Error when using environmental RasterLayer and not RasterStack

[Rekyt]

Hi @sylvainschmitt ,
Thank you for building SSDM!
Because I'm working on simple simulations with a single variable I discovered a simple bug: when using a RasterLayer as the Env variable in the function modelling() I got an error.

Error in modelling("GLM", occ, env_raster, Xcol = "x", Ycol = "y", Pcol = "presence") : 
  Environment data set is not a raster or a raster stack

Here is a minimal reprex:

Minimal Reproductible Example

library("SSDM")
#> Welcome to the SSDM package, you can launch the graphical user interface by typing gui() in the console.
library("raster")
#> Le chargement a nécessité le package : sp

env_raster = raster(matrix(1:10))  # Create simple environmental gradient

occ = data.frame(presence = c(rep(1, 9), 0), cell = 1:10)  # Dummy occurences

# Extract cell coordinates
cell_xy = xyFromCell(env_raster, occ$cell)
occ$x = cell_xy[, "x"]
occ$y = cell_xy[, "y"]

# Not working model ###
raster_layer = modelling("GLM", occ, env_raster, Xcol = "x", Ycol = "y",
                         Pcol = "presence")
#> Data check ...
#> Error in modelling("GLM", occ, env_raster, Xcol = "x", Ycol = "y", Pcol = "presence"): Environment data set is not a raster or a raster stack

# Working model ####
raster_stack = modelling("GLM", occ, stack(env_raster), Xcol = "x", Ycol = "y",
                         Pcol = "presence")
#> Data check ... 
#>    done. 
#> 
#> Model evaluation...
#>    done. 
#> 
#> Model projection...
#>    done. 
#> 
#> Model axes contribution evaluation...
#>    done.

My workaround is currently to wrap my RasterLayer into a call of stack().
But it would great if it wouldn't be necessary when you only need a single environmental variable.

Possible bug origin: I think the problem comes from

SSDM/R/modelling.R

Line 298 in 971795b

if (!inherits(Env, "RasterStack")) {

It could be replaced by if (!inherits(Env, "Raster")) I'm going to check if it doesn't break anything before making a PR ;)

[Rekyt]

If you want to have a look, I did something at https://github.com/Rekyt/SSDM/tree/fix_raster_layer

Sidenote: I added context() to all tests to fix r-lib/testthat#700

I'm having issues currently with

SSDM/R/data.values.R

Line 10 in 971795b

values <- data.frame(extract(Env, cbind(obj@data$X, obj@data$Y)))

Because it creates weird variable names for RasterLayer such as extract.Env..cbind.obj.data.X..obj.data.Y.. and seems to make modelling() fail.

[sylvainschmitt]

There was supposed to be a security for this case that does not seems to work:

SSDM/R/modelling.R

Lines 295 to 297 in 939b53e

	if (is.raster(Env)) {
	Env <- stack(Env)
	}

I will fix it but anyway I would not much use an SDM with only one predictor.