Re: binary data reading problem

One of the issues with posting a GrADS script to ncl-talk
is that the post assumes that people in ncl-talk are familiar
with GrADS. That is not the case.

A few random comments:

   [0] There is a whole section of the Applications page
       devoted to GrADS files. Please read the following:

           http://www.ncl.ucar.edu/Applications/grads.shtml

       I highly recommend that you use the Climate Data Operators
       (CDO) to convert the GrADS binary to netCDF. It can be done
       from the command line.

------
NCL comments:

   [1] GrADS can produce sequential or direct (stream) binary.
       Which did you use? If you did not explicitly specify this,
       then what is the GrADS default?

       sequential: fbinrecread
       direct: fbindirread

   http://www.ncl.ucar.edu/Document/Functions/Built-in/fbinrecread.shtml
   http://www.ncl.ucar.edu/Document/Functions/Built-in/fbindirread.shtml

   [2] Is the binary big-endian or little-endian?

  http://www.ncl.ucar.edu/Document/Functions/Built-in/setfileoption.shtml

       setfileoption("bin","ReadByteOrder","LittleEndian")
                    or
       setfileoption("bin","ReadByteOrder","BigEndian")

   [3] Not sure but I think GrADS grid ordering is like fortran's
       (mlon,nlat,klev,ntim). NCL's is like C (ntim,klev,nlat,mlon)

        Hence, your array order should be the reverse. Also, NCL record
        ordering starts at 0 not 1. Please carefully read the
        documentation. If 'thf' is the 1st record, 'nsf' the 2nd record,
        etc:

        thf = new ((/ntim,klev,nlat,mlon),"float")
        nsf = ths
        .........
        nt = -1
        do nyr=1,nyrs
          do nmo=1,12
             nt = nt+1
             thf(nt,:,:,:) =
               fbindirread(path,nt,(/klev,nlat,mlon/),"float")
or
               fbinrecread(....)
             nsf(nt,:,:,:) = ...

          end do
        end do

same storage order as a FORTRAN array,

On 2/19/14, 3:17 PM, Jatin Kala wrote:
> Hi,
> I’ve had to deal with binary data (had no choice) for a few projects.
> You’ve probably messed up the order of the dimensions or something when reading the variable.
> My advice: Don’t write to binary format, it’s a Pain (in my opinion).
> I don’t use grads, but it seems you can write to netcdf:
> http://www.iges.org/grads/gadoc/gradcomdsdfwrite.html
> I would use that instead, then, reading in NCL is a breeze.
> Cheers,
> Jatin.
>
>
>
> On 20 Feb 2014, at 4:52 am, Colfescu Ioana <colfiii@yahoo.it> wrote:
>
>> Hi,
>>
>> I'm trying to read in NCL binary data I wrote in GrAds and I don't understand what error I'm doing so that NCL mixes the values.
>> The data I m trying to read is written as follows :
>> while( day <= daymax )
>> 'thfrn=thfr-tr(t-'shift')'
>> 'set x 1 128'
>> 'set y 1 64'
>> 'd thfrn'
>> 'd nsfrn'
>> 'd zwsrn'
>> 'd mwsrn'
>> 'd fwfrn'
>> day=day+1
>> endwhile
>> mo=mo+1
>> endwhile
>> year=year+1
>> endwhile
>>
>> I try to read the data like :
>> thf=fbindirread(path,1,(/64,128,5,nt/),"float").
>> I get the values but they are mixed. Does anyone know what am I doing wrong ?
>> Thanks,
>> _______________________________________________
>> ncl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Wed Feb 19 16:39:03 2014