Site format stuff

[Simon Cox]

Philip Verhagen said :

> If we're agreeing that the GRASS data structure should be consistent, why
> not create the same structure that exists for vector and raster-data:
> only one attribute allowed for each site (integer, of course, but in
> future maybe floating-point) and storing the category information in a
> site_cats directory? All the data processing that goes beyond already has to
> be done in a RDBS for vector and raster now. Advantages of this: tools
> like grass-informix won't be dependent on x,y alone, and the display
> commands can be standardized (there already is some stuff around like
> d.what.site and d.site.labels). Just a suggestion...
>

I think there is a big difference between rasters and the
other things which is relvant here.  A raster layer has an
entry for every location, which in turn is stored at a limited
fixed spatial precision.  This allows an implicit indexing
scheme based on row/col.  Multi-attributes are then dealt
with by multiple layers, and tools like mapcalc allow us
to look at these multiple attributes simultaneously, simply by
looking at one pixel (cell) at a time.

With sites (and rasters) we might also want to consider multiple
attributes of the same object.  This means that the attributes
have to be tied to an objects "name" or "code", in order to ensure

that it does refer to the same object (there is no implicit
indexing, and the location precision is not limited).
Thus, if we do not allow multiple attributes per site _within_
GRASS, (but force recourse to an external-ish dbms) then
you are actually penalising or restricting the flexibility of
operations on sites, in comparison with raster operations.

Now, maybe the latter is the price we have to pay by using
a raster-based system (Grass), but I feel that, since there has
been so little development work done to-date on site operation
functionality, then we actually have an opportunity here to
put a structure in place now that will allow this richness
(in the future, perhaps).  A bad choice on the "data-model"
is the surest way to make grass become obselete quicker.
That is not necessary (especially since Helena and Darrell
are such amenable people ;-) )

My farthing's worth

Simon Cox

----
___________________________________________________
      __  L     Dr Simon Cox			
   ,~'  L_|\    VIEPS Department of Earth Sciences, 
,-'         \   Monash University, Clayton Vic 3168        
(            \  Australia
\    ___     /          Phone +61 3 905 5762
 L,~'   "\_x/           Fax   +61 3 905 5062
           u    simon at artemis.earth.monash.edu.au
___________________________________________________