#2 Treat DEF statements correctly

[Fadi Chabarek]

DEF Statements are currently treated like LET
definitions, i.e. with context scope. Thus, the UML
WFRs (in fact no OMG standard document) can be
validated.

This implies that the framework cannot be used for
any practical validation purpose with existing
standards, as it rejects the respective input.

Maybe this should be corrected.
(submitted by jgsuess)

[Fadi Chabarek]

Logged In: YES
user_id=450963

Date: 2004-12-17 13:40
Sender: fchabar
Logged In: YES
user_id=450963

The OCL chapter isn't clear about the scope of DEF
statements. In my opinion the WFRs especially the recursive
use of (DEF) attributes and the use of DEF statements across
contexts are not compliant to the standard (or the standard
does not come up to its needs, s. UML cha. 6.4.3). Recursive
DEF statements of operations are supported.

To develop recursive attributes and statements across
context help is wanted!

Date: 2004-09-24 08:47
Sender: jgsuess
Logged In: YES
user_id=557658

Seems a sensible suggestion. Actually, I thought with the
current version, type declarations were already generally
mandatory. (And I agree they should probably be anyway.) If
cyclic definitions create to many difficulties, than linear
dependencies would probably go a long way in resolving the
problem.

Date: 2004-09-23 19:36
Sender: fchabar
Logged In: YES
user_id=450963

I will see what I can do. The problem is that def statements
must be declared before they are used. But the declared
properties can be used in other def statements. The order of
the calls is not defined, cycles of def properties can exist.
I think I could restrict the language and define that def
properties must be specified with a type. Then I shall be able
to declare these properties in the in-phase while visiting
corresponding AST-nodes so that these properties are
available in the out-phase of the type check/evaluation. This
could work around the need of a type inference algorithm.