udf with z and t

[Matthias Munnich]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Got it.  Try

ga-> set loopdim x

before executing your UDF.

... Matt

Henrique Barbosa wrote:
> Mathias,
>
> The problem is a bit more complicated because the calculation I
> want to do requires that I have, at the same time (i.e. at the same
>  call to the udf func), all Z and T values of the field.
>
> What tloop and define do is (from grads help):
>
> "When time is a varying dimension in the dimension environment, the
>  tloop function evaluates the expr at fixed times, then
> reconstructs the time series to obtain a final result that is time
> varying."
>
> Futher more, the grads help goes on and says:
>
> "The tloop function is provided as a way to obtain time series from
>  functions that themselves are not implemented to be able to
> operate when time is a varying dimension."
>
> That's exactly my point! I want to create an UDF function that IS
> implemented to operate when time is a varying dimension.
>
> As I said on my previous email, it seems it should work, since the
> header of the transfer file used to communicate between grads and
> the external executable has a field to tell which variables is
> inside the archive. From grads help:
>
> "Format of the function data transfer file (...) Argument records:
> (...) 1st record: This record contains 20 values, (...) 2 -- An
> index to identify the i dimension (idim). Options for the index
> are: -1    None 0    X dimension (lon) 1    Y dimension (lat) 2
> Z dimension (lev) 3    T dimension (time) "
>
> But when a call my function with fixed (x,y) and varying (z,t), I
> get idim=-1 and jdim=-1 when I read the transfer file inside the
> fortran code.
>
> Thank you,
>
> Henrique
>
> On 10/30/06, Matthias Munnich <munnich at atmos.ucla.edu> wrote: Hi
> Henrique,
>
> ga-> d  tloop(myudf(expr))
>
> should work.   You may like to read the discussion on "tloop"
> http://grads.iges.org/grads/gadoc/gradfunctloop.html and "set
> loopdim" in the docs to understand why this is necessary.
>
> It also explains why "define allundef=myudf(expr)" should also
> work.
>
> ... Matt
>
> Henrique Barbosa wrote:
>> Hi all,
>
>> I want to write a UDF that deals with field values varying in Z
>> and T dimensions. From the grads help files, it seems that it
>> should work:
>
>> http://www.iges.org/grads/gadoc/udf.html
>
>> ... At least there are fields in the header of the transfer file
>> that tell if the first and second varying dimensions are either
>> x, y, z or t.
>
>> My function is very simple,  it only reads the transfer file and
>> print the header info, and returns a undefined field. But when I
>> call it with fixed (x,y) and varying (Z,T), grads call the
>> function many times.
>
>> I mean, instead of calling the function only once with the 2D
>> subset:
>
>> x=cte, y=cte, z=z_min...z_max,t=t_min...t_max
>
>> Grads calls the udf one time for each pair (z,t), always passing
>> only a single value.
>
>> Does anyone know how to pass a varying (Z,T) to a UDF function?
>
>> Or is that feature that will only be implemented in the future?
>
>> I am using grads-1.9b4, precompiled binaries (rh9), on a Suse
>> 10.1 intel platform (i386).
>
>> Thank you,
>
>> Henrique
>
>
>>

- --
- --------------------------------
Matthias Munnich
Univ. of California, Los Angeles
Inst. of Geophysics and Planetary Physics
3845 Slichter Hall
Los Angeles, CA 90095-1567
Phone: +1-310-794 5899
Fax: +1-310-206 3051
Email: munnich at atmos.ucla.edu
- --------------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFRqfVidaDTh/cfhYRAoWUAJsETqs7ye30M7o8AQkX1AquJKIn/wCdH4Di
wfTGv9eGizCHjqyMVviCNyc=
=lpEJ
-----END PGP SIGNATURE-----