Re: drawing graphs with dia [PATCH]

[Grégoire Dooms <dooms info ucl ac be>]

Lars Clausen wrote:

> On Mon, 2005-03-14 at 15:21 +0100, Grégoire Dooms wrote:
>  
>
> > Hello,
> > Here is a context diff against CVS HEAD.
> > I implemented auto gap + absolute gap for lines and beziers.
> > It works like a charm (AFAIK :-).
> > I added the properties GUI (could be prettier).
> >    
>
> Works nicely, indeed.  Thank you very much for working on this!  There's
> a few things I'd like to see changed, though.
>
> Most importantly, the autogap feature should be dependent on which
> connectionpoint you connect to -- it doesn't make a lot of sense unless
> you're connected to a central connection point (a bit of sense, though,
> but it looks weird).  Making it depend on the CP will also remove those
> options from the properties dialog, always a good thing.
>  
Hmm. Not sure about this. I think compositionality  is a good thing
and as this feature works regardless of the connection point, I'd like 
to be able to use it. On the other side if there is a connection point 
at the center of each object it's a good default and there shouldn't be 
use-cases where on would like to point to an other CP.

> We'll then need a flag on CPs to say that they autogap (are interior).
> I'm thinking the Element class should automatically add this.  For now,
> we can just set it explicitly on the objects that have an interior CP.
> Hmmm... CPs don't have flags as such yet, but they have the directions
> 4-bit which is kinda flaglike.  Nah, it's gotta be a new field flags.
>
> For submitting patches, it'd be nice if you use the -u (unified) flag
> and did not compress it (unless it's >40K large).
>
>  
>
> > The only thing missing for a very nice way to output a graph from 
> > graphviz is the central connection point on the text object.
> >
> > I also attach an example dia file.
> >    
>
> The bezier autogap is significantly different from the line autogap, in
> that it doesn't move the head around the object.
>  
To move the head, move the second handle of the connected point. The 
auto-gap feature on bezier conserves the slope at the endpoint (this was 
my first proposition in a former mail). I chose this one after observing 
the behavior of bezier connected to an object when you drag the other 
point: they don't move around the connected point, they also conserve 
their slope.
In that sense the auto-gap feature conserves the behavior of the 
line/bez object when you move the other point.

> I'd like to see the other simple lines (polyline, zigzagline in
> particular) have gaps too.  And if the gap stuff is moved to the
> Connection/Orthconn classes, it'll suddenly apply to a whole lot of
> other objects.  See how autorouting is handled in OrthConn for examples
> on how to do this.  But this is second priority, first is proper gap
> defs for simple lines.
>
>  
There are a few bugs left.
I'll handle them:
- The absolute end gap in the bezier has its sign interpreted the other 
way around
- Line and bezier miss the copy code so gaps are'nt copied they are 
reset to their defaults.

--
Grégoire Dooms