n) or right (-$n) from the '.'
1927 # $n == 0 means round to integer
1928 # expects and returns normalized numbers!
1929 my $x = shift; my $self = ref($x) || $x; $x = $self->new(shift) if !ref($x);
1931 return $x if $x->modify('bfround');
1933 my ($scale,$mode) = $x->_scale_p($self->precision(),$self->round_mode(),@_);
1934 return $x if !defined $scale; # no-op
1936 # never round a 0, +-inf, NaN
1939 $x->{_p} = $scale if !defined $x->{_p} || $x->{_p} < $scale; # -3 < -2
1942 return $x if $x->{sign} !~ /^[+-]$/;
1944 # don't round if x already has lower precision
1945 return $x if (defined $x->{_p} && $x->{_p} < 0 && $scale < $x->{_p});
1947 $x->{_p} = $scale; # remember round in any case
1948 delete $x->{_a}; # and clear A
1951 # round right from the '.'
1953 return $x if $x->{_es} eq '+'; # e >= 0 => nothing to round
1955 $scale = -$scale; # positive for simplicity
1956 my $len = $MBI->_len($x->{_m}); # length of mantissa
1958 # the following poses a restriction on _e, but if _e is bigger than a
1959 # scalar, you got other problems (memory etc) anyway
1960 my $dad = -(0+ ($x->{_es}.$MBI->_num($x->{_e}))); # digits after dot
1961 my $zad = 0; # zeros after dot
1962 $zad = $dad - $len if (-$dad < -$len); # for 0.00..00xxx style
1964 # p rint "scale $scale dad $dad zad $zad len $len\n";
1965 # number bsstr len zad dad
1966 # 0.123 123e-3 3 0 3
1967 # 0.0123 123e-4 3 1 4
1970 # 1.2345 12345e-4 5 0 4
1972 # do not round after/right of the $dad
1973 return $x if $scale > $dad; # 0.123, scale >= 3 => exit
1975 # round to zero if rounding inside the $zad, but not for last zero like:
1976 # 0.0065, scale -2, round last '0' with following '65' (scale == zad case)
1977 return $x->bzero() if $scale < $zad;
1978 if ($scale == $zad) # for 0.006, scale -3 and trunc
1984 # adjust round-point to be inside mantissa
1987 $scale = $scale-$zad;
1991 my $dbd = $len - $dad; $dbd = 0 if $dbd < 0; # digits before dot
1992 $scale = $dbd+$scale;
1998 # round left from the '.'
2000 # 123 => 100 means length(123) = 3 - $scale (2) => 1
2002 my $dbt = $MBI->_len($x->{_m});
2004 my $dbd = $dbt + ($x->{_es} . $MBI->_num($x->{_e}));
2005 # should be the same, so treat it as this
2006 $scale = 1 if $scale == 0;
2007 # shortcut if already integer
2008 return $x if $scale == 1 && $dbt <= $dbd;
2009 # maximum digits before dot
2014 # not enough digits before dot, so round to zero
2017 elsif ( $scale == $dbd )
2024 $scale = $dbd - $scale;
2027 # pass sign to bround for rounding modes '+inf' and '-inf'
2028 my $m = Math::BigInt->new( $x->{sign} . $MBI->_str($x->{_m}));
2029 $m->bround($scale,$mode);
2030 $x->{_m} = $m->{value}; # get our mantissa back
2036 # accuracy: preserve $N digits, and overwrite the rest with 0's
2037 my $x = shift; my $self = ref($x) || $x; $x = $self->new(shift) if !ref($x);
2039 if (($_[0] || 0) < 0)
2041 require Carp; Carp::croak ('bround() needs positive accuracy');
2044 my ($scale,$mode) = $x->_scale_a($self->accuracy(),$self->round_mode(),@_);
2045 return $x if !defined $scale; # no-op
2047 return $x if $x->modify('bround');
2049 # scale is now either $x->{_a}, $accuracy, or the user parameter
2050 # test whether $x already has lower accuracy, do nothing in this case
