[cad-linux] Re: the parametric project

Jeff, 
thanks for these very valuable hints. Both issues that you mentioned
are obviously correct and right on. They are dependent on the
implementation of the "constraint solver". Now, at this point, I am
only thinking of the actual core of such a system. Whereas what you
have mentioned is already two resp. three steps up. 

I'll keep your email anyway since you have exactly described what this
parametric thing is all about.

Thanks,
Roland

--- Jeffrey McGrew <JMcGrew@xxxxxxxxxxxxxx> wrote:
> 
> > P2 relates to P1, if P1 moves so does P2. Very simple but works.
> The
> > code for this is in the latest upload on kosmos.sf.net.=20
> 
> One thing that I've seen that might help you in your thinking is
> that the systems I've seen that are the most successful at parametric
> all have two major 'elements' or 'themes':
> 
> 
> 1. The associations are bi-directional, and real-time.
> 
> This mean that when I change something, like move a wall, the
> parametric
> instantly 'kick-in' and that wall will not only move elements that
> are
> located within it (like the commonly discussed door example) but also
> update all the views of said objects. This makes it possible for the
> user to understand the relationships between the objects, and to see
> changes instantly. Other parametric systems that I've seen that lack
> this functionality, and require a manual 'update' or 'regen' run to
> see
> changes, are much harder to deal with and waste time. For example, if
> I make a change to something in Revit, Revit updates that change
> everywhere that moment (which sometimes can take a while if the
> change
> is complex). In ArchiCAD, if I make a change, I then have to go to
> the
> views that would 'see' that change (like a section) and tell it to
> update that view; this adds work, and when things don't resolve it
> becomes a mess as to what's not working if you changed more than one
> thing. In Revit, if the change is somehow 'illegal' (like moving a
> door
> outside of a wall) it tells you that moment, allowing you to fix it.
> That real-time feedback is important. Also, the bi-directional, is
> part
> of this; some parametric systems I've seen wouldn't let you make
> changes in any view. For example, in ADT, I don't believe you can
> make
> changes anywhere in the project; you can only change things in
> certain
> views. So you can't change a door on the door schedule and have that
> door re-draw itself automatically, maintaining any relationships you
> might have constrained it to; you have to go to the plan and change
> it
> there. In Revit, you can change anything, anywhere, and any time, and
> that makes it so much more efficient and accessible of a system.
> 
> 
> 2. Make the constraint metaphors consistent with the discipline of
> work.
> 
> To make a simple constrain to something in Revit, you simply draw a
> dimension string between two items, and then 'lock' the string.
> Now those two items will hold that relationship, the distance between
> them, until you 'unlock' it. There are more complex constraints that
> you can do, but, metaphorically, they work the way an Architect would
> think about things. As an Architect, I'm gonna think "I want this
> Desk held off of the face of that Wall there 3 feet" not "I need
> this handle to reference that wall-sub-object at a blah blah blah".
> Inventor, another parametric system, also deal with constraints
> by common Mechanical methods and thinking; so that it's very easy
> from a user to understand how to set up parametric relationships
> between objects. If you system you are planning can be coded in such
> a way that the various disciplines have the ability to use the
> metaphors
> that they are familiar with to set up relationships between object
> than
> your system will be far more accessible to lay people and will get a
> lot
> more people using it.
> 
> Jeffrey
> 
> 
> 

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com

Other related posts: