[Uml-devel] yet another mmgen update
Andrew Sutton
ansutton at kent.edu
Wed Feb 26 12:45:06 UTC 2003
i've been working on dependency generation (#includes) and think i got most of
them figured out. i've verified that the activitygraphs, core and datatypes
packages defined within uml compile correctly - or should. they're building
right now and i've been using them for testing code correctness. and because
i'm bored i'm punishing my computer and letting it build all 3
simultaneously.
on a side note: you'd think that documents published by a standards body would
be correct. i've found a couple of problems with the the certain components
of UML - some relationships define bi-directional associations when there
aren't reciprical references on one of the classes in the relationship. its
easy enough to fix - you can just change the "isNavigable" attribute of the
association end for that object from "true" to "false" to eliminate the
generation of assignment code - but still...
after further investigation, it seems that none of the associations defined
within UML bothered to indicate that any association end was NOT navigable.
in short, i'm going to have to go thru the uml document, figure out which
ends aren't navigable and change the specification. there's only 122
associations in UML. surely some of them are bi-directional.
anyways, its looking pretty good. the todo list is as follows:
- [high] fix association specs
- [high] generate lifecycle semantics (i.e., when and where to delete stuff)
- [high] generate containment code specific to composition relationships
- [high] more Qt signals for other objects (like extents)
- [medium] code to make the metamodel an actual plugin
- [low] generate automake/autoconf stuff for a metamodel
- [low] fix parsing and codegen for nested packages
i think that's about it. anybody wants anything else in there, let me know.
andy
More information about the umbrello-devel
mailing list