[#1194] creating a ncdf file from a stack with named variables

a few examples of code that I had to alter, and I suspect there's a way to make it look simpler.

I read radar images which I have to crop.
I crop them creating a raster that I use as a shape:
shape <- crop(raster(radarbeelden, layer=1), extent(areaExtent[2, 'x'], areaExtent[1, 'x'], areaExtent[2, 'y'], areaExtent[1, 'y']))
shape[] <- 0

I need weighted inverse distance masks and I create them like this:
for(i in 1:nrow(descriptionData)) {
logdebug("computing inverse distance from %s (ground station %d)", descriptionData[i, 'location'], i, logger="fews.diagnostics")
temp <- setValues(shape, NA)
temp[cellIndices[i]] <- 1
temp <- setValues(shape, 1) / distance(temp)
temp[cellIndices[i]] <- 1
temp[cellIndices[-i]] <- 0
buffer <- c(buffer, temp)
}

(cellIndices[i] contains the index of the pixel relative to my i-th location in the cropped image).

previously I could write:
temp <- 1 / distance(temp)

now this shorter instruction gives error:
Error in function (classes, fdef, mtable) :
unable to find an inherited method for function "raster", for signature "RasterLayer"

with the `buffer` vector of rasters I create the brick...
previously I used a stack and I created it like this:
> raster(buffer)
now I use a brick and I write:
> do.call("brick", buffer)

accessing the timestamps is now a lot easier (and a huge lot more readable)!
radarbeelden@zvalue
instead of
sapply(attr(radarbeelden, 'layers'), attr, 'zvalue')

but what happened to a few basic functions `+` and `prod`... first I could write:
calibrated <- calibrated + prod(image, incidence, correctionFactor)
(the arguments in prod are rasters or just numeric)
now I must write this:
addendum <- image * incidence * setValues(shape, correctionFactor)
calibrated[] <- calibrated[] + addendum[]

with the due adjustments, it looks like working, but if you could make my programmer's life easier, I would be thankful!

MF

Mario

I am on US Pacific time.
When raster reads ncdf with something like "minutes since 1970-01-01T00:00:00+0000", it computes the actual data/time and puts that in the zvalue slot (at least when detected (when using the CF standards)). When writing the inverse would needs to be done. This would allow merging different data sets, but it is tricky stuff that perhaps needs to be redesigned.

Because you do the assignment yourself:
result@zvalue <- as.seconds(index(neerslagmeting))

and perhaps the seconds are probably correctly written to the file (coerced to numbers) this may not be an issue for you.

There were some issues with R-Forge this week. Seems better now.

adding a patch that I'm using for the released version.

Some additional arguments:

library(raster)
r <- raster(system.file("external/test.grd",
package="raster"))
b <- brick(r, sqrt(r), r+10)
b@zvalue <- 8:10
nc <- writeRaster(b, filename='netCDF.nc', format="CDF", overwrite=TRUE, xname='x', yname='y', zname='time', zunit='months', varname='rain', varunit='mm', longname='Daily rainfall')
print(nc)
str(nc)