2051 # but do round if the accuracy is the same, since a math operation might
2052 # want to round a number with A=5 to 5 digits afterwards again
2053 return $x if defined $_[0] && defined $x->{_a} && $x->{_a} < $_[0];
2055 # scale < 0 makes no sense
2056 # never round a +-inf, NaN
2057 return $x if ($scale < 0) || $x->{sign} !~ /^[+-]$/;
2059 # 1: $scale == 0 => keep all digits
2060 # 2: never round a 0
2061 # 3: if we should keep more digits than the mantissa has, do nothing
2062 if ($scale == 0 || $x->is_zero() || $MBI->_len($x->{_m}) <= $scale)
2064 $x->{_a} = $scale if !defined $x->{_a} || $x->{_a} > $scale;
2068 # pass sign to bround for '+inf' and '-inf' rounding modes
2069 my $m = Math::BigInt->new( $x->{sign} . $MBI->_str($x->{_m}));
2071 $m->bround($scale,$mode); # round mantissa
2072 $x->{_m} = $m->{value}; # get our mantissa back
2073 $x->{_a} = $scale; # remember rounding
2074 delete $x->{_p}; # and clear P
2075 $x->bnorm(); # del trailing zeros gen. by bround()
2080 # return integer less or equal then $x
2081 my ($self,$x,$a,$p,$r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
2083 return $x if $x->modify('bfloor');
2085 return $x if $x->{sign} !~ /^[+-]$/; # nan, +inf, -inf
2087 # if $x has digits after dot
2088 if ($x->{_es} eq '-')
2090 $x->{_m} = $MBI->_rsft($x->{_m},$x->{_e},10); # cut off digits after dot
2091 $x->{_e} = $MBI->_zero(); # trunc/norm
2092 $x->{_es} = '+'; # abs e
2093 $MBI->_inc($x->{_m}) if $x->{sign} eq '-'; # increment if negative
2095 $x->round($a,$p,$r);
2100 # return integer greater or equal then $x
2101 my ($self,$x,$a,$p,$r) = ref($_[0]) ? (ref($_[0]),@_) : objectify(1,@_);
2103 return $x if $x->modify('bceil');
2104 return $x if $x->{sign} !~ /^[+-]$/; # nan, +inf, -inf
2106 # if $x has digits after dot
2107 if ($x->{_es} eq '-')
2109 $x->{_m} = $MBI->_rsft($x->{_m},$x->{_e},10); # cut off digits after dot
2110 $x->{_e} = $MBI->_zero(); # trunc/norm
2111 $x->{_es} = '+'; # abs e
2112 $MBI->_inc($x->{_m}) if $x->{sign} eq '+'; # increment if positive
2114 $x->round($a,$p,$r);
2119 # shift right by $y (divide by power of $n)
2122 my ($self,$x,$y,$n,$a,$p,$r) = (ref($_[0]),@_);
2123 # objectify is costly, so avoid it
2124 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
2126 ($self,$x,$y,$n,$a,$p,$r) = objectify(2,@_);
2129 return $x if $x->modify('brsft');
2130 return $x if $x->{sign} !~ /^[+-]$/; # nan, +inf, -inf
2132 $n = 2 if !defined $n; $n = $self->new($n);
2133 $x->bdiv($n->bpow($y),$a,$p,$r,$y);
2138 # shift left by $y (multiply by power of $n)
2141 my ($self,$x,$y,$n,$a,$p,$r) = (ref($_[0]),@_);
2142 # objectify is costly, so avoid it
2143 if ((!ref($_[0])) || (ref($_[0]) ne ref($_[1])))
2145 ($self,$x,$y,$n,$a,$p,$r) = objectify(2,@_);
2148 return $x if $x->modify('blsft');
2149 return $x if $x->{sign} !~ /^[+-]$/; # nan, +inf, -inf
2151 $n = 2 if !defined $n; $n = $self->new($n);
2152 $x->bmul($n->bpow($y),$a,$p,$r,$y);
2155 ###############################################################################
2159 # going through AUTOLOAD for every DESTROY is costly, avoid it by empty sub
2164 # make fxxx and bxxx both work by selectively mapping fxxx() to MBF::bxxx()
