ggplot2: multiple plots in a single row with a single legend

Question

I want a combined plot of two plots + their legend like this:

library(ggplot2) 
library(grid)
library(gridExtra)
dsamp <- diamonds[sample(nrow(diamonds), 1000), ]    
p1 <- qplot(price, carat, data=dsamp, colour=clarity)
p2 <- qplot(price, depth, data=dsamp, colour=clarity)
g <- ggplotGrob(p1 + theme(legend.position="bottom"))$grobs
legend <- g[[which(sapply(g, function(x) x$name) == "guide-box")]]
grid.arrange(arrangeGrob(p1+theme(legend.position="right"),p2+theme(legend.position="none"),legend,ncol=3,widths=c(3/7,3/7,1/7)))

However I do not want to guess the width of the plots and legends (and specify ncol) but have it extracted from p1 and p2 as shown here.

So I expect I would need something like this (adapted code from the link):

grid_arrange_shared_legend_row <- function(...) {
  plots <- list(...)
  g <- ggplotGrob(plots[[1]] + theme(legend.position="right"))$grobs
  legend <- g[[which(sapply(g, function(x) x$name) == "guide-box")]]
  lwidth <- sum(legend$width)
  grid.arrange(
    do.call(arrangeGrob, lapply(plots, function(x)
      x + theme(legend.position="none"))),
    legend,
    ncol = length(plots)+1,
    widths = unit.c(rep(unit(1, "npc") - lwidth, length(plots)), lwidth))
}
grid_arrange_shared_legend_row(p1, p2)

but this is not arranging the two plots in one row but rather one column:

This question is similar to this one here but different in that I am asking for the adapted widths as well. I am using code extracts both from that question + answer and the github.

score 8 · Accepted Answer · answered Jan 15 '16 at 15:40

8

Why don't you use facetting?

library(reshape2)
dmelt <- melt(dsamp, id.vars = c("price", "clarity"), measure.vars = c("carat", "depth"))
ggplot(dmelt, aes(x = price, y = value, color = clarity)) +
  geom_point() +
  facet_wrap(~ variable, scales = "free")

answered Jan 15 '16 at 15:40

Roland

127,288
10
191
288

Thank you, this works great actually. Maybe you should answer the other question as well with this same code? It was quite high up my google results and would be a great reference for future users! – mts Jan 15 '16 at 16:01

user5029763 · Answer 2 · 2016-01-18T15:14:52.383

I usually use facet_wrap or facet_grid as @Roland suggested.

A while ago I had to use the grid.arrange (I wanted the labels on the y-axis to be colored in an specific order) and this is the function I came up with:

ggplot_shared_info <- function(...) {
  plots <- list(...)
  g <- ggplotGrob(plots[[1]])$grobs
  legend <- g[[which(sapply(g, function(x) x$name) == "guide-box")]]
  title <- g[[grep("plot.title", sapply(g, function(x) x$name))]]
  xaxis <- g[[grep("axis.title.x", sapply(g, function(x) x$name))]]
  yaxis <- g[[grep("axis.title.y", sapply(g, function(x) x$name))]]

  lwidth <- sum(legend$width)
  theight <- sum(title$height)
  xheight <- sum(xaxis$height)
  ywidth <- sum(yaxis$width)

  grid.arrange(
     title,
     arrangeGrob(
       yaxis,
       do.call(arrangeGrob, c(lapply(plots, function(x)
         x + theme(legend.position="none", 
                   plot.title = element_blank(),
                   axis.title = element_blank())), 
         nrow = 1)),
       legend,
       nrow = 1,
       widths = grid::unit.c(ywidth, unit(1, "npc") - ywidth - lwidth, lwidth)
     ),
     xaxis, 
     heights = grid::unit.c(theight, unit(1, "npc") - theight - xheight, xheight),
     ncol = 1
  )
}

Edit: now user can determine which of the listed plot elements should be 'joined'.

ggplot_shared_info <- function(..., elements = c('legend', 'title', 'yaxis', 'xaxis')) {
  plots <- list(...)
  g <- ggplotGrob(plots[[1]])$grobs

  legend <- g[[which(sapply(g, function(x) x$name) == "guide-box")]]
  lwidth <- sum(legend$width)
  title <- g[[grep("plot.title", sapply(g, function(x) x$name))]]
  theight <- sum(title$height)  
  xaxis <- g[[grep("axis.title.x", sapply(g, function(x) x$name))]]
  xheight <- sum(xaxis$height)
  yaxis <- g[[grep("axis.title.y", sapply(g, function(x) x$name))]]
  ywidth <- sum(yaxis$width)

  plots <- lapply(plots, function(x, elements = elements){
    if('legend' %in% elements) x <- x + theme(legend.position="none")
    if('title' %in% elements) x <- x + theme(plot.title = element_blank())
    if('xaxis' %in% elements) x <- x + theme(axis.title.x = element_blank())
    if('yaxis' %in% elements) x <- x + theme(axis.title.y = element_blank())
    x
  }, elements = elements)
  plots <- do.call(arrangeGrob, c(plots, nrow = 1))

  if('legend' %in% elements) 
    plots <- arrangeGrob(plots, legend, nrow = 1, widths = grid::unit.c(unit(1, "npc") - lwidth, lwidth))
  if('yaxis' %in% elements)
    plots <- arrangeGrob(yaxis, plots, nrow = 1, widths = grid::unit.c(ywidth, unit(1, "npc") - ywidth))
  if('title' %in% elements) 
    plots <- arrangeGrob(title, plots, ncol = 1, heights = grid::unit.c(theight, unit(1, "npc") - theight))
  if('xaxis' %in% elements)     
    plots <- arrangeGrob(plots, xaxis, ncol = 1, heights = grid::unit.c(unit(1, "npc") - xheight, xheight))
  grid.arrange(plots)
}

This works nicely, and should be even more general, thanks! Note that a missing title will lead to an error, i.e. for the given example should be called as: `ggplot_shared_info(p1+ggtitle("p1"), p2)` — mts, Jan 15 '16 at 16:54
Hadn't thought about that! I'll try to edit my code so that the user can determine which elements of the plot he whishes to be 'shared'. — user5029763, Jan 18 '16 at 13:34

ggplot2: multiple plots in a single row with a single legend

2 Answers2

Linked