—package
JE::Number;
our
$VERSION
=
'0.062'
;
use
strict;
# I need constants for inf and nan, because perl 5.8.6 interprets the
# strings "inf" and "nan" as 0 in numeric context.
# This is what I get running Deparse on 5.8.6:
# $ perl -mO=Deparse -e 'print 0+"nan"'
# print 0;
# $ perl -mO=Deparse -e 'print 0+"inf"'
# print 0;
# And here is the output from 5.8.8 (PPC [big-endian]):
# $ perl -mO=Deparse -e 'print 0+"nan"'
# print unpack("F", pack("h*", "f78f000000000000"));
# $ perl -mO=Deparse -e 'print 0+"inf"'
# print 9**9**9;
# I don't know about 5.8.7.
# However, that 'unpack' does not work on little-endian Xeons running
# Linux. What I'm testing it on is running 5.8.5, so the above one-liners
# don't work. But I can use this:
# $ perl -mO=Deparse -mPOSIX=fmod -e 'use constant nan=>fmod 0,0;print nan'
# use POSIX (split(/,/, 'fmod', 0));
# use constant ('nan', fmod(0, 0));
# print sin(9**9**9);
# sin 9**9**9 also works on the PPC.
'""'
=>
sub
{
my
$value
=
$_
[0][0];
$value
== inf ?
'Infinity'
:
$value
== -+inf ?
'-Infinity'
:
$value
==
$value
?
$value
:
'NaN'
},
'0+'
=>
'value'
,
bool
=>
sub
{
my
$value
=
$_
[0][0];
$value
&&
$value
==
$value
;
},
'+'
=>
sub
{
$_
[0]->value +
$_
[1] },
# ~~~ I shouldn’t need this,
# but perl’s magic
# auto-generation
# isn’t so magic.
# cmp => sub { "$_[0]" cmp $_[1] };
;
require
JE::String;
require
JE::Boolean;
require
JE::Object::Number;
# Each JE::Number object is an array ref like this: [value, global object]
sub
new {
my
(
$class
,
$global
,
$val
) =
@_
;
if
(
defined
blessed
$val
and can
$val
'to_number'
) {
my
$new_val
=
$val
->to_number;
ref
$new_val
eq
$class
and
return
$new_val
;
eval
{
$new_val
->isa(__PACKAGE__) } and
$val
=
$new_val
->[0],
goto
RETURN;
}
$val
= _numify(
$val
);
RETURN:
bless
[
$val
,
$global
],
$class
;
}
sub
_numify {
my
$val
=
shift
||0;
# For perls that don't interpret 0+"inf" as inf:
if
(
$val
=~ /^\s*([+-]?)(inf|nan)/i) {
$val
=
lc
$2 eq
'nan'
? nan :
$1 eq
'-'
? -(inf) : inf;
}
else
{
$val
+=0 }
$val
;
}
sub
prop {
if
(
@_
> 2) {
return
$_
[2] }
# If there is a value, just return it
my
(
$self
,
$name
) =
@_
;
$$self
[1]->prototype_for(
'Number'
)->prop(
$name
);
}
sub
keys
{
my
$self
=
shift
;
$$self
[1]->prototype_for(
'Number'
)->
keys
;
}
sub
delete
{1}
sub
method {
my
$self
=
shift
;
$$self
[1]->prototype_for(
'Number'
)->prop(
shift
)->apply(
$self
,
$$self
[1]->upgrade(
@_
)
);
}
sub
value {
shift
->[0]
}
*TO_JSON
=
*value
;
sub
exists
{ !1 }
sub
typeof {
'number'
}
sub
class {
'Number'
}
sub
id {
my
$value
=
shift
->value;
# This should (I hope) take care of systems that stringify nan and
# inf oddly:
'num:'
. (
$value
!=
$value
?
'nan'
:
$value
== inf ?
'inf'
:
$value
== -+inf ?
'-inf'
:
$value
)
}
sub
primitive { 1 }
sub
to_primitive {
$_
[0] }
sub
to_boolean {
my
$value
= (
my
$self
=
shift
)->[0];
JE::Boolean->new(
$$self
[1],
$value
&&
$value
==
$value
);
}
sub
to_string {
# ~~~ I need to find out whether Perl's number
# stringification is consistent with E 9.8.1 for
# finite numbers.
my
$value
= (
my
$self
=
shift
)->[0];
JE::String->_new(
$$self
[1],
$value
== inf ?
'Infinity'
:
$value
== -(inf) ?
'-Infinity'
:
$value
==
$value
?
$value
:
'NaN'
);
}
*to_number
= \& to_primitive;
sub
to_object {
my
$self
=
shift
;
JE::Object::Number->new(
$$self
[1],
$self
);
}
sub
global {
$_
[0][1] }
sub
taint {
my
$self
=
shift
;
tainted
$self
->[0] and
return
$self
;
my
$alter_ego
= [
@$self
];
no
warnings
'numeric'
;
$alter_ego
->[0] +=
shift
();
return
bless
$alter_ego
,
ref
$self
;
}
=head1 NAME
JE::Number - JavaScript number value
=head1 SYNOPSIS
use JE;
use JE::Number;
$j = JE->new;
$js_num = new JE::Number $j, 17;
$perl_num = $js_num->value;
$js_num->to_object; # returns a new JE::Object::Number
=head1 DESCRIPTION
This class implements JavaScript number values for JE. The difference
between this and JE::Object::Number is that that module implements
number
I<objects,> while this module implements the I<primitive> values.
Right now, this module simply uses Perl numbers underneath for storing
the JavaScript numbers. It seems that whether Perl numbers are in accord with the IEEE 754 standard that
ECMAScript uses is system-dependent. If anyone requires IEEE 754
compliancy,
a patch would be welcome. :-)
The C<new> method accepts a global (JE) object and a number as its
two arguments. If the latter is an object with a C<to_number> method whose
return value isa JE::Number, that object's internal value
will be used. Otherwise the arg itself is used. (The precise details of
the behaviour of C<new> when the second arg is a object are subject to
change.) It is numified Perl-style,
so 'nancy' becomes NaN
and 'information' becomes Infinity.
The C<value> method produces a Perl scalar. The C<0+> numeric operator is
overloaded and produces the same.
Stringification and boolification are overloaded and produce the same
results as in JavaScript
The C<typeof> and C<class> methods produce the strings 'number' and
'Number', respectively.
=head1 SEE ALSO
=over 4
=item L<JE>
=item L<JE::Types>
=item L<JE::Object::Number>
=back
=cut