2165 # or falling back to MBI::bxxx()
2166 my $name = $AUTOLOAD;
2168 $name =~ s/(.*):://; # split package
2169 my $c = $1 || $class;
2171 $c->import() if $IMPORT == 0;
2172 if (!method_alias($name))
2176 # delayed load of Carp and avoid recursion
2178 Carp::croak ("$c: Can't call a method without name");
2180 if (!method_hand_up($name))
2182 # delayed load of Carp and avoid recursion
2184 Carp::croak ("Can't call $c\-\>$name, not a valid method");
2186 # try one level up, but subst. bxxx() for fxxx() since MBI only got bxxx()
2188 return &{"Math::BigInt"."::$name"}(@_);
2190 my $bname = $name; $bname =~ s/^f/b/;
2198 # return a copy of the exponent
2199 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2201 if ($x->{sign} !~ /^[+-]$/)
2203 my $s = $x->{sign}; $s =~ s/^[+-]//;
2204 return Math::BigInt->new($s); # -inf, +inf => +inf
2206 Math::BigInt->new( $x->{_es} . $MBI->_str($x->{_e}));
2211 # return a copy of the mantissa
2212 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2214 if ($x->{sign} !~ /^[+-]$/)
2216 my $s = $x->{sign}; $s =~ s/^[+]//;
2217 return Math::BigInt->new($s); # -inf, +inf => +inf
2219 my $m = Math::BigInt->new( $MBI->_str($x->{_m}));
2220 $m->bneg() if $x->{sign} eq '-';
2227 # return a copy of both the exponent and the mantissa
2228 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2230 if ($x->{sign} !~ /^[+-]$/)
2232 my $s = $x->{sign}; $s =~ s/^[+]//; my $se = $s; $se =~ s/^[-]//;
2233 return ($self->new($s),$self->new($se)); # +inf => inf and -inf,+inf => inf
2235 my $m = Math::BigInt->bzero();
2236 $m->{value} = $MBI->_copy($x->{_m});
2237 $m->bneg() if $x->{sign} eq '-';
2238 ($m, Math::BigInt->new( $x->{_es} . $MBI->_num($x->{_e}) ));
2241 ##############################################################################
2242 # private stuff (internal use only)
2248 my $lib = ''; my @a;
2250 for ( my $i = 0; $i < $l ; $i++)
2252 if ( $_[$i] eq ':constant' )
2254 # This causes overlord er load to step in. 'binary' and 'integer'
2255 # are handled by BigInt.
2256 overload::constant float => sub { $self->new(shift); };
2258 elsif ($_[$i] eq 'upgrade')
2260 # this causes upgrading
2261 $upgrade = $_[$i+1]; # or undef to disable
2264 elsif ($_[$i] eq 'downgrade')
2266 # this causes downgrading
2267 $downgrade = $_[$i+1]; # or undef to disable
2270 elsif ($_[$i] eq 'lib')
2272 # alternative library
2273 $lib = $_[$i+1] || ''; # default Calc
2276 elsif ($_[$i] eq 'with')
2278 # alternative class for our private parts()
2279 # XXX: no longer supported
2280 # $MBI = $_[$i+1] || 'Math::BigInt';
2289 # let use Math::BigInt lib => 'GMP'; use Math::BigFloat; still work
2290 my $mbilib = eval { Math::BigInt->config()->{lib} };
2291 if ((defined $mbilib) && ($MBI eq 'Math::BigInt::Calc'))
2293 # MBI already loaded
2294 Math::BigInt->import('lib',"$lib,$mbilib", 'objectify');
2298 # MBI not loaded, or with ne "Math::BigInt::Calc"
2299 $lib .= ",$mbilib" if defined $mbilib;
2300 $lib =~ s/^,//; # don't leave empty
2301 # replacement library can handle lib statement, but also could ignore it
2304 # Perl < 5.6.0 dies with "out of memory!" when eval() and ':constant' is
2305 # used in the same script, or eval inside import().
