Re: equivalents for find function in matlab?

Hi,

The following may be another solution to your specific problem.

  %----matlab----
   Ibad = find(f<=-999); %all bad points
   Igood = find(f>-999); %all good points
 
   f(Ibad) = griddata(x(Igood), y(Igood), f(Igood),x(Ibad), y(Ibad));

  ;---NCL-----
  f=mask(f,f.le.-999,False) ; flag unwanted number(s) as missing
  f=linmsg(f,-1) ; interpolate missing values

Documentation for mask and linmsg can be found here

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

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

Note, this is not as general as the "find" approach mentioned above
using matlab.

saji

* wenlong_at_hpl.umces.edu <wenlong_at_hpl.umces.edu> [2007-02-13 16:53:41 -0500]:

> Hi all,
>
> Thanks a lot to Mateus and Adam and others,
>
> I see the beauty of where() function. And also the ND-1D-ND switch by
> adam.
>
> Yet for a slightly more complicated situation, matlab find() is much
> more powerful.
>
> Say I have x(1:100,1:200), y(1:100,1:200) and f(1:100,1:200), where
> x and y are simply coordinates, f is a function, while some random
> locations of f is missing or nan or -999.0
>
> Then, I will want to substitue the missing/bad values with an
> interpolation instead of a single value for all the missing points. In
> matlab it is:
>
> %----matlab----
> Ibad = find(f<=-999); %all bad points
> Igood = find(f>-999); %all good points
>
> f(Ibad) = griddata(x(Igood), y(Igood), f(Igood),x(Ibad), y(Ibad));
>
> % where the griddata does a scattered data
> % interpolation for the locations of the bad values
> % using the good values
>
> %--------------
>
> I found it pretty hard to do in NCL. Any super suggestions?
>
> I'd really love to see the ind() function is improved so that it can
> handle similar things in the new NCL.
>
> function linmsg() seems to work for missing values, but not good a
> case where the condition f<=-999 is more complicated, for example f<=c,
> where c is a matrix two.
>
>
> Wen
>
>
> Quoting Adam Phillips <asphilli_at_cgd.ucar.edu>:
>
> >Thanks to Mateus for replying...
> >
> >The script Mateus provided will work for a one dimensional array, but
> >not a two dimensional array, as ind only operates on a 1D array. Thus:
> >
> >arr = ndtooned(A)
> >Ifind = ind(arr.gt.10.0)
> >arr(Ifind) = 100.
> >A = (/ onedtond(arr,dimsizes(A)) /)
> >
> >Also, in the next release of NCL a new function named where has been added:
> >http://www.ncl.ucar.edu/Document/Functions/Built-in/where.shtml
> >
> >this would allow the above to be done in one line, as where can operate
> >on a multi-dimensional array:
> >
> >A = where(A.gt.10.0,100,A)
> >
> >Adam
> >
> >Mateus da Silva Teixeira wrote:
> >>Hi Wen,
> >>
> >>In NCL you do it as follow:
> >>
> >> Ifind = ind ( A .gt. 10.0)
> >> A(Ifind) = 100.0
> >>
> >>very easy also!
> >>
> >>Mateus
> >>
> >>wenlong_at_hpl.umces.edu escreveu:
> >>>Hi all,
> >>>
> >>> I'm starting to learn NCL.
> >>>
> >>> Once thing I'd like to do is to set part of an array to a value
> >>>I want. In matlab is is really eazy:
> >>>
> >>> Say A(1:100,1:200) is a 2D float array,
> >>>
> >>> I can do the following to set all locations with value >10 to 100.0
> >>>
> >>>%----matlab----
> >>> Ifind=find(A>10.0); %find the indices that statisfy the condition
> >>> A(Ifind) =100.0; %reset the value to 100.0
> >>>%--------------
> >>>
> >>> Is there a fast way to do this in NCL? I mean, I really don't
> >>>want to use tedious do loops.
> >>>
> >>>
> >>> Thanks
> >>>
> >>> Wen
> >>>
> >>>
> >>>
> >>>_______________________________________________
> >>>ncl-talk mailing list
> >>>ncl-talk_at_ucar.edu
> >>>http://mailman.ucar.edu/mailman/listinfo/ncl-talk
> >>>
> >>
> >>_______________________________________________
> >>ncl-talk mailing list
> >>ncl-talk_at_ucar.edu
> >>http://mailman.ucar.edu/mailman/listinfo/ncl-talk
> >
> >--
> >--------------------------------------------------------------
> >Adam Phillips asphilli_at_ucar.edu
> >National Center for Atmospheric Research tel: (303) 497-1726
> >ESSL/CGD/CAS fax: (303) 497-1333
> >P.O. Box 3000
> >Boulder, CO 80307-3000 http://www.cgd.ucar.edu/cas/asphilli
> >
>
>
>
> _______________________________________________
> ncl-talk mailing list
> ncl-talk_at_ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

-- 
Saji N. Hameed
APEC Climate Center          				+82 51 668 7470
National Pension Corporation Busan Building 12F         
Yeonsan 2-dong, Yeonje-gu, BUSAN 611705			saji_at_apcc21.net
KOREA
_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk