From: Bob Alexander <bobalex@...>
Date: 2002-01-26T14:20:32+09:00
Subject: Re: Subrange of String subclass => invalid object

<nobu.nokada@softhome.net> wrote:
> What did you expect?
> -----------------------------------------
> S(otherValue: 77, "abcdefghij")
> ./rbug.rb:15:in `initialize: wrong number of arguments(1 for 2)
(ArgumentError)
>         from ./rbug.rb:15:in `[]'
>         from ./rbug.rb:15
> -----------------------------------------
> Or?
> -----------------------------------------
> S(otherValue: 77, "abcdefghij")
> S(otherValue: 77, "bcdefghi")
> -----------------------------------------

Well, my example was pretty simple, but in a real life case

  -  It's a pretty good bet that leaving the instance variable uninitialized
does not result in a valid object.

  -  We have no way of knowing whether copying the value (77) from the
original instance results in a valid object. Maybe it's a hash of the
string, or an object sequence number, in which case it should be something
different from 77.

I don't know of a foolproof way for a superclass to create a valid instance
of the subclass without knowledge of the subclass.

This is a difficult problem. The current solution is too sloppy for such a
well thought out language as Ruby. I hope collectively the community can
come up with a better way.

A simple answer that probably most folks won't like is: the superclass could
return an instance of itself (seems natural enough!). The subclass can then
do what it wants with it (in an overriding method), including returning a
new subclass instance, modifying the existing self instance, returning the
superclass instance, or anything else it chooses. This is a bit more work
for the subclasser, but it is airtight.

Bob