NICE-ESG-Libs Digest        Thu, 22 Feb 96       Volume 2 : Issue  17 

Today's Topics:
fwd: [jacob@blackbox.enmu.edu] Re: strip and ARRAY[ANY] (NICE-ESG-Libs Digest V2 #16)


  NICE Eiffel Standards Group -- Library Committee Mailing List

  To post to list:
     NICE-ESG-Libs@atlanta.twr.com
  To send mail to the Chairman of the committee:
     NICE-ESG-Libs-chair@atlanta.twr.com
  Administrative matters (sign up, unsubscribe, mail problems, etc):
     NICE-ESG-Libs-request@atlanta.twr.com

 
Date: Thu, 22 Feb 96 16:43:49 EST
From: tynor (Steve Tynor)
Subject: fwd: [jacob@blackbox.enmu.edu] Re: strip and ARRAY[ANY] (NICE-ESG-Libs Digest V2 #16)
To: nice-esg-libs

Clearly, sometimes one will want a stronger (or weaker) condition, in
which case Jacob's `deep_equal', or more selective use if deep equality
on some fields, and shallow equality on others, or 
might be appropriate.

I was not attempting to propose that my proposed `items_are_equal'
function was a "one size fits all" solution to all cases where one wants
to use strip -- mearly a necessary one for implementing the semantics
(which I believe are) intended in the ETL examples. If these semantics
are regarded as unimportant, then the need for this new function is
of course diminished. 

Steve

------- start of forwarded message (RFC 934 encapsulation) -------
From: "Jacob Gore" 
To: tynor@atlanta.twr.com (Steve Tynor)
Subject: Re: strip and ARRAY[ANY] (NICE-ESG-Libs Digest V2 #16)
Date: Thu, 22 Feb 1996 14:25:49 MST

Hi, Steve.

I've thought about the "nothing else changes" postcondition (and wrote 
about it in a couple sections of my book), and your solution promising, 
though I have a different motivation than the expanded to nonexpanded field 
promotion.

Suppose that none of the attributes were expanded.  There is still a 
problem with

	p1: strip(x).is_equal(old strip(x))

The problem is that it is NOT the "nothing else changes" postcondition: 
For example, the routine could contain the code

	y.change

and p1 would still hold, since y's identity didn't change.

My solution was to use instead

	p2: strip(x).is_deep_equal(old strip(x))

but a couple of the reviewers pointed out that this may be too 
restrictive, as in the case when one of the other fields z refers to the 
same object as x (directly or indirectly).

Your solution,

	p3: strip(x).items_are_equal(old strip(x))

is halfway between p1 and p2.  It will often compute a lot faster, and it 
eliminates the concern about the case where z indirectly refers to the same 
object as x, but it is still overly restrictive if z and x refer to the 
same object.

Finally, both p2 and p3 will miss the following change to the object:

	y := clone(y)

Jacob

P.S.  Feel free to forward some or all of this message to the list if you 
feel it will help the discussion.
------- end -------