mask not using rasterOptions(tmpdir)

[Guillaume Drolet]

Line 141 in fmask.R:

r <- mask(r, n.ms, maskvalue = 0)

does not write to tmpdir even if rasterOptions(tmpdir = outdir) is called before running fmask() and most temporary rasters are effectively written to outdir. Looks like a bug in raster::mask.

Tried: r <- mask(r, n.ms, maskvalue = 0, filename = rasterTmpFile())

Didn't work. Still writing to C:\Users...\ (which is not the working directory) and filling up memory, which results in a crash of the function.

[etiennebr]

Add

clusterExport(cl, list(op = rasterOptions()))
clusterEvalQ(cl, c(library(raster), library(rgdal), rasterOptions(op))

[Guillaume Drolet]

added:

td <- options("rasterTmpDir")[[1]]
clusterExport(cl, list("td"), envir = environment())
clusterEvalQ(cl, c(library(raster), library(rgdal), rasterOptions(tmpdir = td)))

[etiennebr]

Is it robust ? Can you do that in a new session where rasterOptions() has never been called ? Doesn't seem to work.

[Guillaume Drolet]

its not robust and I experienced it this afternoon when I ran the code without first calling rasterOptions(tmpdir = tmpdir). It's a temporary fix...

[etiennebr]

It's not a fix then :)

[Guillaume Drolet]

Temporary: clunky but it works IF you do the right things.

Guillaume Drolet ing.f M.Sc. Chercheur en télédétection et dynamique forestière Direction de la recherche forestière Ministère des ressources naturelles du Québec Québec (Québec), Canada G1P 3W8 T : [phone] #6727 F : [phone]

http://www.mrn.gouv.qc.ca/forets

[etiennebr]

We should probably get rid of rasterOptions() and explicitly load the appropriate parameters. See #24 for other problems.