2306 require Math::BigInt;
2307 Math::BigInt->import( lib => $lib, 'objectify' );
2311 my $rc = "use Math::BigInt lib => '$lib', 'objectify';";
2317 require Carp; Carp::croak ("Couldn't load $lib: $! $@");
2319 $MBI = Math::BigInt->config()->{lib};
2321 # any non :constant stuff is handled by our parent, Exporter
2322 # even if @_ is empty, to give it a chance
2323 $self->SUPER::import(@a); # for subclasses
2324 $self->export_to_level(1,$self,@a); # need this, too
2329 # adjust m and e so that m is smallest possible
2330 # round number according to accuracy and precision settings
2331 my ($self,$x) = ref($_[0]) ? (undef,$_[0]) : objectify(1,@_);
2333 return $x if $x->{sign} !~ /^[+-]$/; # inf, nan etc
2335 my $zeros = $MBI->_zeros($x->{_m}); # correct for trailing zeros
2338 my $z = $MBI->_new($zeros);
2339 $x->{_m} = $MBI->_rsft ($x->{_m}, $z, 10);
2340 if ($x->{_es} eq '-')
2342 if ($MBI->_acmp($x->{_e},$z) >= 0)
2344 $x->{_e} = $MBI->_sub ($x->{_e}, $z);
2345 $x->{_es} = '+' if $MBI->_is_zero($x->{_e});
2349 $x->{_e} = $MBI->_sub ( $MBI->_copy($z), $x->{_e});
2355 $x->{_e} = $MBI->_add ($x->{_e}, $z);
2360 # $x can only be 0Ey if there are no trailing zeros ('0' has 0 trailing
2361 # zeros). So, for something like 0Ey, set y to 1, and -0 => +0
2362 $x->{sign} = '+', $x->{_es} = '+', $x->{_e} = $MBI->_one()
2363 if $MBI->_is_zero($x->{_m});
2366 $x; # MBI bnorm is no-op, so dont call it
2369 ##############################################################################
2373 # return number as hexadecimal string (only for integers defined)
2374 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2376 return $x->bstr() if $x->{sign} !~ /^[+-]$/; # inf, nan etc
2377 return '0x0' if $x->is_zero();
2379 return $nan if $x->{_es} ne '+'; # how to do 1e-1 in hex!?
2381 my $z = $MBI->_copy($x->{_m});
2382 if (! $MBI->_is_zero($x->{_e})) # > 0
2384 $MBI->_lsft( $z, $x->{_e},10);
2386 $z = Math::BigInt->new( $x->{sign} . $MBI->_num($z));
2392 # return number as binary digit string (only for integers defined)
2393 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2395 return $x->bstr() if $x->{sign} !~ /^[+-]$/; # inf, nan etc
2396 return '0b0' if $x->is_zero();
2398 return $nan if $x->{_es} ne '+'; # how to do 1e-1 in hex!?
2400 my $z = $MBI->_copy($x->{_m});
2401 if (! $MBI->_is_zero($x->{_e})) # > 0
2403 $MBI->_lsft( $z, $x->{_e},10);
2405 $z = Math::BigInt->new( $x->{sign} . $MBI->_num($z));
2411 # return copy as a bigint representation of this BigFloat number
2412 my ($self,$x) = ref($_[0]) ? (ref($_[0]),$_[0]) : objectify(1,@_);
2414 my $z = $MBI->_copy($x->{_m});
2415 if ($x->{_es} eq '-') # < 0
2417 $MBI->_rsft( $z, $x->{_e},10);
2419 elsif (! $MBI->_is_zero($x->{_e})) # > 0
2421 $MBI->_lsft( $z, $x->{_e},10);
2423 $z = Math::BigInt->new( $x->{sign} . $MBI->_num($z));
2430 my $class = ref($x) || $x;
2431 $x = $class->new(shift) unless ref($x);
2433 return 1 if $MBI->_is_zero($x->{_m});
2435 my $len = $MBI->_len($x->{_m});
2436 $len += $MBI->_num($x->{_e}) if $x->{_es} eq '+';
2440 $t = $MBI->_num($x->{_e}) if $x->{_es} eq '-';
2451 Math::BigFloat - Arbitrary size floating point math package
2458 $x = Math::BigFloat->new($str); # defaults to 0
2459 $nan = Math::BigFloat->bnan(); # create a NotANumber
2460 $zero = Math::BigFloat->bzero(); # create a +0
2461 $inf = Math::BigFloat->binf(); # create a +inf
2462 $inf = Math::BigFloat->binf('-'); # create a -inf
2463 $one = Math::BigFloat->bone(); # create a +1
2464 $one = Math::BigFloat->bone('-'); # create a -1
2467 $x->is_zero(); # true if arg is +0
2468 $x->is_nan(); # true if arg is NaN
2469 $x->is_one(); # true if arg is +1
2470 $x->is_one('-'); # true if arg is -1
2471 $x->is_odd(); # true if odd, false for even
2472 $x->is_even(); # true if even, false for odd
2473 $x->is_pos(); # true if >= 0
2474 $x->is_neg(); # true if < 0
2475 $x->is_inf(sign); # true if +inf, or -inf (default is '+')
2477 $x->bcmp($y); # compare numbers (undef,<0,=0,>0)
2478 $x->bacmp($y); # compare absolutely (undef,<0,=0,>0)
2479 $x->sign(); # return the sign, either +,- or NaN
2480 $x->digit($n); # return the nth digit, counting from right
2481 $x->digit(-$n); # return the nth digit, counting from left
2483 # The following all modify their first argument. If you want to preserve
2484 # $x, use $z = $x->copy()->bXXX($y); See under L<CAVEATS> for why this is
2485 # neccessary when mixing $a = $b assigments with non-overloaded math.
