[st-users] Some thoughts on ST-Developer v7

[David Loffredo]

Here's some thoughts for folks to chew on.  In the next release of
developer, I'm making some changes to the core object indices that
should dramatically increase traverse-by-type speed and should also 
eventually reduce memory usage.  

The only externally visible side effect from all of this will be that
objects in a part 21 file will be written out sorted by type rather
than in order of creation as they are now.  For example, instead of
the following:

    #10 = POINT(1.,0.);
    #20 = POINT(2.5,4.);
    #30 = POINT(5.,0.);
    #40 = LINE(#10,#20);
    #50 = CIRCLE(1.5,#60);
    #60 = POINT(0.,0.);
    #70 = TEXT('A Little Picture',#30);
    #80 = CIRCLE(1.5,#20);

The files would look like this:

    #10 = POINT(1.,0.);
    #20 = POINT(2.5,4.);
    #30 = POINT(5.,0.);
    #40 = POINT(0.,0.);
    #50 = CIRCLE(1.5,#40);
    #60 = CIRCLE(1.5,#20);
    #70 = LINE(#10,#20);
    #80 = TEXT('A Little Picture',#30);

I'm curious what folks think about this side-effect.  In my mind, the
locality of reference in the original scheme was handy sometimes, but
this new way can be really nice too, since you have a great picture of
the types of entities in the file, and they are conveniently grouped.

It also opens up a bunch of other options, like writing the object
types in alphabetic order, using a graph ordering alogorithm to
minimize forward references and so forth ....

						- Dave