How to use facets with a dual y-axis ggplot

Question

I have been trying to extend my scenario from here to make use of facets (specifically facet_grid()).

I have seen this example, however I can't seem to get it to work for my geom_bar() and geom_point() combo. I attempted to use the code from the example just changing from facet_wrap to facet_grid which also seemed to make the first layer not show.

I am very much a novice when it comes to grid and grobs so if someone can give some guidance on how to make P1 show up with the left y axis and P2 show up on the right y axis that would be great.

Data

library(ggplot2)
library(gtable)
library(grid)
library(data.table)
library(scales)

grid.newpage()

dt.diamonds <- as.data.table(diamonds)

d1 <- dt.diamonds[,list(revenue = sum(price),
                        stones = length(price)),
                  by=c("clarity","cut")]

setkey(d1, clarity,cut)

p1 & p2

p1 <- ggplot(d1, aes(x=clarity,y=revenue, fill=cut)) +
  geom_bar(stat="identity") +
  labs(x="clarity", y="revenue") +
  facet_grid(. ~ cut) +
  scale_y_continuous(labels=dollar, expand=c(0,0)) + 
  theme(axis.text.x = element_text(angle = 90, hjust = 1),
        axis.text.y = element_text(colour="#4B92DB"), 
        legend.position="bottom")

p2 <- ggplot(d1, aes(x=clarity, y=stones, colour="red")) +
  geom_point(size=6) + 
  labs(x="", y="number of stones") + expand_limits(y=0) +
  scale_y_continuous(labels=comma, expand=c(0,0)) +
  scale_colour_manual(name = '',values =c("red","green"), labels = c("Number of Stones"))+
  facet_grid(. ~ cut) +
  theme(axis.text.y = element_text(colour = "red")) +
  theme(panel.background = element_rect(fill = NA),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        panel.border = element_rect(fill=NA,colour="grey50"),
        legend.position="bottom")

Attempt to combine (based on example linked above) This fails in the first for loop, I suspect to the hard coding of geom_point.points, however I don't know how to make it suit my charts (or fluid enough to suit a variety of charts)

# extract gtable
g1 <- ggplot_gtable(ggplot_build(p1))
g2 <- ggplot_gtable(ggplot_build(p2))

combo_grob <- g2
pos <- length(combo_grob) - 1
combo_grob$grobs[[pos]] <- cbind(g1$grobs[[pos]],
                                 g2$grobs[[pos]], size = 'first')

panel_num <- length(unique(d1$cut))
for (i in seq(panel_num))
{
   grid.ls(g1$grobs[[i + 1]])
  panel_grob <- getGrob(g1$grobs[[i + 1]], 'geom_point.points',
                        grep = TRUE, global = TRUE)
  combo_grob$grobs[[i + 1]] <- addGrob(combo_grob$grobs[[i + 1]], 
                                       panel_grob)
}       


pos_a <- grep('axis_l', names(g1$grobs))
axis <- g1$grobs[pos_a]
for (i in seq(along = axis))
{
  if (i %in% c(2, 4))
  {
    pp <- c(subset(g1$layout, name == paste0('panel-', i), se = t:r))

    ax <- axis[[1]]$children[[2]]
    ax$widths <- rev(ax$widths)
    ax$grobs <- rev(ax$grobs)
    ax$grobs[[1]]$x <- ax$grobs[[1]]$x - unit(1, "npc") + unit(0.5, "cm")
    ax$grobs[[2]]$x <- ax$grobs[[2]]$x - unit(1, "npc") + unit(0.8, "cm")
    combo_grob <- gtable_add_cols(combo_grob, g2$widths[g2$layout[pos_a[i],]$l], length(combo_grob$widths) - 1)
    combo_grob <- gtable_add_grob(combo_grob, ax,  pp$t, length(combo_grob$widths) - 1, pp$b)
  }
}

pp <- c(subset(g1$layout, name == 'ylab', se = t:r))

ia <- which(g1$layout$name == "ylab")
ga <- g1$grobs[[ia]]
ga$rot <- 270
ga$x <- ga$x - unit(1, "npc") + unit(1.5, "cm")

combo_grob <- gtable_add_cols(combo_grob, g2$widths[g2$layout[ia,]$l], length(combo_grob$widths) - 1)
combo_grob <- gtable_add_grob(combo_grob, ga, pp$t, length(combo_grob$widths) - 1, pp$b)
combo_grob$layout$clip <- "off"

grid.draw(combo_grob)

EDIT to attempt to make workable for facet_wrap

The following code still works with facet_grid using ggplot2 2.0.0

g1 <- ggplot_gtable(ggplot_build(p1))
g2 <- ggplot_gtable(ggplot_build(p2))

pp <- c(subset(g1$layout, name == "panel", se = t:r))
g <- gtable_add_grob(g1, g2$grobs[which(g2$layout$name == "panel")], pp$t,
                     pp$l, pp$b, pp$l)
# axis tweaks
ia <- which(g2$layout$name == "axis-l")
ga <- g2$grobs[[ia]]
ax <- ga$children[[2]]
ax$widths <- rev(ax$widths)
ax$grobs <- rev(ax$grobs)
ax$grobs[[1]]$x <- ax$grobs[[1]]$x - unit(1, "npc") + unit(0.15, "cm")
g <- gtable_add_cols(g, g2$widths[g2$layout[ia, ]$l], length(g$widths) - 1)
g <- gtable_add_grob(g, ax, unique(pp$t), length(g$widths) - 1)

# Add second y-axis title
ia <- which(g2$layout$name == "ylab")
ax <- g2$grobs[[ia]]
# str(ax) # you can change features (size, colour etc for these -
# change rotation below
ax$rot <- 90
g <- gtable_add_cols(g, g2$widths[g2$layout[ia, ]$l], length(g$widths) - 1)
g <- gtable_add_grob(g, ax, unique(pp$t), length(g$widths) - 1)

# Add legend to the code
leg1 <- g1$grobs[[which(g1$layout$name == "guide-box")]]
leg2 <- g2$grobs[[which(g2$layout$name == "guide-box")]]

g$grobs[[which(g$layout$name == "guide-box")]] <-
  gtable:::cbind_gtable(leg1, leg2, "first")

grid.draw(g)

Answer 1

Now that ggplot2 has secondary axis support this has become much much easier in many (but not all) cases. No grob manipulation needed.

Even though it is supposed to only allow for simple linear transformations of the same data, such as different measurement scales, we can manually rescale one of the variables first to at least get a lot more out of that property.

library(tidyverse)

max_stones <- max(d1$stones)
max_revenue <- max(d1$revenue)

d2 <- gather(d1, 'var', 'val', stones:revenue) %>% 
  mutate(val = if_else(var == 'revenue', as.double(val), val / (max_stones / max_revenue)))

ggplot(mapping = aes(clarity, val)) +
  geom_bar(aes(fill = cut), filter(d2, var == 'revenue'), stat = 'identity') +
  geom_point(data = filter(d2, var == 'stones'), col = 'red') +
  facet_grid(~cut) +
  scale_y_continuous(sec.axis = sec_axis(trans = ~ . * (max_stones / max_revenue),
                                         name = 'number of stones'),
                     labels = dollar) +
  theme(axis.text.x = element_text(angle = 90, hjust = 1),
        axis.text.y = element_text(color = "#4B92DB"),
        axis.text.y.right = element_text(color = "red"),
        legend.position="bottom") +
  ylab('revenue')

It also works nicely with facet_wrap:

Other complications, such as scales = 'free' and space = 'free' are also done easily. The only restriction is that the relationship between the two axes is equal for all facets.

Answer 2

EDIT: UPDATED TO GGPLOT 2.2.0
But ggplot2 now supports secondary y axes, so there is no need for grob manipulation. See @Axeman's solution.

facet_grid and facet_wrap plots generate different sets of names for plot panels and left axes. You can check the names using g1$layout where g1 <- ggplotGrob(p1), and p1 is drawn first with facet_grid(), then second with facet_wrap(). In particular, with facet_grid() the plot panels are all named "panel", whereas with facet_wrap() they have different names: "panel-1", "panel-2", and so forth. So commands like these:

pp <- c(subset(g1$layout, name == "panel", se = t:r))
g <- gtable_add_grob(g1, g2$grobs[which(g2$layout$name == "panel")], pp$t,
                     pp$l, pp$b, pp$l)

will fail with plots generated using facet_wrap. I would use regular expressions to select all names beginning with "panel". There are similar problems with "axis-l".

Also, your axis-tweaking commands worked for older versions of ggplot, but from version 2.1.0, the tick marks don't quite meet the right edge of the plot, and the tick marks and the tick mark labels are too close together.

Here is what I would do (drawing on code from here, which in turn draws on code from here and from the cowplot package).

# Packages
library(ggplot2)
library(gtable)
library(grid)
library(data.table)
library(scales)

# Data 
dt.diamonds <- as.data.table(diamonds)
d1 <- dt.diamonds[,list(revenue = sum(price),
                        stones = length(price)),
                  by=c("clarity", "cut")]
setkey(d1, clarity, cut)

# The facet_wrap plots
p1 <- ggplot(d1, aes(x = clarity, y = revenue, fill = cut)) +
  geom_bar(stat = "identity") +
  labs(x = "clarity", y = "revenue") +
  facet_wrap( ~ cut, nrow = 1) +
  scale_y_continuous(labels = dollar, expand = c(0, 0)) + 
  theme(axis.text.x = element_text(angle = 90, hjust = 1),
        axis.text.y = element_text(colour = "#4B92DB"), 
        legend.position = "bottom")

p2 <- ggplot(d1, aes(x = clarity, y = stones, colour = "red")) +
  geom_point(size = 4) + 
  labs(x = "", y = "number of stones") + expand_limits(y = 0) +
  scale_y_continuous(labels = comma, expand = c(0, 0)) +
  scale_colour_manual(name = '', values = c("red", "green"), labels = c("Number of Stones"))+
  facet_wrap( ~ cut, nrow = 1) +
  theme(axis.text.y = element_text(colour = "red")) +
  theme(panel.background = element_rect(fill = NA),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank(),
        panel.border = element_rect(fill = NA, colour = "grey50"),
        legend.position = "bottom")


# Get the ggplot grobs
g1 <- ggplotGrob(p1)
g2 <- ggplotGrob(p2)

# Get the locations of the plot panels in g1.
pp <- c(subset(g1$layout, grepl("panel", g1$layout$name), se = t:r))

# Overlap panels for second plot on those of the first plot
g <- gtable_add_grob(g1, g2$grobs[grepl("panel", g1$layout$name)], 
      pp$t, pp$l, pp$b, pp$l)


# ggplot contains many labels that are themselves complex grob; 
# usually a text grob surrounded by margins.
# When moving the grobs from, say, the left to the right of a plot,
# Make sure the margins and the justifications are swapped around.
# The function below does the swapping.
# Taken from the cowplot package:
# https://github.com/wilkelab/cowplot/blob/master/R/switch_axis.R 

hinvert_title_grob <- function(grob){

  # Swap the widths
  widths <- grob$widths
  grob$widths[1] <- widths[3]
  grob$widths[3] <- widths[1]
  grob$vp[[1]]$layout$widths[1] <- widths[3]
  grob$vp[[1]]$layout$widths[3] <- widths[1]

  # Fix the justification
  grob$children[[1]]$hjust <- 1 - grob$children[[1]]$hjust 
  grob$children[[1]]$vjust <- 1 - grob$children[[1]]$vjust 
  grob$children[[1]]$x <- unit(1, "npc") - grob$children[[1]]$x
  grob
}

# Get the y axis title from g2
index <- which(g2$layout$name == "ylab-l") # Which grob contains the y axis title?   EDIT HERE
ylab <- g2$grobs[[index]]                # Extract that grob
ylab <- hinvert_title_grob(ylab)         # Swap margins and fix justifications

# Put the transformed label on the right side of g1
g <- gtable_add_cols(g, g2$widths[g2$layout[index, ]$l], max(pp$r))
g <- gtable_add_grob(g, ylab, max(pp$t), max(pp$r) + 1, max(pp$b), max(pp$r) + 1, clip = "off", name = "ylab-r")

# Get the y axis from g2 (axis line, tick marks, and tick mark labels)
index <- which(g2$layout$name == "axis-l-1-1")  # Which grob.    EDIT HERE
yaxis <- g2$grobs[[index]]                    # Extract the grob

# yaxis is a complex of grobs containing the axis line, the tick marks, and the tick mark labels.
# The relevant grobs are contained in axis$children:
#   axis$children[[1]] contains the axis line;
#   axis$children[[2]] contains the tick marks and tick mark labels.

# First, move the axis line to the left
# But not needed here
# yaxis$children[[1]]$x <- unit.c(unit(0, "npc"), unit(0, "npc"))

# Second, swap tick marks and tick mark labels
ticks <- yaxis$children[[2]]
ticks$widths <- rev(ticks$widths)
ticks$grobs <- rev(ticks$grobs)

# Third, move the tick marks
# Tick mark lengths can change. 
# A function to get the original tick mark length
# Taken from the cowplot package:
# https://github.com/wilkelab/cowplot/blob/master/R/switch_axis.R 
plot_theme <- function(p) {
  plyr::defaults(p$theme, theme_get())
}

tml <- plot_theme(p1)$axis.ticks.length   # Tick mark length
ticks$grobs[[1]]$x <- ticks$grobs[[1]]$x - unit(1, "npc") + tml

# Fourth, swap margins and fix justifications for the tick mark labels
ticks$grobs[[2]] <- hinvert_title_grob(ticks$grobs[[2]])

# Fifth, put ticks back into yaxis
yaxis$children[[2]] <- ticks

# Put the transformed yaxis on the right side of g1
g <- gtable_add_cols(g, g2$widths[g2$layout[index, ]$l], max(pp$r))
g <- gtable_add_grob(g, yaxis, max(pp$t), max(pp$r) + 1, max(pp$b), max(pp$r) + 1, 
   clip = "off", name = "axis-r")

# Get the legends
leg1 <- g1$grobs[[which(g1$layout$name == "guide-box")]]
leg2 <- g2$grobs[[which(g2$layout$name == "guide-box")]]

# Combine the legends
g$grobs[[which(g$layout$name == "guide-box")]] <-
    gtable:::cbind_gtable(leg1, leg2, "first")

# Draw it
grid.newpage()
grid.draw(g)