ggRGB unable to plot masked image

[bappa10085]

ggRGB unable to plot the masked image. Here is a minimal, reproducible example

library(terra)
library(ggplot2)
library(RStoolbox)

lsat

plotRGB(lsat, r = 3, g = 2, b = 1, stretch = "hist")


# SpatVector as standalone object
m <- structure(list(geom = "POLYGON ((620947.959184 -412745.345369, 621312.951334 -414249.11303, 623254.709576 -414526.507064, 623897.095761 -413154.136578, 621152.354788 -411767.166405, 620947.959184 -412745.345369))"),
               class = "data.frame",
               row.names = c(NA, -1L)
)

# To SpatVector with the right CRS
m <- vect(m$geom)
crs(m) <- crs(lsat)

plot(m, add = TRUE, col = "red")

lsat.masked <- crop(lsat, m, mask = T)

plotRGB(lsat.masked, r = 3, g = 2, b = 1, stretch = "hist")

ggplot() +
  ggRGB(img = lsat.masked,
    r = 3,
    g = 2,
    b = 1,
    stretch = "hist",
    ggLayer = T)

It returns the following error

Error in grid.Call.graphics(C_raster, x$raster, x$x, x$y, x$width, x$height, :
cannot allocate memory block of size 67108864 Tb
In addition: Warning message:
In matrix(z, nrow = nrow(rr), ncol = ncol(rr), byrow = TRUE) :
data length [5888] is not a sub-multiple or multiple of the number of rows [88]

[KonstiDE]

Hey,

unfortunately, I am not really able to reproduce the error. For me, already the crop call yields [crop] cannot crop a SpatRaster with an empty extent. Furthermore, the polygon you create has no CRS, valid extent whatsoever?

I recommend getting the last.rds from the inst/external directory path, save it as tif, load it into qgis and draw your desired polygon there. Then export back into R crop.

Would that be a solution?

[bappa10085]

I have edited the code to make the error reproducible. Please have a look.

[KonstiDE]

Interestingly, I still cant reporduce the following part of the error:

Error in grid.Call.graphics(C_raster, x$raster, x$x, x$y, x$width, x$height, :
cannot allocate memory block of size 67108864 Tb

Anyhow, you seem to be correct, I'll look into the error this week or next weekend. For now, just do not plot the RGB as a ggLayer, but as a base ggplot using the following snippet (also do not forget to set geom_raster = T):

library(terra)
library(ggplot2)
library(RStoolbox)

lsat

plotRGB(lsat, r = 3, g = 2, b = 1, stretch = "hist")


# SpatVector as standalone object
m <- structure(
  list(geom = "POLYGON ((620947.959184 -412745.345369, 621312.951334 -414249.11303, 623254.709576 -414526.507064, 623897.095761 -413154.136578, 621152.354788 -411767.166405, 620947.959184 -412745.345369))"),
  class = "data.frame",
  row.names = c(NA, -1L)
)

# To SpatVector with the right CRS
m <- vect(m$geom, crs = crs(lsat))

plot(m, add = TRUE, col = "red")

lsat.masked <- crop(lsat, m, mask = T)

plotRGB(lsat.masked, r = 3, g = 2, b = 1, stretch = "hist")

ggRGB(img = lsat.masked, r = 3, g = 2, b = 1, stretch = "hist", geom_raster = T) + 
   geom_whatever()

[KonstiDE]

Hi, soooo this isn't actually a bug. We are just missing scale_fill_identity(). In ggRGB() the scale_fill_identity() is only returned from the method if ggLayer is false and a whole complete ggplot() build up.

When trying to return the scale_fill_identity() property without a ggplot() directly appended, a ggproto type of object is created which somehow cannot be passed back to the ggplot() + call where is replaces ggRGB(). You should be good to go with the following snippet:

library(terra)
library(ggplot2)
library(RStoolbox)

lsat

plotRGB(lsat, r = 3, g = 2, b = 1, stretch = "hist")


# SpatVector as standalone object
m <- structure(
  list(geom = "POLYGON ((620947.959184 -412745.345369, 621312.951334 -414249.11303, 623254.709576 -414526.507064, 623897.095761 -413154.136578, 621152.354788 -411767.166405, 620947.959184 -412745.345369))"),
  class = "data.frame",
  row.names = c(NA, -1L)
)

# To SpatVector with the right CRS
m <- vect(m$geom, crs = crs(lsat))

plot(m, add = TRUE, col = "red")

lsat.masked <- crop(lsat, m, mask = T)

plotRGB(lsat.masked, r = 3, g = 2, b = 1, stretch = "hist")

ggplot() + 
   ggRGB(img = lsat.masked, r = 3, g = 2, b = 1, stretch = "hist", geom_raster = T, ggLayer = T) + 
   scale_fill_identity()

If this does not work for you, please reopen this issue. Imma close now :)