[cad-linux] Re: symlinks as blocks, parts, etc

> The following was supposedly scribed by
> phrostie
> on Wednesday 17 September 2003 06:06 pm:

>i liked the symlink idea.
>most cad systems have some form of sharing geometry.
>they all have their own names whether it blocks, details, patterns, dittos,
>instances, or what ever.  don't chunck the symlink idea just yet.

The difficulty is to get a method of linking geometry that is flexible and 
usable by all of these systems.

Also, how would you describe such things as "attributes".  Possibly the 
inserted item has a file (or item) in the model tree that represents that 
entity and holds values specific to that instance.

Example:

In the openDWG toolkit (apparently (I've yet to try using them)) a block has a 
structure associated with its definition, and a member of that structure 
holds a blockobjhandle (or whatever they call it) which is read with the same 
methods as a drawing (I think that each instance of the block must have a 
pointer to the same content-definition memory address.)

I don't think a symlink is really the appropriate way to handle this (and I'm 
not quite sold on the parceled-out files approach either) but it may be a 
good way to think about it as a starting point.  Maybe the definition is via 
one entity which holds the attributes for the instance and a pointer to the 
data object (or the root of the tree which defines the block.)  

This pointer could be a url.

The data in the block could be the toplevel of a fs tree or maybe a file with 
a list of url-attribute pairs.  The root of the tree would be the most 
dynamically defined setup.

--Eric

Other related posts: