[GRASS5] Re: Named attribute support in GRASS

[Eric G. Miller]

On Thu, 07 Feb 2002 19:53:54 -0600, Helena <hmitaso at unity.ncsu.edu> wrote:

[snip]
> the current sites format was a temporary solution until an adequate capability
> would be available in vector format - we had 3D soils data with many attributes,
> as well as some 3D pollution data and sites combined with awk provided
> lots of functionality that we needed. I don't think that the current sites format
> should be
> further developed, however we cannot throw it away until its
> functionality is fully replaced (my entire work now depends on it as all the
> measured
> data that I get are ascii point data and it has been very easy to handle them in
> GRASS).
> But it was really a temporary solution (which has been in place for over six
> years),

I'm not advocating throwing it away today.  Obviously we'd want a migration
path and would want to retarget the interesting sites programs to use a
unified vector map approach.

My short list of problems with the sites format:

1.  No data definition in the header is required, therefore a heuristic is
used to parse.  Information in the header is mostly window dressing, since
it can't be relied upon.

2.  Limited data "types" (coordinates, doubles, a "category" number, and
strings).  No integers (other than "cat"), no dates, no booleans.

3.  No way to reference most attribute fields other than type/index combination.

4.  No NULL data support (missing data breaks the parser).

5.  Uses too many metacharacters, making parsing more complex than necessary
and requiring character escaping on input.

-- 
Eric G. Miller <egm2 at jps.net>