[Uml-devel] code generation changes

[Paulo Sehn]

Hi,

I've done some changes in order to fix these issues, I'm sending a new
patch with the fixes plus an option in the cppcodegenerator dialog to
generate accessors as public, I guess it will be useful.

Paulo Sehn

On 7/18/05, Oliver.Kellogg at t-online.de <Oliver.Kellogg at t-online.de> wrote:
> On Monday 18 July 2005 07:01, Paulo Sehn wrote:
> > I'm posting some changes I've done in umbrello to increase the
> > coverage of cpp code generation, this changes are not to be sent to
> > subversion yet, I've just wold like to receive comments about this.
> >
> > Changes:
> > ======
> > 1) assign void to operations that don't have return type and aren't
> > constructors/destructors
> 
> As long as you're only doing this within cppsourcecodeoperation.cpp and
> friends then that's fine.
> BTW,  to identify a destructor in several places you write something like
>    ... methodName.find(i18n("~")) ...
> That i18n() is superfluous - the tilde is part of the C++ syntax and as such
> should not be localized?
> Also some documentation of the arguments of applyStereotypes() in
> cppheadercodeoperation.h wouldn't hurt.
> 
> > 2) removal of the generate makefile entry in the cpp code generation
> > wizard - code generation options dialog - language options tab. It was
> > not working.
> 
> IMHO that's fine - If nobody objects then I'll commit it.
> 
> > 3) changes in the xmi file generation in order to remove the entries
> > for native data types. Just to see what I'm talking about: open
> > umbrello, set the language to cpp, save the empty file, open the xmi
> > file and search for fileName. You'll see things like int, float, and
> > so. I don't think we should have ,h/.cpp entries for native types. So,
> > I've removed this and the makefile entry too.
> 
> I believe it could be okay as long as you're only doing this inside the
> <codegeneration> part of <XMI.extension>, but I haven't tested it yet.
> (I.e. I haven't yet changed the "#if 0" in ClassifierCodeDocument::
> saveToXMI() into "#if 1")
> 
> > 4) I've changed the the cpp code generation wizard - code generation
> > options dialog to manage the options generate accessors, accessors are
> > inline and operations are inline.
> 
> Okay.
> 
> > 5) I've changed the umloperationdialog changing the stereotype editbox
> > for a stereotype combobox, this combo is filled with stereotypes
> > generated by the codegenerator (i.e. cppcodegenerator) plus the
> > stereotypes the user had defined.
> 
> Haven't tried it yet.
> 
> > 6) I've defined the friend, virtual and constructor stereotypes to the
> > cpp language
> 
> Neat.
> 
> > 7) I've changed the cppcodegenerator to handle these stereotypes, now
> > we can create friend operations, we can identify constructors, and we
> > can generate virtual operations that aren't abstract.
> 
> Wonderful.
> 
> > 8) I've changed the cpp code import classes to identify constructors
> > and friend operations.
> 
> Ditto.
> 
> > I would like to get a feedback about these changes and to know if I
> > have broken somethingelse :)
> 
> One thing - I notice that when using the old C++ generator (see the checkbox
> in General Settings/Miscellaneous) a constructor is generated twice, one with
> "void" in front and one without.
> 
> I'm appending the diff against current svn trunk with some minor changes.
> I was in the directory kdesdk/umbrello/umbrello when making this diff.
> Please send your patches as proper text attachments or gzipped -
> I had a hard time cutting and pasting the diff from your mail.
> 
> Oliver
> 
> 
>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: patch.txt
URL: <http://mail.kde.org/pipermail/umbrello-devel/attachments/20050723/bd574ee6/attachment.txt>