From: Peter Vanbroekhoven <calamitates@...>
Date: 2005-10-18T09:15:09+09:00
Subject: Integer#** weirdness

Hello,

I've encountered a bit of a problem when I was trying out Ruby's numeric 
coercion:

    class SomeNumberLikeThing

      attr :value

      def initialize(value)
        @value = value
      end

      def **(other)
        SomeNumberLikeThing.new(value ** other.value)
      end

      def coerce(other)
        [SomeNumberLikeThing.new(other), self]
      end

    end

    p 2.0 ** SomeNumberLikeThing.new(2)
      # OK, returns #<SomeNumberLikeThing:0x4021c32c @value=4.0>
    p 2 ** SomeNumberLikeThing.new(2)
      # ERROR: /home/peterv/local/lib/ruby/1.8/rational.rb:339:in `**':
           undefined method `>=' for #<SomeNumberLikeThing:0x4021c2dc
           @value=2> (NoMethodError)

Seems that Integer#** tries to compare its argument to zero before 
coercing. Problem is that I don't want this SomeNumberLikeThing to be 
comparable. This is not something very exotic, as not all "number-like" 
things are (totally) ordered, e.g., polynomials, matrices and complex 
numbers. So IMO Integer#** should coerce before doing anything else with 
its argument.

As a side note, Complex#<=> is defined, but it violates the definition of 
an order: Complex(1, 1) >= Complex(1, -1) and Complex(1, 1) <= Complex(1, 
-1) but not Complex(1, 1) == Complex(1, -1), which is a violation of the 
antisymmetry law.

Peter