The relevant link to the ggplot2 approach was given in the comments. Copied from there:
library(grid)
g <- rasterGrob(blues9, width=unit(1,"npc"), height = unit(1,"npc"),
interpolate = TRUE)
# grid.draw(g)
library(ggplot2)
ggplot(mtcars, aes(factor(cyl))) + # add gradient background
annotation_custom(g, xmin=-Inf, xmax=Inf, ymin=-Inf, ymax=Inf) +
geom_bar() # add data layer
My own approach:
As usual, I cannot compete with the simple elegance of baptiste's solutions for problems with grid graphics, but here is my approach since I went to all that work:
gg.background.fill <- function(gg.plot, cols = "white", which = "x") {
#does not work with facets
stopifnot(which %in% c("x", "y"))
which1 <- if (which == "x") "width" else "height"
require(gridExtra)
g <- ggplotGrob(gg.plot)
#g <- ggplotGrob(p)
gg <- g$grobs
findIt <- vapply(gg, function(x) grepl("GRID.gTree", x$name, fixed = TRUE), TRUE)
n1 <- getGrob(gg[findIt][[1]], "grill.gTree", grep=TRUE)$name
n2 <- getGrob(gg[findIt][[1]], "panel.background.rect", grep=TRUE)$name
gg[findIt][[1]]$children[[n1]]$children[[n2]]$gp$fill <- cols
x <- gg[findIt][[1]]$children[[n1]]$children[[n2]][[which]]
w <- gg[findIt][[1]]$children[[n1]]$children[[n2]][[which1]]
attr <- attributes(x)
x <- seq(0 + c(w)/length(cols)/2, 1 - c(w)/length(cols)/2, length.out = length(cols))
attributes(x) <- attr
gg[findIt][[1]]$children[[n1]]$children[[n2]][[which]] <- x
w <- c(w)/length(cols)
attributes(w) <- attr
gg[findIt][[1]]$children[[n1]]$children[[n2]][[which1]] <- w
g$grobs <- gg
class(g) = c("arrange", "ggplot", class(g))
g
}
p1 <- gg.background.fill(p, colorRampPalette(c("red", "blue"))(100))
print(p1)
p2 <- gg.background.fill(p, colorRampPalette(c("red", "blue"))(100), "y")
print(p2)
This modifies the existing background which might be considered an advantage, but in contrast to the annotation_custom
approach it doesn't work with faceting. More work would be required for that.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…