[cad-linux] Re: An open CAD file format

  • From: Eric Wilhelm <ewilhelm@xxxxxxxxxxxxx>
  • To: cad-linux@xxxxxxxxxxxxx
  • Date: Sun, 7 Sep 2003 01:17:18 -0500

> The following was supposedly scribed by
> Massimiliano Mirra
> on Sunday 07 September 2003 12:36 am:

>Yes, I have some of them in my saved posts archive.  Personally I
>think using a database would gain speed but would lose (too) much ease
>of access to the data.  Right now I can make quick changes to objects
>even with ed...

So what you need is db-ed.

I'm now working on a write-up aimed more at developers than my previous paper.  
I'll post a link as soon as it is ready.

In looking at yours and Bruno's pages, I am seeing the benefits to using the 
filesystem and other existing operating-system functions.  However, I have 
been working in this same mode for six months now and have a project 
directory that looks something like this:
$ls -lR |grep -c ""
        15967

That is a lot of files to keep track of and my head is nearly exploding with 
the complexity.  What is probably worse (since my head explodes with 
complexity on a fairly regular basis and I'm really starting to get used to 
it) is that everyone else in the office has to ask me what all of these files 
mean if they want to get anything out of the system.

The real issue here is interface.  I have built a Perl module for this project 
which basically lays out where to find anything and facilitates loading each 
file into some structure or another.  This module is only 714 lines of code 
and has mostly encapsulated the entire organizational system.

While understanding how to use the 714 line module is more difficult than 
modifying a single file, using the module leaves you less-likely to damage 
any of the files with regard to other programs utilizing the module.

Interface is the reason that I shy away from the raw-filesystem implementation 
because informing others in the office of the content of some files has 
resulted in over-written configuration data (you ever try to keep permissions 
straight on 15967 files and work on a team?)

Given an abstracted interface which restricts / tracks / facilitates use and 
modifications to the data, the filesystem approach could work, but now it is 
starting to sound a lot like a database.

--Eric

-- 
"Left to themselves, things tend to go from bad to worse."
                                        --Murphy's Corollary


Other related posts: