Re: nc_add_recs.m

["David Wang" <climater@xxxxxxxxx>]

Hi Dave and John,

thanks for providing alternative workarounds. i actually came up with a somewhat awkward solution. i dumped the time series of the dimensional order (1, 1, ntime) into the netcdf file using nc_add_recs.m, then permuted the dimensions to the correct order (ntime, 1, 1) using ncpdq. it's not an elegant solution, but a working one.

thanks,
d.

On 8/4/07, John Caron <caron@xxxxxxxxxxxxxxxx> wrote:

My understanding is that you can use the netcdf-java library in MatLab. I dont know the details, but Rich Signell has contributed this doc:

   http://www.unidata.ucar.edu/software/netcdf-java/nj22matlab.m

David Wang wrote:
> Hello,
>
> i've been scratching my head how to add a three-dimensional (time, lat,
> lon) variable into a netcdf file where time is the unlimited dimension
> and both lat and lon are dimensions of size 1. i work with matlab
> (nc_add_recs.m in snctools) but matlab doesn't allow trailing singleton
> dimensions. so while nc_add_recs.m (and nc_varput.m as well presumably)
> requires an input of rank 3, i'm not able to provide it an array of size
> (ntime, 1, 1). it's a pain in the neck. if there is any workaround out
> there, please give me a hint.
>
> thanks,
> david
>

-- 
"Now you have found your last wish", she said finally, "what you really and truly want is to love."
"But why can't I, Dame Eyola?"
"You won't be able to until you have drunk of the Water of Life", she said, "And you can't go back to your own world unless you take some of it back for others."

THE NEVERENDING STORY