2488 $x->bzero(); # set $i to 0
2489 $x->bnan(); # set $i to NaN
2490 $x->bone(); # set $x to +1
2491 $x->bone('-'); # set $x to -1
2492 $x->binf(); # set $x to inf
2493 $x->binf('-'); # set $x to -inf
2495 $x->bneg(); # negation
2496 $x->babs(); # absolute value
2497 $x->bnorm(); # normalize (no-op)
2498 $x->bnot(); # two's complement (bit wise not)
2499 $x->binc(); # increment x by 1
2500 $x->bdec(); # decrement x by 1
2502 $x->badd($y); # addition (add $y to $x)
2503 $x->bsub($y); # subtraction (subtract $y from $x)
2504 $x->bmul($y); # multiplication (multiply $x by $y)
2505 $x->bdiv($y); # divide, set $x to quotient
2506 # return (quo,rem) or quo if scalar
2508 $x->bmod($y); # modulus ($x % $y)
2509 $x->bpow($y); # power of arguments ($x ** $y)
2510 $x->blsft($y); # left shift
2511 $x->brsft($y); # right shift
2512 # return (quo,rem) or quo if scalar
2514 $x->blog(); # logarithm of $x to base e (Euler's number)
2515 $x->blog($base); # logarithm of $x to base $base (f.i. 2)
2517 $x->band($y); # bit-wise and
2518 $x->bior($y); # bit-wise inclusive or
2519 $x->bxor($y); # bit-wise exclusive or
2520 $x->bnot(); # bit-wise not (two's complement)
2522 $x->bsqrt(); # calculate square-root
2523 $x->broot($y); # $y'th root of $x (e.g. $y == 3 => cubic root)
2524 $x->bfac(); # factorial of $x (1*2*3*4*..$x)
2526 $x->bround($N); # accuracy: preserve $N digits
2527 $x->bfround($N); # precision: round to the $Nth digit
2529 $x->bfloor(); # return integer less or equal than $x
2530 $x->bceil(); # return integer greater or equal than $x
2532 # The following do not modify their arguments:
2534 bgcd(@values); # greatest common divisor
2535 blcm(@values); # lowest common multiplicator
2537 $x->bstr(); # return string
2538 $x->bsstr(); # return string in scientific notation
2540 $x->as_int(); # return $x as BigInt
2541 $x->exponent(); # return exponent as BigInt
2542 $x->mantissa(); # return mantissa as BigInt
2543 $x->parts(); # return (mantissa,exponent) as BigInt
2545 $x->length(); # number of digits (w/o sign and '.')
2546 ($l,$f) = $x->length(); # number of digits, and length of fraction
2548 $x->precision(); # return P of $x (or global, if P of $x undef)
2549 $x->precision($n); # set P of $x to $n
2550 $x->accuracy(); # return A of $x (or global, if A of $x undef)
2551 $x->accuracy($n); # set A $x to $n
2553 # these get/set the appropriate global value for all BigFloat objects
2554 Math::BigFloat->precision(); # Precision
2555 Math::BigFloat->accuracy(); # Accuracy
2556 Math::BigFloat->round_mode(); # rounding mode
2560 All operators (inlcuding basic math operations) are overloaded if you
2561 declare your big floating point numbers as
2563 $i = new Math::BigFloat '12_3.456_789_123_456_789E-2';
2565 Operations with overloaded operators preserve the arguments, which is
2566 exactly what you expect.
2568 =head2 Canonical notation
2570 Input to these routines are either BigFloat objects, or strings of the
2571 following four forms:
2585 C</^[+-]\d+E[+-]?\d+$/>
2589 C</^[+-]\d*\.\d+E[+-]?\d+$/>
2593 all with optional leading and trailing zeros and/or spaces. Additonally,
2594 numbers are allowed to have an underscore between any two digits.
