Re: Incorrect netCDF file growth

• To: netcdfgroup@xxxxxxxxxxxxxxxx (NetCDF group)
• Subject: Re: Incorrect netCDF file growth
• From: gabor@xxxxxxxxxxxxxxxxxxxxxx
• Date: Thu, 26 Mar 1992 16:57:20 CST

>      From: "Glenn P. Davis" <davis@xxxxxxxxxxxxxxxx>

> Byte and short variable sizes are rounded up to the next even 32 bit boundary.
> This is so that 'seeks' are always done "words", not bytes within a word.
> This is true within "record" varables as well.
> 

Yes, it sounds reasonable, but seems not to be true in all cases.
For example, you have one variable and 9 NC_BYTE-long vector-records
were written into the file.  main() is fed by 1,2,3,4,5 etc., the
number of records:

---------------------------------------------------------------------*/
#include <stdio.h>
#include "netcdf.h"

int cdfid;
int v1_id, v2_id;
short v1;
char  v2;
char* cp;

main(argc,argv)
int argc;
char **argv;
{
        int n,i,ix[1];
        if(argc < 2)
          {
            n = 1;                              /* default to one record */
            cp="1";
          }
        else
          {
            n = atoi(argv[1]);  /* or request number of records */
            cp=argv[1];
          }
        docdf();                                /* define and create the file */
        v1 = 1;
        v2 = 2;
        for(i = 0; i < n; i++)  /* write the requested records */
        {
                ix[0] = i;
                ncvarput1(cdfid, v1_id, ix, (void *) &v1);
        }
        ncclose(cdfid);
}       
/*---------------------------------------------------------------------*/
docdf()
{
        int dims[3];
        int record_dim;
        int vector_dim;
        char name[100];
                
        strcat(name, "gabor.");
        strcat(name, cp);
        cdfid = nccreate( name, NC_CLOBBER);

        record_dim = ncdimdef(cdfid, "record", NC_UNLIMITED);
        vector_dim = ncdimdef(cdfid, "vector", 9);

        dims[0] = record_dim;
        dims[1] = vector_dim;

        v1_id = ncvardef (cdfid, "var1", NC_BYTE, 2, dims);
        ncattput(cdfid,NC_GLOBAL,"title",NC_CHAR,13,(void *)"CDF TEST DATA");
        ncendef (cdfid);
}
/*----------------------------------------------------------------------*/

And see the sizes I received:

-rw-r-----  1 gabor         152 Mar 26 16:27 gabor.1  <--   1 record
-rw-r-----  1 gabor         161 Mar 26 16:27 gabor.2  <--   2 records
-rw-r-----  1 gabor         170 Mar 26 16:27 gabor.3  <--   3 records
-rw-r-----  1 gabor         179 Mar 26 16:27 gabor.4  <--   4 records
-rw-r-----  1 gabor         188 Mar 26 16:27 gabor.5  <--   5 records

Absolute file size is not adjusted to word boundary. The increment
caused by adding one more record is always 9 bytes, it is not rounded
to word boundary.

********************************************************************************
********************************************************************************

And now the honey on the cake: I had TWO variables, Each contained
1,2,3 records. The record was 1,2,3 and 5 NC_BYTE long.

record = 1 NC_BYTE
variable 1 - NC_BYTE(1) : variable 2 - NC_BYTE(1)
file size w/ 1 record = 188  - 2 records = 196 - 3 records = 204 (diff = 8)

                                                                should be 2
record = 2 NC_BYTEs
variable 1 - NC_BYTE(2) : variable 2 - NC_BYTE(2)
file size w/ 1 record = 188  - 2 records = 196 - 3 records = 204 (diff = 8)
                                                                should be 4

record = 3 NC_BYTEs
variable 1 - NC_BYTE(3) : variable 2 - NC_BYTE(3)
file size w/ 1 record = 188  - 2 records = 196 - 3 records = 204 (diff = 8)
                                                                should be 6

record = 5 NC_BYTEs
variable 1 - NC_BYTE(5) : variable 2 - NC_BYTE(5)
file size w/ 1 record = 196  - 2 records = 212 - 3 records = 228 (diff = 16)
                                                                should be 10


NOW it was rounded!!! But when I had only ONE variable, it was not
rounded!?!? How does it come? //
********************************************************************************
********************************************************************************

                                Gabor Fichtinger

                                Scientific Visualization Group,
                                Center for High Performance Computing
                                The University of Texas System
                                Balcones Research Center, 1.154 CMS
                                10100 Burnet Road, Austin, TX, 78758-4497
                                Ph.  : (512) 471 2409
                                Ph.  : 1-800-262-2472/2409 (toll free)
                                Fax  : (512) 471 2445   
                                Email: gabor@xxxxxxxxxxxxxxx
                              __o
             __o             -\<,
            -\<,  __________O / O
___________O / O