Charles Oliver Nutter
4/25/2009 11:09:00 AM
Javier Esteve wrote:
> I'm learning ruby, and I want to know if it is any way to know the
> biggest number to be represented with a specific type of number (Fixnum,
> Bignum, Float).
Float is a standard 64-bit IEEE floating point value. There are several
constants on the Float class that can tell you whatever you need:
â?? jruby -e "p Float::constants"
["MAX", "MAX_10_EXP", "MANT_DIG", "RADIX", "DIG", "MIN", "ROUNDS",
"MIN_EXP", "EPSILON", "MIN_10_EXP", "MAX_EXP"]
Fixnum has no such constants, but on most implementations (CRuby, JRuby,
Rubinius at least) its max size is dependent on how large its basis
representation is. I believe it can be either "32" or "64" bits on
CRuby, where on JRuby it is always "64" bits:
â?? ruby -e "p (1 << (1.size * 8 - 1)) - 1"
2147483647
â?? ruby -e "p -(1 << (1.size * 8 - 1))"
-2147483648
â?? jruby -e "p (1 << (1.size * 8 - 1)) - 1"
9223372036854775807
â?? jruby -e "p -(1 << (1.size * 8 - 1))"
-9223372036854775808
The loss of one bit of precision on most impls is to emulate CRuby's
implementation of Fixnum as a tagged pointer, where the top bit
indicates whether an object pointer is a real pointer or a Fixnum value).
And Bignum has no min or max size, other than the maximum size of a
byte[] on your system.
- Charlie