From: David Balmain <dbalmain.ml@...>
Date: 2006-06-04T16:38:27+09:00
Subject: Re: Ruby's Bignum

On 6/4/06, David Balmain <dbalmain.ml@gmail.com> wrote:
> On 6/4/06, Aaron Patterson <aaron_patterson@speakeasy.net> wrote:
> > Hi everyone.  I'm porting some C code to pure ruby, and I would like to
> > keep Ruby from turning my numbers in to Bignums.  Is that possible?
> >
> > For example, in C:
> >
> >   printf("number: %d\n", 0xd76aa478);
> >
> > Prints "number: -680876936"
> >
> > In ruby:
> >
> > irb(main):001:0> puts 0xd76aa478
> > 3614090360
> > => nil
> >
> > Is it possible to get ruby to behave the same way as C?  This seems like
> > something easy to do, I'm just having a hard time figuring it out.
> > Thank you!
>
> Hi Aaron,
>
> Short answer, no. In Ruby, as you know, everything is an object and in
> the C code implementing ruby, this is represented by pointers to these
> objects. These pointers have a type of VALUE which is basically a
> 32-bit integer. But it would be pretty inefficient to store integers
> as memory-allocated objects so instead it stores them in the same
> 32-bit integer. (This is true for FixInt, BigNum is an object ref). So
> how does the interpreter know the difference between an object
> reference and a FixInt? It relies on the fact that all memory pointers
> are are on 4 or 8-byte boundaries so that leaves at least the last 2
> bits free. These last two bits are used to specify whether VALUE is an
> integer, a refence or possibly a symbol, true or false value or nil.
> This means that there are only 30 bits left to specify the value of
> FixInt. So the maximum value you can store in a FixInt is 0x3FFFFFFF
> and the minimum is -0x40000000.
>
> irb(main):007:0> 0x40000000.class
> => Bignum
> irb(main):008:0> 0x3fffffff.class
> => Fixnum
> irb(main):009:0> -0x40000000.class
> => Fixnum
> irb(main):010:0> -0x40000001.class
> => Bignum
>
> Hope that makes sense.
>
> Cheers,
> Dave

Sorry, this answer should have been under "and I would like to keep
Ruby from turning my numbers in to Bignums.  Is that possible?"

To cast an Integer to a 32-bit signed integer you can use pack/unpack
like Park said, or you could do it like this;

def to_32_bit_signed(val)
    val |= -0x100000000 if (val & 0x80000000) > 0
    val
end