Skip to content

writeRaster does not preserve names when writing to NetCDF

3 messages · Barry Rowlingson, Gmail, Roger Bivand

Original
Barry Rowlingson
# 
This chunk of code writes a character vector to a file.  It needs two
dimensions, one that has to be big enough for the longest character
string, and one which has to be the length of the vector.

writeChars <- function(fn){
    nameDim = dim.def.ncdf("NamesDim","", 1:3)
    charDim = dim.def.ncdf("CharsDim","", 1:10)
    nameVar = var.def.ncdf("Names","",list(charDim,nameDim),"NA")
    nc = create.ncdf(fn, list(nameVar))
    v = c("red  ","green","blue ")
    for(i in 1:length(v)){
        put.var.ncdf(nc, nameVar, v[i], start=c(1,i), count=c(-1,1))
    }
    close.ncdf(nc)
}

Dumping the NetCDF gives:

data:
 CharsDim = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ;
 NamesDim = 1, 2, 3 ;
 Names =
  "red  ",
  "green",
  "blue " ;
}

Whether this is a good idea or not... Hmmmm....

I've noticed if you use `writeRaster` with format="CDF" you get the
data in one variable called "value", but if you use format="netCDF"
you get one named variable for each layer. I'm pretty sure these names
could be set and recovered, but currently they seem to be called Band1
etc. I think thats the GDAL driver at work:

"""
   This driver supports creation of netCDF file following the CF-1
convention. You may create set of 2D
   datasets. Each variable array is named Band1, Band2, ... BandN.
""" http://www.gdal.org/frmt_netcdf.html

Also, this section in writeRaster:

          NetCDF files have the following additional, optional,
          arguments: ?varname?, ?varunit?, ?longname?, ?xname?,
          ?yname?, ?zname?, ?zunit?

only seems to apply to format="CDF" and not "NetCDF", which seems to
ignore at least the ones I've tried. The "NetCDF" format seems to be
done via GDAL.

Is it worth putting any effort into this? I mean, is it worth *Robert*
putting any effort into this?

Barry
12 days later
Gmail
# 
Dear all,

I keep getting a memory error (vector too big) when overlaying hundreds 
of thousands of points with a polygon layer. The computer I am running 
this on is quite powerful, thus the error is just internal to R because 
of reaching a vector allocation too big in size. Is there any workaround 
to execute this operation? My goal is to overlay several point onto a 
habitat polygon layer, and remove all points falling outside that polygon.

Any help is much appreciated.

Thank you,
Francesco
Roger Bivand
#
On Mon, 25 Aug 2014, Francesco Tonini wrote:

            

      
      
      
    

        
Please recall that nobody can see over your shoulder. Always state the 
output of sessionInfo(), the code causing the error, the output of 
traceback() after the error, and some details of the objects concerned 
(counts of points, for example). Are you for example using over() methods 
from the sp package?

Powerful may also be misconfigured, so available RAM and any adjustments 
you have made to R defaults should be declared.

Do read the posting guide!

Roger