Feature #2347: Math::INFINITY

2009/11/9 Marc-Andre Lafortune <[e-mail]>:
> It is easy to get an infinity in Ruby (ex: 1.0/0) but that may not be obvious to everyone.
>
> Could we have Math::INFINITY which would make code using it cleaner?

As my understand, Ruby has not had Math::INFINITY because C does not have it.
Introducing infinity might decrease portability of Ruby. But I think
actually it is not a problem.

Now Ruby assumes the environment has native threading and US-ASCII
compatible locale.
Here are a problem. Is there an environment which satisfies the
conditions but does not have infinity?

Thanks,
-- Yuki Sonoda (Yugui)

2009/11/9 Tanaka Akira <[e-mail]>:
>> Here are a problem. Is there an environment which satisfies the
>> conditions but does not have infinity?
>
> NetBSD on VAX, I guess.

VAX!

OK. I understand. I'm against for introducing Math::INFINITY.

-- Yuki Sonoda (Yugui)

Hello,

In message "[ruby-core:26632] [Feature #2347] Math::INFINITY"
    on Nov.09,2009 04:53:15, <[e-mail]> wrote:
> +    rb_define_const(rb_mMath, "INFINITY", DBL2NUM(1.0/0.0));

FYI, some strict compilers doesn't permit to divide by zero
literal.


Regards,
-- 
U.Nakamura <[e-mail]>

Hi,

2009/11/9 Yugui <[e-mail]>:
>> Could we have Math::INFINITY which would make code using it cleaner?
>
> As my understand, Ruby has not had Math::INFINITY because C does not have it.
> Introducing infinity might decrease portability of Ruby. But I think
> actually it is not a problem.

I proposed Float::INFINITY in [ruby-list:7023], and it was discussed
in the thread
and a thread from [ruby-dev:1657].  1.0 / 0.0 now produces an
infinity, but it raised
ZeroDivisionError at that time.

I don't remember the reason why Float::INFINITY was rejected, but we
have already
Float#infinite?, so Math::INFINITY or Float::INFINITY which are
provided only on the
platform supportng IEEE 754 may not be so bad.

Apart from portability issues, someone proposed a class whose instance
represents
a mathematical infinity at that time, not a IEEE 754 infinity.  It may
be worth considering.

-- 
Shugo Maeda

Hi,

In message "Re: [ruby-core:26644] Re: [Feature #2347] Math::INFINITY"
    on Mon, 9 Nov 2009 12:17:41 +0900, Shugo Maeda <[e-mail]> writes:

|I don't remember the reason why Float::INFINITY was rejected, but we
|have already Float#infinite?

It's not because infinity is not defined by Ruby's number system, but
because there's no portable way to generate infinity from C.  1.0/0.0
may not be compiled on some compiler.

							matz.

I'm quite ignorant about the differences between compilers and implementations (and I wouldn't mind remaining this way! ;-)

The fact is, Ruby _already_ generates infinities. Line 2471 of numeric.c:

#define infinite_value() ruby_div0(1.0)

It's used in fix_pow.

I think there is a C macro called INFINITY also:
http://www.gnu.org/software/libtool/manual/libc/Infinity-and-NaN.html

If it is still an imperative that Ruby supports non IEEE architectures, then maybe that constant could be conditional on that?

I do not want to judge the decision to support an old platforms or not. I think it would be sad if Ruby was kept back because of supporting them though.

Hi,

2009/11/10 Marc-Andre Lafortune <[e-mail]>:
> I'm quite ignorant about the differences between compilers and implementations (and I wouldn't mind remaining this way! ;-)
>
> The fact is, Ruby _already_ generates infinities. Line 2471 of numeric.c:
>
> #define infinite_value() ruby_div0(1.0)

FYI, I heard that some compilers doesn't permit zero divisions by
literals, but the following
code can be used to avoid the check:

  tmp = 1.0;
  inf = 1.0 / (tmp - tmp);

GNU Octave has more complicated code to generate an infinity:

        volatile double tmp_inf;

#if defined (SCO)
        volatile double tmp = 1.0;
        tmp_inf = 1.0 / (tmp - tmp);
#elif defined (__alpha__) && defined (__osf__)
        extern unsigned int DINFINITY[2];
        tmp_inf =  (*(X_CAST(double *, DINFINITY)));
#else
        double tmp = 1e+10;
        tmp_inf = tmp;
        for (;;)
          {
            tmp_inf *= 1e+10;
            if (tmp_inf == tmp)
              break;
            tmp = tmp_inf;
          }

-- 
Shugo Maeda

• Status changed from Open to Assigned

• Priority changed from Normal to Low

• Status changed from Assigned to Closed
• % Done changed from 0 to 100

This issue was solved with changeset r26197.
Marc-Andre, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.