12

Update 2023:

There is now a great package, ggh4x by @teunbrand
See this answer to the related the question

Q: In a ggplot, how can we modify the colors of the facet background based on the group.

Can there be more than one color for facet backgrounds?

Example:

I am using facet_grid (not facet_wrap) with multiple layers.

## Sample data
dat <- mtcars
## Add in some colors based on the data
dat$facet_fill_color <- c("red", "green", "blue", "yellow", "orange")[dat$gear]

## Create main plot
library(ggplot2)
P <- 
  ggplot(dat, aes(x = cyl, y = wt)) + 
  geom_point(aes(fill = hp)) + 
  facet_grid(gear + carb ~ .)

The background of ALL of the strips can be changed using:

P + theme(strip.background = element_rect(fill="red"))

main challenge:

However, I would like to change the color differently for different groups.

Something like the following (which of course does not work)

P + theme(strip.background = element_rect(fill=dat$facet_fill_color))
P + theme(strip.background = element_rect(aes(fill=facet_fill_color)))

related, but not an actual answer to this current question ggplot2: facet_wrap strip color based on variable in data set)

Ricardo Saporta
  • 54,400
  • 17
  • 144
  • 178

4 Answers4

14

For what it's worth, it's pretty straight-forward to adapt that previous gtable hack.

enter image description here

## Sample data
require(ggplot2)
dat <- mtcars
## Add in some colors based on the data
dat$facet_fill_color <- c("red", "green", "blue", "yellow", "orange")[dat$gear]

## Create main plot
p <- ggplot(dat, aes(x=cyl, y=wt)) + 
  geom_point(aes(fill=hp)) + facet_grid(gear+carb ~ .) +
  theme(strip.background=element_blank())

dummy <- p
dummy$layers <- NULL
dummy <- dummy + geom_rect(data=dat, xmin=-Inf, ymin=-Inf, xmax=Inf, ymax=Inf,
                           aes(fill = facet_fill_color))

library(gtable)

g1 <- ggplotGrob(p)
g2 <- ggplotGrob(dummy)

gtable_select <- function (x, ...) 
{
  matches <- c(...)
  x$layout <- x$layout[matches, , drop = FALSE]
  x$grobs <- x$grobs[matches]
  x
}

panels <- grepl(pattern="panel", g2$layout$name)
strips <- grepl(pattern="strip-right", g2$layout$name)
g2$grobs[strips] <- replicate(sum(strips), nullGrob(), simplify = FALSE)
g2$layout$l[panels] <- g2$layout$l[panels] + 1
g2$layout$r[panels] <- g2$layout$r[panels] + 2

new_strips <- gtable_select(g2, panels | strips)
grid.newpage()
grid.draw(new_strips)

gtable_stack <- function(g1, g2){
  g1$grobs <- c(g1$grobs, g2$grobs)
  g1$layout <- rbind(g1$layout, g2$layout)
  g1
}
## ideally you'd remove the old strips, for now they're just covered
new_plot <- gtable_stack(g1, new_strips)
grid.newpage()
grid.draw(new_plot)
Community
  • 1
  • 1
baptiste
  • 75,767
  • 19
  • 198
  • 294
4

This feels like a horrible hacky thing to do (to the extent that I am almost embarrassed to post this as an answer), but it is possible...

require(ggplot2);require(grid)

# Facet strip colours
cols <- rep( c("red", "green", "blue", "yellow", "orange")[rep( c(4,3,5),times=c(4,3,4))] , 2 )

# Make a grob object
Pg <- ggplotGrob(P)

# To keep track of strip.background grobs
idx <- 0 

# Find each strip.background and alter its backround colour...
for( g in 1:length(Pg$grobs) ){
    if( grepl( "strip.absoluteGrob" , Pg$grobs[[g]]$name ) ){
        idx <- idx + 1
        sb <- which( grepl( "strip\\.background" , names( Pg$grobs[[g]]$children ) ) )
        Pg$grobs[[g]]$children[[sb]][]$gp$fill <- cols[idx]

    }
}

# Plot
grid.newpage()
grid.draw(Pg)

enter image description here

baptiste
  • 75,767
  • 19
  • 198
  • 294
Simon O'Hanlon
  • 58,647
  • 14
  • 142
  • 184
4

Alternative search-and-replace strategy inspired by @SimonO'Hanlon's solution,

strips <- grep(pattern="strip-right", Pg$layout$name)

refill <- function(strip, colour){
  strip[["children"]][[1]][["gp"]][["fill"]] <- colour
  strip
}
cols <- rep_len(rep.int(c("blue", "green", "red"), c(4,3,4)), length(strips))
Pg$grobs[strips] <- mapply(refill, 
                           strip = Pg$grobs[strips], 
                           colour = cols,
                           SIMPLIFY = FALSE)

grid.newpage()
grid.draw(Pg)
baptiste
  • 75,767
  • 19
  • 198
  • 294
  • baptiste - @ handles only work in comments and chat! You're lucky I checked back to see this. +1. Still better than mine. – Simon O'Hanlon Jun 12 '14 at 08:24
  • it's a variation of the same idea, so I thought you might want to insert it in your answer. As you prefer, up to you. – baptiste Jun 12 '14 at 10:33
  • I think it's sufficiently different and useful to have an answer slot of it's own. I like both of these and will use them (when needs must I do something simillar to my solution - but usually to set border widths around `pch=21` points - the outline width is not set-able within ggplot2 and scales with point size which I don't like). – Simon O'Hanlon Jun 12 '14 at 10:35
2

There's a very nice and simple answer that is buried in the similar question ggplot2: facet_wrap strip color based on variable in data set that avoids the hacks, although you still have to manually specify the colors you want in strip_themed().

library(ggh4x)
#> Loading required package: ggplot2

# Only colour strips in x-direction
strip <- strip_themed(background_x = elem_list_rect(fill = rainbow(7)))

# Wrap variant
ggplot(mpg, aes(displ, hwy)) +
  geom_point() +
  facet_wrap2(~ class, strip = strip)
Cliff Bueno
  • 86
  • 4