[Uml-devel] Reverse engineering

[Andrea Aime]

Hi Paul, hi everyone else,
as I already said, I propose myself for reverse engineering sources into
UML diagrams. I already taken a look at some different approaches, namely:
* kUML 0.51 approach. I like the class structure, but it doesn't compile on
my
  Mandrake 8.0, and as far I can see from the sources, it doesn't support
templates.
  Moreover, the parser has been made with a Java lexer/parser generator,
ANTLR,
  that seems more difficult to learn than the lex/yacc couple, althought is
more
  complete as well. Finally, it uses exceptions, this means more clear code,
but
  also more bloated one until gcc people fix this problem;
* Doxygen approach: only one flex file and lots of global data structures...
the code
   seems too much oriented onto the production of documentation (well,
that's Doxygen
   work in fact) but can handle templates and the ability to get also
comments is
   good if you are willing to get not only a diagram, but also some
explaination in
   the form of comments
I'm also willing to take a look at KDOC sources, althought is more limited
than
Doxygen (can only parse C++ and IDL, whilst Doxygen can handle also Java).

I also found some reference about automatic diagram layout (necessary after
reverse engineering...)

For what concerns XMI, I've taken a look at the standard, and it's really
big...
too much for a direct reverse engineering into XMI format. I will handle
only
the reverse engineering problem.
Well, comments and suggestions are welcomed :-)
Best regards
Andrea Aime

PS: for what concerns the file format, I agree with option 1, but before
freezing
      everything I think that we should support also activity and state
charts... or
      well, at least state charts only...