From: Dave Thomas <dave@...>
Date: 2004-03-06T12:09:19+09:00
Subject: Re: Duck typing chapter


On Mar 5, 2004, at 17:44, Warren Brown wrote:

>     One thing that stuck out to me is that after explaining that in 
> Ruby
> the class is not the type, the class Roman tests the class of its
> parameters (i.e. "if Integer === other").  In this case you could (and
> probably will) argue that the class *is* the type, but this should
> probably be specifically stated to the reader to avoid confusion.

Well, I think I'd probably say that we're interested in the class here, 
not the type. The classes contain the implementation of methods sych as 
'+', are we're trying to determine is we have compatible a compatible 
receiver and argument for addition. So I'm not really interested in the 
type as much as I am in the implementation.

Does that make sense?

>
>     I'm sure you'll find all of the easy typos, but there is a subtle
> one in the Roman#initialize method (no not the "muts" one).  The error
> message should say "<= #{MAX_ROMAN}", not "< #{MAX_ROMAN}".
>

Oh! Good catch.

>     The only other thing that struck me as odd was the assumption in
> Roman#coerce that any parameter that is not a Fixnum should be 
> converted
> to a Float.  This seems a bit simplistic, especially with Bignum as a
> built-in class, and would seem to preclude using some other 
> Integer-like
> class (Chinese number?).  Perhaps a check for a to_int method would be
> more appropriate?
>

This whole example is screwy, but let me tell you my initial reasoning.

If the argument is a Roman number, then the result might be a Roman 
number.

If the result is a Fixnum, then the result might be a Roman number

But is the argument is a bignum or anything else, the result can't be a 
Roman number (because bignums would be too big).

I can't test for to_int, because that would lose precision (Float 
supports to_int, which I'm going to post about separately).

Having said that, the more I look at this code, the less I like it. 
Roman numbers are representations, and don't really support arithmetic 
directly.

Time for a new example, methinks


Cheers

Dave