Part 25, UML and BAG or LIST

[Bernd G. Wenzel]

Dave,

the meeting I referred to was a discussion between the ISO 19103
team (UML profile for GIS/data exchange) and OMG's ADPTF (incl.
Jim Rumbaugh). The main part of the meeting was a discussion of
problems identified by the ISO 19103 team, and explanations from
the ADPTF folks, and primarily Jim, why they didn't exist. It was
nothing formal though.

Nevertheless, which UML spec are you referring to? When I read
the UML reference manual (by Rumbaugh, Jacobson, and Booch), the
text is pretty much the one you quoted except that there is no
such uniqueness constraint on associations/association ends.
There seems to be some confusion.

In addition, the definition of the concept equality in UML is not
clear to me. Is it based on (in EXPRESS terms) instance or value
equality, or on a user defined comparison operator? In the latter
case, what does it mean if such operator has not been defined?

If your version of the UML spec does not answer these questions,
we have found just another example of semantic weakness in UML.
Should this be the case, we are facing an undefined constraint,
which we can safely ignore IMHO.

:-) Bernd

----- Original Message -----
From: David Price <dmprice at us.ibm.com>
To: Bernd G. Wenzel <Bernd.Wenzel at eurostep.com>
Cc: Stephen Brodsky <sbrodsky at us.ibm.com>; <wg10 at steptools.com>;
<xmlsc4 at nist.gov>; <wg11 at smiling.net>
Sent: Monday, March 12, 2001 8:52 PM
Subject: Re: Part 25, UML and BAG or LIST



Bernd,

The UML experts may have agreed on this question at some point in
time but
there is clear text in the UML Spec on Association saying it is a
mathematical set. It says:

>"An association defines a semantic relationship between
classifiers. The
instances of an
>association are a set of tuples relating instances of the
classifiers.
Each tuple value may
>appear at most once."

It's pretty clear that "may appear at most once" in the second
sentence
confirms that duplicates are not allowed.

If you read further on AssociationEnd you find the following text
describing ordering :

>"ordered - the elements of the set have an ordering, but
duplicates are still
>prohibited. This generic specification includes all kinds of
ordering. This may be
>specified by the keyword syntax "{ordered}.."

Again, it's clear that duplicates are not allowed for
Associations.

Either the consensus you described has changed or not yet made it
into the
UML specification. If the latter is the case then I'm fine with
using
Association.

Thanks,
David

IBM Corporation
5300 International Blvd.
N. Charleston, SC 29418, USA
dmprice at us.ibm.com
Phone : +1 (843) 760-4341   Fax : +1 (843) 760-3349
Oooo.
(UNC)
 ) /
(*/


"Bernd G. Wenzel" <Bernd.Wenzel at eurostep.com> on 03/12/2001
02:19:35 PM

Please respond to "Bernd G. Wenzel" <Bernd.Wenzel at eurostep.com>

To:   David Price/Bethesda/IBM at IBMUS
cc:   Stephen Brodsky/Santa Teresa/IBM at IBMUS,
<wg10 at steptools.com>,
      <xmlsc4 at nist.gov>, <wg11 at smiling.net>
Subject:  Re: Part 25, UML and BAG or LIST



Dave,

about a year ago, there was a discussion between the ISO 19103
team (including myself) with OMG's UML gurus. The consensus there
was, that the word set is used in its common sense, not its
mathematical sense in the UML spec. In other words, a UML set is
a collection, where duplicates are allowed.

Another recommendation was to use tagged values to express
optionality and uniqueness. This is however only useful in the
case of a UML profile.

:-) Bernd

----- Original Message -----
From: David Price <dmprice at us.ibm.com>
To: <wg10 at steptools.com>; <xmlsc4 at nist.gov>; <wg11 at smiling.net>
Cc: Stephen Brodsky <sbrodsky at us.ibm.com>
Sent: Monday, March 12, 2001 6:06 PM
Subject: Part 25, UML and BAG or LIST


>
> Folks,
>
> I've read the UML sections on Association and Attribute several
times now
> and it's become clear to me that UML has no concept capable of
supporting
> the EXPRESS concept of BAG or LIST with duplicates.  In every
case, the UML
> spec uses the word "set".
>
> Here's the relevent UML 1.3 text:
>
> UML Association
>
> "The Association represents a set of connections among
instances of the Classifiers. An instance of an Association is a
> Link, which is a tuple of Instances drawn from the
corresponding Classifiers."
>
> UML Attribute (multiplicity)
>
> "The possible number of data values for the attribute that may
be held by an instance. The cardinality of the set of values is
an
> implicit part of the attribute. In the common case in which the
multiplicity is 1..1, then the attribute is a scalar (i.e., it
holds
> exactly one value)."
>
>  I'm at a loss as to what to do.  It seems to me that whatever
we do is
> going to violate UML. Perhaps we should chose the option that
makes that
> violation visible? That could mean we'd need to make sure
something in the
> UML gets named Bag-xxx. Anyone have any bright ideas?
>
> Thanks,
> David
>
> IBM Corporation
> 5300 International Blvd.
> N. Charleston, SC 29418, USA
> dmprice at us.ibm.com
> Phone : +1 (843) 760-4341   Fax : +1 (843) 760-3349
> Oooo.
> (UNC)
>  ) /
> (*/
>