2596 Empty strings as well as other illegal numbers results in 'NaN'.
2598 bnorm() on a BigFloat object is now effectively a no-op, since the numbers
2599 are always stored in normalized form. On a string, it creates a BigFloat
2604 Output values are BigFloat objects (normalized), except for bstr() and bsstr().
2606 The string output will always have leading and trailing zeros stripped and drop
2607 a plus sign. C<bstr()> will give you always the form with a decimal point,
2608 while C<bsstr()> (s for scientific) gives you the scientific notation.
2610 Input bstr() bsstr()
2612 ' -123 123 123' '-123123123' '-123123123E0'
2613 '00.0123' '0.0123' '123E-4'
2614 '123.45E-2' '1.2345' '12345E-4'
2615 '10E+3' '10000' '1E4'
2617 Some routines (C<is_odd()>, C<is_even()>, C<is_zero()>, C<is_one()>,
2618 C<is_nan()>) return true or false, while others (C<bcmp()>, C<bacmp()>)
2619 return either undef, <0, 0 or >0 and are suited for sort.
2621 Actual math is done by using the class defined with C<with => Class;> (which
2622 defaults to BigInts) to represent the mantissa and exponent.
2624 The sign C</^[+-]$/> is stored separately. The string 'NaN' is used to
2625 represent the result when input arguments are not numbers, as well as
2626 the result of dividing by zero.
2628 =head2 C<mantissa()>, C<exponent()> and C<parts()>
2630 C<mantissa()> and C<exponent()> return the said parts of the BigFloat
2631 as BigInts such that:
2633 $m = $x->mantissa();
2634 $e = $x->exponent();
2635 $y = $m * ( 10 ** $e );
2636 print "ok\n" if $x == $y;
2638 C<< ($m,$e) = $x->parts(); >> is just a shortcut giving you both of them.
2640 A zero is represented and returned as C<0E1>, B<not> C<0E0> (after Knuth).
2642 Currently the mantissa is reduced as much as possible, favouring higher
2643 exponents over lower ones (e.g. returning 1e7 instead of 10e6 or 10000000e0).
2644 This might change in the future, so do not depend on it.
2646 =head2 Accuracy vs. Precision
2648 See also: L<Rounding|Rounding>.
2650 Math::BigFloat supports both precision and accuracy. For a full documentation,
2651 examples and tips on these topics please see the large section in
2654 Since things like sqrt(2) or 1/3 must presented with a limited precision lest
2655 a operation consumes all resources, each operation produces no more than
2656 the requested number of digits.
2658 Please refer to BigInt's documentation for the precedence rules of which
2659 accuracy/precision setting will be used.
2661 If there is no gloabl precision set, B<and> the operation inquestion was not
2662 called with a requested precision or accuracy, B<and> the input $x has no
2663 accuracy or precision set, then a fallback parameter will be used. For
2664 historical reasons, it is called C<div_scale> and can be accessed via:
2666 $d = Math::BigFloat->div_scale(); # query
2667 Math::BigFloat->div_scale($n); # set to $n digits
2669 The default value is 40 digits.
2671 In case the result of one operation has more precision than specified,
2672 it is rounded. The rounding mode taken is either the default mode, or the one
2673 supplied to the operation after the I<scale>:
2675 $x = Math::BigFloat->new(2);
2676 Math::BigFloat->precision(5); # 5 digits max
2677 $y = $x->copy()->bdiv(3); # will give 0.66666
2678 $y = $x->copy()->bdiv(3,6); # will give 0.666666
2679 $y = $x->copy()->bdiv(3,6,'odd'); # will give 0.666667
2680 Math::BigFloat->round_mode('zero');
2681 $y = $x->copy()->bdiv(3,6); # will give 0.666666
2687 =item ffround ( +$scale )
2689 Rounds to the $scale'th place left from the '.', counting from the dot.
2690 The first digit is numbered 1.
2692 =item ffround ( -$scale )
2694 Rounds to the $scale'th place right from the '.', counting from the dot.
2698 Rounds to an integer.
2700 =item fround ( +$scale )
2702 Preserves accuracy to $scale digits from the left (aka significant digits)
2703 and pads the rest with zeros. If the number is between 1 and -1, the
2704 significant digits count from the first non-zero after the '.'
