Re: 20020618: netCDF: Problem displaying array of values

>To: address@hidden
>From: Naveenta Anand <address@hidden>
>Subject: Re: 20020618: netCDF: Problem displaying array of values
>Organization: .
>Keywords: 

Hi Naveenta,

> On June 20, 2002 1:05 PM, Russ Rew [SMTP:address@hidden] wrote:
> ...
> > The reason ncdump does not show any of the other variables, is because
> > they are record variables (use the unlimited dimension) and you have
> > not written any records yet:
> >
> >     PRIMARY_DIMENSION = UNLIMITED ; // (0 currently)
> >
> > The (0 currently) says there are currently no records.  Only the
> > LATITUDE and LONGITUDE variables are not record variables, so they
> > appear in the ncdump output, each having a single fill value:
> >
> >     data:
> >
> >      LATITUDE = _ ;
> >
> >      LONGITUDE = _ ;
> >
> > The reason you aren't writing any records (or any data) is because you
> > are specifying a 0 count for how much data to write:
> >
> >     size_t count[] = {primary_dimension_len};
> 
> ------------> Ok! I agree to this point. I changed the count now but
> there is no change and thus no display in netcdf file. Why would
> that be?  And also how do UNLIMITED dimensions actually work? I
> cannot specify the unlimited_dimesnion_length, right? But I do want
> the length to grow with my # of records and then how does the length
> reflect in definition in netcdf file [PRIMARY_DIMENSION = UNLIMITED
> ; // (0 currently)].And there is something else not right that I
> cannot figure out!! I am attaching my function along (ascii file) so
> you can take a look at the program (function for Russ) as a whole.

You don't specify the unlimited dimension length, but it is stored in
the file as the number of records written, and it automatically gets
increased when you write new records.  So for example, if you write
the 100th record of a variable without having written records 0 to 99,
all the records get written with fill values before record 100 is
written.  Then later if you write record 50, that overwrites the fill
values, but the file still has 100 records.  The length of the
unlimited dimension is the greatest number of records written for any
record variable so far.  You could write the records backward if you
want, starting with record 100, record 99, record 0, and then later
write record 101.

The record number is just the value of the first index (index[0]) of a
record variable.  You write records just like you would write other
slices of data, specifying the corner where to start the write with
the "index" array and specifying how far along each dimension to write
with the "count" array.  So if you are writing 2 records starting at
record 50, you would have index[0] = 49 and count[0] = 2.

> I am also attaching the current netcdf output (again ascii file called 
> temp.netcdf)

I'm not sure I see how you changed the count or whether it is right.
You have the statements

        size_t index[]  = {0};
        size_t count[]  = {1}; 
        size_t index1[] = {0};
        size_t count1[] = {1};
      ...
        count[0] = records;

indicating you want to write "records" records, but what is the value
of "records"?  It looks like it should be the number of input records
you read above:

  int records = 0;

  printf("Entered the function \n");


    while(read_twl_station(infile)!=1)
    { 
      records++;
  ...
    }
  printf(" %d records have been read \n",records);

but you didn't send the output from running the program, so I don't
know from looking at the code what the value of "records" is.  Also
from the structure of your code, it looks like you are trying to read
all the ASCII data in initially in the above while loop, and then
trying to write all the data to the netCDF file.  At least that's what
I think setting

        count[0] = records;

means.  But then you only give the address of a single value,

        stat = nc_put_vara_float(ncid,sea_level_id,index,count,&fill);

so if count[0] was anything except 1, this would not work.  For
example if records was 2, count[0] would be 2, and you would have to
provide the address of an array of two values instead of just &fill.

Maybe the structure you want is to just write a single record of data
after you read a record of data from the input.  In that case, your
program would look something like:

    // create the netCDF file, defining dimensions, variable, and attributes
   ...
    // Leave define mode
   ...
    // read in each record and write it to the netCDF file
    while(read_twl_station(infile)!=1)
    { 
      // read in a record of data
      ...
      // write the record of data to the netCDF file
        size_t index[]  = {0};
        size_t count[]  = {1}; 
        size_t index1[] = {0};
        size_t count1[] = {1};
      // call nc_put_var_... for all the record variables

    }

where you write each record out after you read it in, instead of
reading in all the data and then writing it all out.  In this case,
you would definitely want to just leave count[0] set to 1 to indicate
you are writing one record's worth of data and increment index[0] for
each record in the loop to indicate that you are adding a record.

--Russ

_____________________________________________________________________

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