From: Rich Kilmer <rich@...>
Date: 2002-04-01T22:17:41+09:00
Subject: RE: Creating an instance of a class without calling initialize()



> -----Original Message-----
> From: Harry Ohlsen [mailto:harryo@zip.com.au]
> Sent: Saturday, March 30, 2002 6:08 PM
> To: ruby-talk ML
> Subject: Re: Creating an instance of a class without calling
initialize()
> 
> "Joseph Girgis" <jbgirgis@msn.com> wrote in message
> news:<OE74EzcT1ePyu91wCJW00002df4@hotmail.com>...
> > Why don't you want the initialize() method?
> 
> Glad someone finally asked :-).
> 
> I've been working with Chris Morris's XML serialisation module, which
is
> really nice. However, what it currently requires is that your
initialize
> either takes no parameters or has defaults for everything, so that it
can
> be called with none.  What the deserialisation does is something like
...
> 
>     obj = new
> 
>     FOR EACH ATTRIBUTE
>         obj.send("#{attributeName}=", attributeValue)
>     END

You actually may want to recommend that he use:

obj = new

FOR EACH ATTRIBUTE
  obj.instance_eval "@#{attributeName}=#{attributeValue}"
END

That way it works for attributes that do not have writer methods but
only reader methods (ie. Foo.x but not Foo.x=)

> 
> where the attribute names and values are retrieved from the XML data.
> 
> However, we've been discussing how this is a bit of a restriction,
> especially if you're retrofitting the serialisation to a class
> hierarchy you've already written.
> 
> What you really want is to bypass any initialize that exists, not
> force the class designer to make their initialize callable with
> no parameters.
> 
> I'd been able to make everything work the way I want, but only by
> having a separate method in every class that generates a prototype
> object (ie, it just calls initialize with some valid set of values)
> and wanted to remove that requirement.
> 
> The idea about aliasing and then unaliasing initialize should work
> for me; I really should have thought of that !!  I'm also intrigued
> by the reference to YourClass.allocate() mechanism that's been
> mentioned as being in 1.7, so I'll look into that.
> 
> Thanks to everyone for the replies.  I just KNEW it could be done in
> pure Ruby, but wasn't thinking carefully enough !!