2706 =item fround ( -$scale ) and fround ( 0 )
2708 These are effectively no-ops.
2712 All rounding functions take as a second parameter a rounding mode from one of
2713 the following: 'even', 'odd', '+inf', '-inf', 'zero' or 'trunc'.
2715 The default rounding mode is 'even'. By using
2716 C<< Math::BigFloat->round_mode($round_mode); >> you can get and set the default
2717 mode for subsequent rounding. The usage of C<$Math::BigFloat::$round_mode> is
2718 no longer supported.
2719 The second parameter to the round functions then overrides the default
2722 The C<as_number()> function returns a BigInt from a Math::BigFloat. It uses
2723 'trunc' as rounding mode to make it equivalent to:
2728 You can override this by passing the desired rounding mode as parameter to
2731 $x = Math::BigFloat->new(2.5);
2732 $y = $x->as_number('odd'); # $y = 3
2738 =head1 Autocreating constants
2740 After C<use Math::BigFloat ':constant'> all the floating point constants
2741 in the given scope are converted to C<Math::BigFloat>. This conversion
2742 happens at compile time.
2746 perl -MMath::BigFloat=:constant -e 'print 2E-100,"\n"'
2748 prints the value of C<2E-100>. Note that without conversion of
2749 constants the expression 2E-100 will be calculated as normal floating point
2752 Please note that ':constant' does not affect integer constants, nor binary
2753 nor hexadecimal constants. Use L<bignum> or L<Math::BigInt> to get this to
2758 Math with the numbers is done (by default) by a module called
2759 Math::BigInt::Calc. This is equivalent to saying:
2761 use Math::BigFloat lib => 'Calc';
2763 You can change this by using:
2765 use Math::BigFloat lib => 'BitVect';
2767 The following would first try to find Math::BigInt::Foo, then
2768 Math::BigInt::Bar, and when this also fails, revert to Math::BigInt::Calc:
2770 use Math::BigFloat lib => 'Foo,Math::BigInt::Bar';
2772 Calc.pm uses as internal format an array of elements of some decimal base
2773 (usually 1e7, but this might be differen for some systems) with the least
2774 significant digit first, while BitVect.pm uses a bit vector of base 2, most
2775 significant bit first. Other modules might use even different means of
2776 representing the numbers. See the respective module documentation for further
2779 Please note that Math::BigFloat does B<not> use the denoted library itself,
2780 but it merely passes the lib argument to Math::BigInt. So, instead of the need
2783 use Math::BigInt lib => 'GMP';
2786 you can roll it all into one line:
2788 use Math::BigFloat lib => 'GMP';
2790 It is also possible to just require Math::BigFloat:
2792 require Math::BigFloat;
2794 This will load the neccessary things (like BigInt) when they are needed, and
2797 Use the lib, Luke! And see L<Using Math::BigInt::Lite> for more details than
2798 you ever wanted to know about loading a different library.
2800 =head2 Using Math::BigInt::Lite
2802 It is possible to use L<Math::BigInt::Lite> with Math::BigFloat:
2805 use Math::BigFloat with => 'Math::BigInt::Lite';
2807 There is no need to "use Math::BigInt" or "use Math::BigInt::Lite", but you
2808 can combine these if you want. For instance, you may want to use
2809 Math::BigInt objects in your main script, too.
2813 use Math::BigFloat with => 'Math::BigInt::Lite';
2815 Of course, you can combine this with the C<lib> parameter.
2818 use Math::BigFloat with => 'Math::BigInt::Lite', lib => 'GMP,Pari';
2820 There is no need for a "use Math::BigInt;" statement, even if you want to
2821 use Math::BigInt's, since Math::BigFloat will needs Math::BigInt and thus
2822 always loads it. But if you add it, add it B<before>:
2826 use Math::BigFloat with => 'Math::BigInt::Lite', lib => 'GMP,Pari';
2828 Notice that the module with the last C<lib> will "win" and thus
2829 it's lib will be used if the lib is available:
2832 use Math::BigInt lib => 'Bar,Baz';
2833 use Math::BigFloat with => 'Math::BigInt::Lite', lib => 'Foo';
2835 That would try to load Foo, Bar, Baz and Calc (in that order). Or in other
2836 words, Math::BigFloat will try to retain previously loaded libs when you
2837 don't specify it onem but if you specify one, it will try to load them.
