Re: UML module hackery

[David Hickerson <dahicke datawest net>]

Lars Clausen wrote:

> On Sun, 10 Nov 2002, David Hickerson wrote:
>
> > I have done enough class diagrams in Rat Rose to hate the way they do
> > lines, and I am sure that
> > everyone has opinion one way or the other. I suggest the API
> > programmer's approach: "Extend
> > do not change". Make it an option of the line to behave either way or
> > make additional new lines.
>
> I'd like to hear what is considered better with the 'middle' connection.
> If it's the ease of connecting to the object, we could have it snap to the
> nearest connection point.  If it's about the limited number of connection
> points, that's more tricky.  I think the first is a more valid concern --
> not having to find a connection point would be great, interface-wise.
I have found that you can take a line an add a whole bunch of connection 
point to it, then attach  the line to the class box at the corners.

> > There is definely one improvement needed for lines. And that is the
> > arrow head needs to also be
> > manully rotatable. Having the line horizontal and the arrow head
> > vertical when connected to a
> > class because the line is pixel hair off the point of connection is
> > pain. Give me a control point for
> > the rotation of the arrow head.
>
> Before we go in and add more control points, let's consider when it is
> really necessary.  One change that's started is to have zig-zag lines be
> more intelligent about their orientation.  It hasn't propagated beyond the
> standard zigzagline and box yet, and only works at creation (though I have
> a way to have it work for later, too).  It makes sure that the arrow points
> into the object, instead of being placed on the line of the object.  The
> only case where you could get a funny angled arrow is at a corner
> connection point.  Try it out and see if you still feel that arrow handles
> are necessary.
I played with class shape and the zigzag line a little bit. I still find 
it hard is some cases to get he line to match up with the connection 
point exactly to make the arrow head point into the class shape. This is 
especially the case when zoom in on the class. Also, this behavior only 
happens when connecting to the sides of a shape. Connecting from the top 
or bottom is fine, the arrow head always points correctly. This is do to 
the end control point behavior of expanding in only the vertical 
direction and the middle segment handling the horizontal direction 
expansion. I don't know hard this would be, but could the zigzag line be 
given an attribute to act in either horzontal manner or a vertical 
manner at the end control points, basically the ability to turn it 90 
degrees. This I think would fix side attachments to a shape.


On another note. I am interested in creating and contributing some 
shapes, specifically software design patterns. I have tried to create 
one pattern and export it to a shape, however it loses its UML 
properties in the translation to a shape. Is there a way to do this a shape?


Dave