From: Gavin Sinclair <gsinclair@...>
Date: 2002-11-20T12:56:52+09:00
Subject: Re: Defining <=>

From: "Yukihiro Matsumoto" <matz@ruby-lang.org>


> Hi,
>
> In message "Defining <=>"
>     on 02/11/20, "Tim Hunter" <cyclists@nc.rr.com> writes:
>
> |I need to define <=> in a class. What is the appropriate thing to do when
> |the "other" argument is not an object in the class? For example, comparing
> |myObject <=> nil, which actually happens in the debugger. Should I just
> |send it to the superclass?
>
> Return nil.
>
> matz.


If you return nil, then future operations like < and sort will raise the
following error (1.6.5 Cygwin):

  TypeError: no implicit conversion from nil

This is a bit cryptic.  Is there any benefit to *not* just raising a TypeError
yourself (with a more relevant message) in the definition of <=> ?

Gavin