2839 Actually, the lib loading order would be "Bar,Baz,Calc", and then
2840 "Foo,Bar,Baz,Calc", but independend of which lib exists, the result is the
2841 same as trying the latter load alone, except for the fact that one of Bar or
2842 Baz might be loaded needlessly in an intermidiate step (and thus hang around
2843 and waste memory). If neither Bar nor Baz exist (or don't work/compile), they
2844 will still be tried to be loaded, but this is not as time/memory consuming as
2845 actually loading one of them. Still, this type of usage is not recommended due
2848 The old way (loading the lib only in BigInt) still works though:
2851 use Math::BigInt lib => 'Bar,Baz';
2854 You can even load Math::BigInt afterwards:
2858 use Math::BigInt lib => 'Bar,Baz';
2860 But this has the same problems like #5, it will first load Calc
2861 (Math::BigFloat needs Math::BigInt and thus loads it) and then later Bar or
2862 Baz, depending on which of them works and is usable/loadable. Since this
2863 loads Calc unnecc., it is not recommended.
2865 Since it also possible to just require Math::BigFloat, this poses the question
2866 about what libary this will use:
2868 require Math::BigFloat;
2869 my $x = Math::BigFloat->new(123); $x += 123;
2871 It will use Calc. Please note that the call to import() is still done, but
2872 only when you use for the first time some Math::BigFloat math (it is triggered
2873 via any constructor, so the first time you create a Math::BigFloat, the load
2874 will happen in the background). This means:
2876 require Math::BigFloat;
2877 Math::BigFloat->import ( lib => 'Foo,Bar' );
2879 would be the same as:
2881 use Math::BigFloat lib => 'Foo, Bar';
2883 But don't try to be clever to insert some operations in between:
2885 require Math::BigFloat;
2886 my $x = Math::BigFloat->bone() + 4; # load BigInt and Calc
2887 Math::BigFloat->import( lib => 'Pari' ); # load Pari, too
2888 $x = Math::BigFloat->bone()+4; # now use Pari
2890 While this works, it loads Calc needlessly. But maybe you just wanted that?
2892 B<Examples #3 is highly recommended> for daily usage.
2896 Please see the file BUGS in the CPAN distribution Math::BigInt for known bugs.
2902 =item stringify, bstr()
2904 Both stringify and bstr() now drop the leading '+'. The old code would return
2905 '+1.23', the new returns '1.23'. See the documentation in L<Math::BigInt> for
2906 reasoning and details.
2910 The following will probably not do what you expect:
2912 print $c->bdiv(123.456),"\n";
2914 It prints both quotient and reminder since print works in list context. Also,
2915 bdiv() will modify $c, so be carefull. You probably want to use
2917 print $c / 123.456,"\n";
2918 print scalar $c->bdiv(123.456),"\n"; # or if you want to modify $c
2922 =item Modifying and =
2926 $x = Math::BigFloat->new(5);
2929 It will not do what you think, e.g. making a copy of $x. Instead it just makes
2930 a second reference to the B<same> object and stores it in $y. Thus anything
2931 that modifies $x will modify $y (except overloaded math operators), and vice
2932 versa. See L<Math::BigInt> for details and how to avoid that.
2936 C<bpow()> now modifies the first argument, unlike the old code which left
2937 it alone and only returned the result. This is to be consistent with
2938 C<badd()> etc. The first will modify $x, the second one won't:
2940 print bpow($x,$i),"\n"; # modify $x
2941 print $x->bpow($i),"\n"; # ditto
2942 print $x ** $i,"\n"; # leave $x alone
2948 L<Math::BigInt>, L<Math::BigRat> and L<Math::Big> as well as
2949 L<Math::BigInt::BitVect>, L<Math::BigInt::Pari> and L<Math::BigInt::GMP>.
2951 The pragmas L<bignum>, L<bigint> and L<bigrat> might also be of interest
2952 because they solve the autoupgrading/downgrading issue, at least partly.
2955 L<http://search.cpan.org/search?mode=module&query=Math%3A%3ABigInt> contains
2956 more documentation including a full version history, testcases, empty
2957 subclass files and benchmarks.
2961 This program is free software; you may redistribute it and/or modify it under
2962 the same terms as Perl itself.
2966 Mark Biggar, overloaded interface by Ilya Zakharevich.
2967 Completely rewritten by Tels http://bloodgate.com in 2001, 2002, and still