Re: 19980611: Inquiry about netCDF

>To: "'address@hidden'" <address@hidden>
>From: "Yoon, Hoon (CICG - NY Program Trading)" <address@hidden>
>Subject: Inquiry about netCDF
>Organization: .
>Keywords: 199806111528.JAA12217

Hi Hoon,

>    I am interested in netCDF. I have questions that I want to address prior
> to spending resources. I promise they are simple.
> 
> 1.  How does one represent null or NA values?

If you want a variable to have null or NA values, create a netCDF
attribute for that variable that specifies the special value you will
use for such values.  A predefined attribute you can use is
"_FillValue", so for example if you want to specify "-99" as the null
value for integers stored in a variable named foo, just define

  foo:_FillValue = -99

Then if you try to read foo data that hasn't been written yet, it will
return this value.

If you have multiple kinds of null values, you can define additional
variable attributes, for example foo:missing_value, etc.  The
"_FillValue" attribute is treated specially by the netCDF library: it is
used for reading data that wasn't written.  More information on missing
values is available in the User's Guide

  
http://www.unidata.ucar.edu/packages/netcdf/guidec/guidec-12.html#HEADING12-1398
and
  http://www.unidata.ucar.edu/packages/netcdf/guidec/guidec-13.html#MARKER-2-175

> 2.  Can I mix char and numeric in an array?

No.  An array is defined to be a set of values of the same type.  If you
want to store associated character and numeric values, you will need to
store them in two different arrays.  If the two arrays are of the same
shape, you can associate the numeric value in one with the character
value in the other array at the same index.  If you want to store a
string associated with each value, you will need to add an extra
dimension to the character array for the maximum string length.

> 3.  If I had 2 or 3 dimentional array, do I need to redimention it every
> time if I need to get it to increase on the other (non-planned side)?
> Example:
>           IBM, AAPL, BORL
> 19980101   100,  23,  8
> 19980102   101,  24,  7
>     If I need additional date and that was the planned add dimention, I
> guess that won't be problem; however, if I need to increase number of
> stocks, do I have to create a new storage array and copy over? Could you
> tell me what this really takes to get it done?

Yes, to change the shape of an array in any way other than increasing
the size of the unlimited dimension requires creating a new array of the
desired shape and copying the old data into the new array.

The best way to do this would probably be to just create a new netCDF
file with the desired shape of variables and copy the old data into it.
This is clearly not something you would want to do very often; it's
similar to redefining the schema of a database, which is usually
costly.  You can choose to avoid this at the expense of preallocating
variables at the maximum size you anticipate using, or some strategy
that doubles the dimension size whenever a dimension grows beyond the
current maximum ...

If you need to do this frequently, netCDF might not be the right choice
for storing your data in a platform-independent way.  It's not a
database system, but an interface for array-oriented data that can be
represented as named variables with named attributes and dimensions.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu