[Uml-devel] XMI reading (fwd)

[Jens Krüger]

Am Samstag, 5. Oktober 2002 15:11 schrieb Egon Willighagen:
> On Saturday 05 October 2002 14:30, Jens Krüger wrote:
> > I have a question to your file. How did you create this file? I find the
> > reason for the crash but it makes no sense for me to create the missed
> > item, if you not know what type of listitem you have to create.

Hi Egon,

I can. The problem is, that you created a listitem for a operation without a 
listitem for a class. If you create a class and add operations and attributes 
to this class, umbrello writes a listitem with child listitems for operations 
and attributes. Your tool wrote only the operations in the listitem part 
without any information of the class to which its belongs. The problem is 
based in the design of the umbrello. There are some informations about the 
UML objects itself and some other for the represantion of these UML objects 
im umbrello. 

Here is a sequence of listitems the should work correct:

  <listview>
   <listitem open="1" type="800" id="-1" label="Views" >
    <listitem open="1" type="801" id="-1" label="Logical View" >
     <listitem open="1" type="813" id="2" label="Crystal">
      <listitem open="0" type="815" id="2000" label="setA" />
	.
	.
	.
     </listitem>
    </listitem>
   </listitem>
  </listview>

The missing listitem of type="813" (class) is the reason for the crash.

I know that is difficult to understand, but at this time we have this problem.

In the future we should remove this splitting of informations.

>
> Ok, can you be a bit more specific? The XMI dtd is very, very flexible and
> allows for almost no information... I guess that read <UML.*> objects
> should automatically be added to the list... and the element name tells you
> what kind of object you are dealing with. Like <UML.Class> is a class... ;)
> Not sure what the name is in Umbrello...
>
> > But you are
> > right, the umbrello should not crash in this case. It should handle
> > corrupted files in a non-crashing way.
>

Yes, it's true, but for the first time we could live with this solution I 
guess. The error handling has to improve in the future. 

> Yes, it could better give an informative error...
> The file was made with that JavaDoc Doclet that I am writing... It is not
> finished yet, but you can download it from:
>
> http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cdk/cdk/reports/javadoc/sour
>ce/
>
> Egon
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Uml-devel mailing list
> umbrello-devel at kde.org
> https://mail.kde.org/mailman/listinfo/umbrello-devel

--