How to place legends at different sides of plot (bottom and right side) with ggplot2?

Question

I'm trying to get a plot with two legends, one at the bottom and the other at the right side of the plot.

library(tidyverse)

df <- tibble(names = mtcars %>% 
  rownames(),
  mtcars)

#plot 1
p1 <- df %>% 
  filter(names == "Duster 360" | names == "Valiant") %>% 
ggplot(aes(x = as.factor(cyl), y = mpg, color = names)) +
  geom_point() +
  geom_hline(aes(yintercept = 20, linetype = "a")) +
  theme(legend.position = "bottom")

p1

Desired plot (legend should be outside of plot):

Answer 1

Update

As of ggplot2 >= 3.5.0 there is no longer a need for hacky approaches., Instead it is now possible to place the legends individually using the position= argument of the guides (see here):

library(ggplot2)

df <- mtcars
df$names <- rownames(mtcars)
df <- subset(df, names == "Duster 360" | names == "Valiant")

ggplot(df, aes(x = as.factor(cyl), y = mpg, color = names)) +
  geom_point() +
  geom_hline(aes(yintercept = 20, linetype = "a")) +
  guides(
    color = guide_legend(position = "right"),
    linetype = guide_legend(position = "bottom")
  )

Original Answer

Making use of cowplot you could do:

1. Extract the color guide from a plot without a linetype guide and legend.position = "right" using cowplot::get_legend
2. Making use of cowplot::plot_grid make a grid with two columns where the first column contains the plot without the color guide and the linetype guide placed at the bottom, while the color guide is put in the second column.

library(tidyverse)

df <- tibble(names = mtcars %>% 
               rownames(),
             mtcars)

p1 <- df %>% 
  filter(names == "Duster 360" | names == "Valiant") %>% 
  ggplot(aes(x = as.factor(cyl), y = mpg, color = names)) +
  geom_point() +
  geom_hline(aes(yintercept = 20, linetype = "a"))

library(cowplot)

guide_color <- get_legend(p1 + guides(linetype = "none"))

plot_grid(p1 + 
            guides(color = "none") + 
            theme(legend.position = "bottom"), 
          guide_color, 
          ncol = 2, rel_widths = c(.85, .15))