The DAM annex B algorithm again

[Wilson, Peter R]

    I have implemented my checking algorithm (at least for a single schema)
for evaluated sets. It produces the results as given in Annex B for examples
155 and 156. It does not agree with the results as given for example 157,
which shows a final tally of 27 members of the evaluated set. There should
be either 26 or 39 members.
 
    The 27th member is the combination: (a b d l x y z). To me this is
a valid combination even though there is no common subtype of a or x
present. If this is a valid combination via the given algorithm then there
should be 39 combinations in all. If it is not meant to be valid, then
either the algorithm is wrong or it has been interpreted incorrectly.
Currently my algorithm produces the (a z + possibly subtypes) combinations
--- it is a lot simpler that way.

    I have tried some simpler versions of example 157. They all have the
same supsub graph, but I have used different versions of a
SUBTYPE_CONSTRAINT. The ENTITY model is:

{ iso standard 10303 part (11) version (4) }
SCHEMA annexb;

ENTITY a; END_ENTITY;
ENTITY b SUBTYPE OF (a); END_ENTITY;
ENTITY c SUBTYPE OF (a); END_ENTITY;    
ENTITY d SUBTYPE OF (a); END_ENTITY;
ENTITY f SUBTYPE OF (a); END_ENTITY;
(* 
-- constraint 1
SUBTYPE_CONSTRAINT a1 FOR a;
  ONEOF(b, c) AND d ANDOR f;
END_SUBTYPE_CONSTRAINT;

-- constraint 2
SUBTYPE_CONSTRAINT a2 FOR a;
  (ONEOF(b, c) AND d) ANDOR f;
END_SUBTYPE_CONSTRAINT;

-- constraint 3
SUBTYPE_CONSTRAINT a3 FOR a;
  ONEOF(b, c) AND d;
END_SUBTYPE_CONSTRAINT;

*)
END_SCHEMA;

    My algorithm produces the following evaluated sets:
constraint 1 (9 members):
(a) (abd) (acd) (af) (abf) (acf) (adf) (abdf) (acdf)

constraint 2 (9 members):
(a) (abd) (acd) (af) (abf) (acf) (adf) (abdf) (acdf)

constraint 3 (6 members):
(a) (abd) (acd) (af) (abdf) (acdf)

    There is a significant difference between the first two and the third
one, which depends on whether or not the ANDOR is included in the
SUBTYPE_CONSTRAINT.

    Does the DAM algorithm show the same difference? What is the `real'
meaning of an ANDOR? Does it differ if the ANDOR is explicit or implicit?

    I gather, although I have never seen one in action, that there are
implementations of the Edition 1 Annex B algorithm. Has anyone implemented
the DAM algorithm, and if so does it give the same results as shown in the
DAM, the same results as the Edition 1 algorithm, the ...? Working through
the published algorithm by hand is highly error prone (dropping the &
notation in favour of the algebraic multiplication notation (i.e., a&b&c =>
abc) would much improve the clarity of the examples, and probably any hand
calculation). I think it very desirable that there should be at least one
implementation of the algorithm before the DAM is published.

Peter W.

Dr Peter R. Wilson
Boeing Commercial Airplanes
PO Box 3707, MS 6H-AF, Seattle, WA 98124-2207
(Package Delivery: MS 6H-AF, 1601 E. Valley Frontage Road, Renton, WA 98055)
Tel: (425) 237-3506, Fax: (425) 237-3428
Email: peter.r.wilson at boeing.com
--------------------------------
Any opinions expressed above are personal;
they shall not be construed as representative of any organisation.
--------------------------------