Re: Chapin Charts with Dia?

[Cyrille Chepelov <chepelov calixo net>]

Le lun, mai 28, 2001, à 06:49:22 +0200, Hubert Figuiere a écrit:

> Well. objects are defined in the objects directory. They a categorized
> by sheet. So UML objects are in objects/UML, network objects are in
> objects/network, and so on. So you'll have to create a subdir (like
> objects/chaplin for example). As for the objects, I think that you
> should take the generic objects / shapes approach: create a generic
> Chaplin object then make shapes for each different objects.
>
> For the drawing it may be a little bit more difficult has you'll have to
> implement a layout code that work shape-to-shape. Seriously I don't
> know if it hard to do or not.

One thing that could be done: implement these as connection objects (with
the extremities laid out that way:
                -----------X
                |          |
                |          |
                |          |
                |          |
                |          |
                |          |
                |          |
                X-----------

Then, for objects which can have embed other objects (all but instruction,
it seems), add *internal* connection points, like this:

        -------------------X
        |\                /|
        ----------+--------+
        |         |        |
        |         |        |
        |         |        |
        |         |        |
        |         |        |
        |         |        |
        *---------+---------  

(where X is an extremity, + is a regular connection point, * is both in the
same place, perhaps with tweaks so that an object's extremity doesn't
connect to its own connection points. Aah, I need to do that to the grafcet
steps and transitions too <grin/>)

Instruction could be done that way:
        +----------------X
        |                |
        X----------------+

Multichoice will be a bit trickier to do, but certainly not unfeasible, you
"just" have to handle a dynamic amount of connection points (you might find
connpoint_line a useful container for such CP's).

Finally, begin by deriving a sufficiently "modern" object, ie, one that uses
Standard Properties all along. I'd say, one which uses object_copy_props()
should be "modern" enough (but I'm a bit biased towards that way of doing
things).

        -- Cyrille

-- 
Grumpf.