SEDS reports for ISO 10303-11

[Phil Spiby]

Jochen,

I think we need to discuss these at the Myrtle Beach meeting. However, I
would like to start the ball rolling on the discussions now, to allow those
people who cannot attend to have some input into this.

RE Typeof and schema names.
---------------------------
The developers of the original EXPRESS edition 2 (now the stalled edition
3!) tried to resolve this problem by introducing the typename data type.
This was a major change to the type model of EXPRESS as it currently stands
and was rejected for inclusion into the current edition 2 proposal (Draft
Amendment).
Lets look at the problem in detail:
1. I agree that string handling is not really appropriate for these
expressions.
2. Now for a common mis-conception, when an interfaced object is referred to
in this manner, all of its names are available (see 15.25 (4)), so if we
have:
SCHEMA s1;
  ENTITY e;
  END_ENTITY;
END_SCHEMA;

SCHEMA s2;
USE FROM s1 (e AS ent);
  TYPE sel = SELECT(ent, other);
  END_TYPE;
  ENTITY other;
  END_ENTITY;

  ENTITY test;
    attrib : sel;
  WHERE
    wr1: 'S2.ENT' IN TYPEOF(attrib);
    wr2: 'S1.E' IN TYPEOF(attrib);
  END_ENTITY;
END_SCHEMA;

Then both wr1 and wr2 are testing the same thing.

3. Your proposed solution runs into the major problem that we faced with
Edition 3. Where would your proposal fit in the syntax and, more
importantly, what are the datatypes of the operands (if this were to be used
as an expression?)
My guess is that you want IS to be a new operator, in which case you have to
look at how it fits into the expression syntax and semantics. I further
guess that you want to deal with the above problem via:
attrib IS ent
But this leads to other possibilities, such as:
3 + 5 IS INTEGER
[2, 'string'] IS SET OF my_funny_select
5 IS 27
Also you would like to store values of the right-hand portion of the
expression, how would we do this?
ENTITY tester;
  attrib : NUMBER;
DERIVE
  test : XXX := INTEGER;
WHERE
  WR1: attrib IS test;
END_ENTITY;

As you can see, and we found in Edition 3, by trying to solve this in what
seems a simple modification we can open a whole Pandora's box of problems.

RE: Longforms.
--------------
This is a long standing argument between to original WG5 and WG7! WG5
Created the EXPRESS language, warts and all!, with the goal that it should
support re-usability, and this was done using the USE and REFERENCE
constructs. WG7 needed a fixed specification of what was to be implemented,
and did not like the USE and REFERENCE constructs. As a result WG7 came up
with the idea of a long form to drive the implementation methods, based on
an understanding of what USE and REFERENCE meant. The EXPRESS committee,
believing that there was no need for the longform, since Clause 11 of
EXPRESS specified what was meant did not include any formal specification of
how to create a longform. OK history lesson over!
The question is "Is there still a need for longform schemas, given that Part
21 can now handle multiple schema data sections?". On your proposal, there
may be complications with the introduction of a keyword that effectively
means "this was referenced", since this could, I guess, be used in
non-generated schemas, and could be a means by which private sections of a
schema may be specified.

Phil


> -----Original Message-----
> From: owner-wg11 at steptools.com [mailto:owner-wg11 at steptools.com]On
> Behalf Of Jochen Haenisch
> Sent: Tuesday, February 19, 2002 8:33 AM
> To: 'SEDS initial'
> Cc: 'WG11 exploder'; Support Folder EPMTech
> Subject: SEDS reports for ISO 10303-11
>
>
> Please, consider the two attached SEDS reports.
> Best regards, Jochen Haenisch.
>
>  <<SEDS_P11_20020219.txt>>  <<SEDS_P11_20020219_2.txt>>
>
> ____________________________________________________________
> Jochen Haenisch			E-mail:
> Jochen.Haenisch at epmtech.jotne.com
> EPM Technology AS		Home of the EXPRESS Data Manager
> P.O Box 6629 Etterstad		Tel: Int + 47 23 17 17 26;
> mobile: Int + 47
> 922 60 274
> N-0607 Oslo				Fax: Int + 47 23 17 17 01
> Norway					Web:
> http://www.epmtech.jotne.com
>
>