[R-pkg-devel] Problem with S4 method for plotting a raster file

[Duncan Murdoch]

On 14/04/2021 5:41 a.m., Jacob Nabe-Nielsen wrote:
> Dear all,
> 
> I have just discovered an error in a package that I have already uploaded to
> CRAN, and cannot figure out how to fix it. All help will be greatly appreciated.
> The package is called DEPONS2R, and is available from:
> https://cran.r-project.org/web/packages/DEPONS2R/index.html
> 
> The problem is related to one of the included examples, where I plot the coastline
> on top of a raster file. In the example, where the plotting is embedded in
> an S4 method called plot.DeponsRaster, the coastline does not show. But if
> I run the functions that the method is composed of, everything works fine (see
> below). So I am wondering if the graphics devise is not passed on correctly
> within the package, or what actually happens.
> 
> The example that does not run correctly is available from ?plot.DeponsRaster:
> 
> library(DEPONS2R)
> 
> data("bathymetry")
> plot(bathymetry)
> data("coastline")
> library(rgdal)
> # Change projection of coastline to match that of bathymetry data
> coastline2 <- spTransform(coastline, crs(bathymetry))
> plot(coastline2, add=TRUE, col="lightyellow2")
> 
> 
> Here the coastline is not plotted. The method is defined as:
> 
> 
> setMethod("plot", signature("DeponsRaster", "ANY"),
>            function(x, y, maxpixels=500000, col, alpha=NULL, colNA=NA, add=FALSE,
>                     ext=NULL, useRaster=TRUE, interpolate=FALSE, addfun=NULL,
>                     nc, nr, maxnl=16, main, npretty=0, axes=TRUE,
>                     legend=TRUE, trackToPlot=1, ...)  {
>              oldpar <- graphics::par(no.readonly = TRUE)
>              on.exit(graphics::par(oldpar))
>              if (missing(main)) {
>                main <- paste(x using landscape, x using type, sep=" - ")
>              }
>              # Define colours specific or 'type'
>              if(missing(col) && x using type=="bathymetry") {
>                tmp.col <- grDevices::rainbow(1000)[501:800]
>                col <- c(tmp.col[1:100], rep(tmp.col[101:250], each=5))
>              }
>              # Use {raster}-package for plotting
>              if(x using crs=="NA") {
>                crs2 <- sp::CRS(as.character(NA))
>              } else {
>                crs2 <- sp::CRS(x using crs)
>              }
>              rdata <- raster::raster(x=x using data, xmn=x using ext$xleft, xmx=x using ext$xright,
>                                      ymn=x using ext$ybottom, ymx=x using ext$ytop,
>                                      crs=crs2)
>              raster::plot(rdata, col=col, main=main,
>                           alpha=alpha, add=add, ext=ext, axes=axes, legend=legend)
>            }
> )
> 
> 
> But when I use the functions that are at the core of this method it is
> not a problem adding the coastline correctly:
> 
> x <- bathymetry
> crs2 <- raster::crs(x using crs)
> rdata <- raster::raster(x = x using data, xmn = x using ext$xleft,
> xmx = x using ext$xright, ymn = x using ext$ybottom, ymx = x using ext$ytop,
> crs = crs2)
> raster::plot(rdata)
> plot(coastline2, add=TRUE, col="lightyellow2")
> 
> If would be fantastic if someone can figure out how I should modify the method
> to plot the coastline correctly on top of the map.
> 
> 

I'm not sure exactly what is going wrong, but if you plot the coastline 
first and the bathymetry second, things do show up.  Maybe that's enough 
of a hint for you:

library(DEPONS2R)
data("coastline")
library(rgdal)
coastline2 <- spTransform(coastline, crs(bathymetry))
plot(coastline2, col="lightyellow2")

data("bathymetry")
plot(bathymetry, add=TRUE)

This puts the legend on top of the plotted coastline data, so my guess 
is that the limits of the bathymetry plot aren't being made available to 
the coastline plot when you do things in the original order.

Duncan Murdoch