[GRASSLIST:2028] Re: Representation of "no data": we need your opinion

[Justin Hickey]

Hello Rod

Rod Paterson wrote:
> Why not allow the NULL value to be defined in the raster header file;
> This would be more flexible and only involves editing of the header;

One of the main reasons is that it would lead to inconsistancy with how
floating point data handles NULL. For floating point values, there is a
bit pattern that is an invalid floating point number. Thus, it is a
great candidate for use as the NULL value. Therefore, it would be a good
thing to fix the NULL value for floating point data to this bit pattern.
But this would be inconsistant with CELL data, since the user could
choose any value. Thus, this would lead to confusion. Users would either
try to pick a NULL value for floating point data and get an error
message, or they would forget to select one for CELL data and be
confused when their data that matches the default value comes up as
NULL.

> Using raster data from other software which uses this approach (ie 
> ERMapper) would be easier removing the need for any conversion.

There would still be a conversion to perform since GRASS has several
support files for raster maps which would need to be created. The
easiest way to import data in this manner is to use a conversion
program.
 
> Is there a performance penalty in doing this?

I don't think so, but dealing with the NULL file is definitely a
performance penalty.

-- 
Sincerely,

Jazzman (a.k.a. Justin Hickey)  e-mail: jhickey at hpcc.nectec.or.th
High Performance Computing Center
National Electronics and Computer Technology Center (NECTEC)
Bangkok, Thailand
==================================================================
People who think they know everything are very irritating to those
of us who do.  ---Anonymous

Jazz and Trek Rule!!!
==================================================================