Plot coordinates (image device) flipped on MacOSX about spatstat HOT 12 CLOSED

At present, reversing the order of ylim in a call to plot.owin only has an effect inside that plot call and not in subsequent plot commands. I'll investigate whether it's possible to cause a side-effect that persists in subsequent plot commands, but this may depend on the plot device.

For now, your best solution is to actually flip the objects W and X using affine()

from spatstat.

My comments above are incorrect - reversing the order of ylim in the call to plot.owin causes the coordinate system to be flipped (as one can check by inspecting par("usr") for example), and this coordinate system should be respected by subsequent plots when add=TRUE. So this is a bug somewhere, not a missing feature.

Could you please try adding the argument useRaster=FALSE to the image plot, i.e.

plot(density(X,adjust=0.2), add=TRUE, useRaster=FALSE)

Please try in the cases where you were getting the wrong behaviour before. This will give us a good idea whether it's a bug in the device driver.

from spatstat.

Thanks for the fast reply and analysis.
Indeed, the argument useRaster=FALSE did the trick.
plot(density(X,adjust=0.2),add=T,useRaster=F)
I tested the command on MacOSX and now it is displayed correctly.

Result:

Only the legend in the density plot is missing. But this is a result of the created axis and master plot I think.

from spatstat.

OK, thank you. So the fix is to set useRaster=FALSE when plotting the image.

This suggests that there is a bug in R's image display code (graphics::rasterImage) which I will chase up.

To get the legend in the image plot in your example, you can set show.all=TRUE. See help(plot.im)

from spatstat.

This appears to be a known R bug that was thought to have been repaired but is now happening again on some devices. I will pursue it. Thanks again for drawing this to our attention.

from spatstat.

The following code provides an example demonstrating the problem, that does not use spatstat at all.

M <- c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1,
   1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0,
   1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0,
   0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1,
   1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1,
   1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0,
   0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1,
   1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1,
   1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
   0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1,
   1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0,
   0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1,
   1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
   1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1,
   1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
   0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)

M <- matrix(M, 16, 21)

tryone <- function(flip=TRUE, raz=TRUE) {
  xl <- yl <- c(0,1)
  if(flip) xl <- rev(xl)
  plot(0,0,xlim=xl ,ylim=yl, type="n", xlab="", ylab="",
      main=paste("flip =", flip, ", useRaster =", raz))
  xx <- seq(0,1,length=16)
  yy <- seq(0,1,length=21)
  image(xx, yy, M, add=TRUE, useRaster=raz)
}

tryall <- function() {
  opa <- par(mfrow=c(2,2))
  tryone(FALSE, FALSE)
  tryone(FALSE, TRUE)
  tryone(TRUE, FALSE)
  tryone(TRUE, TRUE)
  par(opa)
  invisible(NULL)
}

tryall()

from spatstat.

It seems unlikely that this bug in the device drivers will be fixed soon. The bug tracker says that this bug was reported in 2014. The R help files for image.default and rasterImage contain disclaimers about the rendering of images on some devices.

from spatstat.

Thanks for the analysis and the example. So the solution at the moment is to use the raster argument.

Apropos, the argument show.all=TRUE only works if I plot the density in a default main plot. If I add it to an existing plot (see code above) it isn't displayed.

from spatstat.

Apropos, the argument show.all=TRUE only works if I plot the density in a default main plot. If I add it to an existing plot (see code above) it isn't displayed.

Hmm, ... that is supposed to work..... I can only imagine that perhaps the originally-declared plot area is not large enough to contain the ribbon at the intended position. Could you please try the original code but just insert W <- grow.rectangle(Frame(W), 0.2 * shortside(Frame(W))) before the plot commands, and then see if show.all=TRUE is working?

from spatstat.

In the current development version of spatstat I have given plot.im an extra argument workaround which should also fix the problem of incorrect spatial orientation of the image. Set workaround=TRUE to apply the workaround - but only when you know that the device driver is showing the image incorrectly.

from spatstat.

Yes you are right. if I increase the plot area it is displayed.
Here the result with 0.1:

X<- ppp(Particles$X, Particles$Y, c(0,imageSizeX), c(0,imageSizeY))
W <- Window(X)
W <- grow.rectangle(Frame(W), 0.1 * shortside(Frame(W)))
#Plot is visualized in ImageJ (Java) coordinates so we need an inverted coordinate parent plot!
plot(W, ylim=rev(W$yrange),main=NULL)
#Add a density plot and avoiding a main title here!
plot(density(X,adjust=0.2),add=T,useRaster=FALSE,show.all=TRUE,main="")
#Add the points!
plot(X, add=T)
#Add axis and title in custom font size!
axis(1,cex.axis=2)
axis(2, cex.axis=2)
title(main="Density",xlab="Width", ylab="Height", cex.lab=1.5,cex.main=3) 

Great, thanks, also for the workaround.

from spatstat.

That's good!

I've also upgraded grow.rectangle so that instead of grow.rectangle(W, 0.1 * shortside(Frame(W))) you can now just type grow.rectangle(W, fraction=0.1)

I think we can close this issue now.

from spatstat.