boxKernel() in rescale() produces "checkered" plot

[EhrmannS]

Hi there, I think I found another bug. Not 100% sure though, since am not fully able to deduce what the boxKernel() is actually supposed to do in rescale(). Anyway, when I do

  mat <- raster::as.matrix(spObj)
  mat_resc <- mmand::rescale(mat, 2, kernel)
  spObj_resc <- raster::raster(mat_resc)
  raster::extent(spObj_resc) <- raster::extent(spObj)
  sp::proj4string(spObj_resc) <- sp::proj4string(spObj)

spObj obviously is a binary raster with certain dimensions (97, 201, 19497 (nrow, ncol, ncell)) and spObj_resc with double the nrows/ncols and four times the ncells (194, 402, 77988 (nrow, ncol, ncell)), as expected. What is not expected though, is that when plotting spObj_resc (raster::plot(), but its the same with display() on mat_resc), I get a funny looking raster, where only the lower right cell (of four cells) is "filled" with a value, while the other three (presumably new created cells) have the value 0. This is also irrespective of binary or continuous raster.

Moreover, if I chose a values for factor of for instance 0.5, something similar arises, however on the scale of the whole plot, only the upper left fourth of the downscaled plot is shown.

These issues don't seem to arise at least with the mnKernel(), didn't test the others though.
spObj is still the same.

[jonclayden]

Thanks for the report, Steffen. I think this is fixed on the master branch now, but I'd be grateful if you could give it a try.

[EhrmannS]

Hi Jon, I just looked at it and ran a couple of tests. For rescale(x, factor = 0.5, kernel = boxKernel()) it seems to work now. The new object will have the value of the upper left cell. Here I was wondering, if it would be possible to set some offset in boxKernel()? Maybe mean and/or range of the "box"-graph?

However, if I go for rescale(x, factor = 2, kernel = boxKernel()), then I still get the above described behaviour, where only the lower right of four cells will have the value of the previous object. Is up-scaling even a built-in feature?

[jonclayden]

Hi Steffen. Yes, upscaling should work. To take a simple example, the following is the behaviour I see based on the current tip of the master branch.

library(mmand)
m <- matrix(c(1,1,1,0,1,1,0,0,1),3,3)
m
#      [,1] [,2] [,3]
# [1,]    1    0    0
# [2,]    1    1    0
# [3,]    1    1    1
rescale(m, 2, boxKernel())
#      [,1] [,2] [,3] [,4] [,5] [,6]
# [1,]    1    1    0    0    0    0
# [2,]    1    1    0    0    0    0
# [3,]    1    1    1    1    0    0
# [4,]    1    1    1    1    0    0
# [5,]    1    1    1    1    1    1
# [6,]    1    1    1    1    1    1
rescale(m, 0.5, boxKernel())
#      [,1] [,2]
# [1,]    1    0
# [2,]    1    1

This looks right to me. Do you get the same? If not, could you post your result and sessionInfo() please?

[EhrmannS]

All right, after restarting my R-Session your example (and also mine) now works. Thanks!