Re: [CEDET-devel] cogre questions

[<er...@si...>]

Hi,

  It has been quite a while since I have had time to develop cogre, and I
really don't remember the answers, but I suspect the current answers are
all the same: There is no user facing menu or M-x command for those things.

  When I was first developing the UML part of cogre I ran into limitations
of semantic (such as the stubbed out class you mentioned.)  More work is needed
in that area.

  You can fix the stubbed class by making sure that the source to the parent
class has been seen by semantic (using one of the many semanticdb mechanisms
for doing so) such that it is findable.  Using the senator menu to jump to
it by name is a simple mechanism for determining if it is visible.

  You can add in a single class from a semantic tag by inserting a node
of type cogre-semantic-class.

  As for if it can be done, the answer is yes.  Someone needs to write the
software.  I see the following needed tasks to be done:

  1) debug the cogre/semantic2 tag usage.  I suspect there are still
     bugs hiding in there.  In addition, the naming convention hasn't been
     updated yet.
  2) Write a layout engine for a directed graph
  3) Write code to convert a semantic-class-database into a diagram.
  4) Write code to query datatypes to link up aggrigates.
  5) Write code to make semantic tag based diagrams saveable.
  6) Fix whatever the save/load problem is.
  7) Write code that detects links when new nodes are added from semantic
     tags.

  I'm pretty sure that would keep someone busy for a while. ;)

  There is also code to create DOT diagrams from semantic hierarcies.  If you
have DOT installed you can make large class diagrams pretty easily.  This also
needs aggrigate logic added, and could be a base for a cogre diagram generation
mechanism.

Eric

On Tue, 28 Sep 2004 16:41:39 +0200, Joakim Verona wrote:
> 
> Eric,
> 
> I checked out and tested your fixes.
> 
> - cogre-uml-quick-class now generates a cogre diagram with a cogre
>   node representing the class from a java class.
> 
> This is plenty cool! Semantic also figures out inheritance and makes
> an inheritance link nicely to the base class.
> 
> However:
> 
> - the base clas isnt found, but is "stubbed in". Ok, but can I then
>   later "bind" it to the real class?
> 
> - What I really would like to do right now is to make diagram with
>   about 5 classes and show their aggregation relations. 
> 
> - the "quick-class" method seems to generate an entire graph. Is it
> possible to somehow insert the class node in an existing graph and
> create  the aggregation relations afterwards? (because it wont be
> possible for semantic to determine these relations)
> 
> Cheers,
> /Joakim
> 
> 
> > On Tue, 28 Sep 2004 11:40:19 +0200, Joakim Verona wrote:
> >> Hello list,
> >> 
> >> I'm toying with cogre and find it pretty cool.
> >
> > Nifty
> >  
> >> I understand that very little of the long term goals are actually
> >> implemented, and my question is what actually works.
> >
> > Most of the complex cogre stuff I would like to do resides in the
> > semantic package which is still a work in progress.  Some has been
> > written there, but has not been used cogre yet.
> >
> >> - I can create a simple diagram with boxes and lines. I can move
> >> them about. Im having difficulty restoring the graph after I saved
> >> it though. Is this supposed to work?
> >
> > I must profess, I really don't remember.  COGRE uses the eieio
> > persistent object system which is heavilly utilized in semantic and
> > that works well.  I'm pretty sure it still had isues with graphs
> > generated from semantic though.
> >
> >> - If I load cogre-uml, im able to create a "class" node if I press
> >> c-u first, to specify the class. Its difficult to specify the "tags"
> >> manually though. I cant seem to change the "name" of a class.
> >
> > For UML diagrams, cogre uses the semantic tag format.  COGRE
> > documentation was then never updated to describe the new semantic 2
> > tag format which looks like this:
> >
> > ( "name" variable ( :attrib1 "value" ...) ( :prop1 "value" ) [ start
> > end ] )
> >
> > though I didn't have much luck with the conversion this morning.  I'll
> > need to investigate more.
> >
> >> - I can specify "aggregation" links the same way
> >> 
> >> - the "l" refers to a non existing function
> >> 
> >> - the cogre-uml-create and cogre-uml-quick-class functions appear
> >> broken
> >
> > I checked in a fix which was a compatibility issue with the new
> > version of semantic.  I found the display with semantic's images seems
> > to have done something strange though.
> >
> >> - this has really nothing to do with with this but anyway: when
> >> running in a terminal I only get 16 colors, which is to little for
> >> cogre. Is there some way to increase this? (other than running emacs
> >> over X which is slow)
> >
> > Not that I know of.  You miss some of the display features that way,
> > like overlines and such.
> >
> >> - It would be nice to have a little demo class hierarchy to test
> >> cogre with, like the diagrams on the screenshot page for cog
> >
> > Yup.  It would.  Any assistance improving cogre would be appreciated.
> >
> > Eric