From: Bulat Ziganshin <bulatz@...>
Date: 2002-09-27T16:10:11+09:00
Subject: Re: adding overload to ruby

Hello Yukihiro,

Friday, September 27, 2002, 10:48:25 AM, you wrote:

|bclrS20tse>>   method overloading. If you really believe that type checking
|bclrS20tse>>   is that important, then you should probably be using another
|bclrS20tse>>   language. 
YM> |
YM> |you are not matz ;)

YM> If you really believe that compile time type checking is that
YM> important, then you should probably be using another language.

YM>                                                 matz. ;-)

YM> p.s.
YM> In fact, I'm *not* against method overloading because it helps "divide
YM> and conquer" the problem, not for type checking.

1) i say about dynamic type checking, as does overload.rb. see start of
this thread

2) "type checking" now goes to "protocol checking". i agree with that,
but class/mixin must be one variant of protocol. i think that "pattern
matching" is too far from ruby to be practical

3) code from date.rb:

  def - (x)
    case x
    when Numeric; return type.new0(@rjd - x, @sg)
    when Date;    return @rjd - x.rjd
    end
    raise TypeError, 'expected numeric or date'
  end

uses dispatch-on-type, not specially designed protocol. it is common
practice now. i don't understand why implementing this practice in
lang itself is bad idea:

def -(Numeric x)  type.new0(@rjd - x, @sg) end
def -(Date x)     @rjd - x.rjd end

this code is just the same as current one

-- 
Best regards,
 Bulat                            mailto:bulatz@integ.ru