ggplot2 add annotation spanning facet edges

Question

I want to draw a rectangular annotation that will span over the facet borders in ggplot.

library(ggplot2)
myPlot <- ggplot(mpg, aes(displ, hwy)) + 
    geom_point() + 
    facet_grid(class ~ .)

# add annotation
myPlot +
  annotate("rect", xmin = 3, xmax = 4, ymin = -Inf, ymax = Inf, fill = "green", alpha = 1/5)

What I have so far:

I'd like to draw 1 large rectangle that spans the facet edges like so:

Is there a way to do this with built-in ggplot2 code or with ggforce or do I have to mess around with grid? My ideal use-case would still allow me to have myPlot as a ggplot object which is why I've avoided any complicated grid stuff until this point.

Answer 1

An approach that uses grid functions to edit your plot.

It is easy to draw a rectangle within a grid viewport. Is it possible to construct a grid viewport that will overlay the set of ggplot panels exactly? The answer is "Yes". The trick, for drawing the rectangle, is to get the x-axis "native" coordinates from the ggplot_build information into the grid viewport.

library(ggplot2)
library(grid)

plot <- ggplot(mpg, aes(displ, hwy)) + 
    geom_point() + 
    facet_grid(class ~ .) 

plot

# Construct a grid.layout that is the same as the ggplot layout
gt = ggplotGrob(plot)
lay = grid.layout(nrow = length(gt$heights), ncol = length(gt$widths),
                  widths = gt$widths, heights = gt$heights)

# Push to this viewport
pushViewport(viewport(layout = lay))

# Within the layout, push to a viewport that spans the plot panels.
pos = gt$layout[grep("panel", gt$layout$name), c("t", "l")]  # Positions of the panels
pushViewport(viewport(layout.pos.col = pos$l, layout.pos.row = min(pos$t):max(pos$t)))

# Get the limits of the ggplot's x-scale, including any expansion.
## For ggplot ver 2.2.1
# x.axis.limits = ggplot_build(plot)$layout$panel_ranges[[1]][["x.range"]]

## For ver 3.0.0
# axis.limits = ggplot_build(plot)$layout$panel_params[[1]]$x.range
# But possibly the following will be more immune to future changes to ggplot
x.axis.limits = summarise_layout(ggplot_build(plot))[1, c('xmin', 'xmax')]


# Set up a data viewport,
# so that the x-axis units are, in effect, "native", 
# but y-axis units are, in effect, "npc".
# And push to the data viewport.
pushViewport(dataViewport(yscale = c(0, 1), 
                          xscale = x.axis.limits))

# Draw the rectangle
grid.rect(x = 3, y = 0,
          width = 1, height = 1,
          just = c("left", "bottom"), default.units = "native",
          gp = gpar(fill = "green", col = "transparent", alpha = .2))

# Back to the root viewport
upViewport(0)