How to modify a continous ggplot legend in R

Question

I'm using ggplot to graph the bias of the mean temperature in South America.

In order to graph this, I'm using the following code:

ggplot(bias.df2) +  
  geom_tile(aes(x=x, y=y, fill=pr), alpha=0.8) + 
  scale_fill_viridis(na.value="white") +
  coord_equal() +
  theme(legend.position="bottom") +
  theme(legend.key.width=unit(2, "cm"))+
  scale_color_continuous(limits=c(-10,10),breaks=brkbias)

Which generates the following color bar:

But I need the color bar to look like this:

I've tried modifying the limits of the colourbar but nothing seems to generate the kind of plot that I need. Would really appreciate any help to solve this.

EDIT: The data I'm using can be found in this csv file.

What part of the second colourbar would you like? The discreteness of the colours or the triangle ends? Triangle ends are discussed [here](https://stackoverflow.com/q/68440366/11374827). — teunbrand, Jun 01 '22 at 15:09
I was trying to get both the discreteness and the triangles. — jnava1612, Jun 01 '22 at 15:29
@teunbrand while using your suggestion I was able to get the triangles in the scale, but I can't get the descretization of the scale values and colors, as the colors in the new scale aren't being applied to the graph. The resulting graph looks like [this](https://drive.google.com/file/d/1ad3tRg_kzNJm0KlC-7KsMVtlSqQm-JSg/view?usp=sharing). — jnava1612, Jun 01 '22 at 16:07
Yeah I suppose you'd have to do that manually. I'm currently looking into making these types of guides in an easier way, but haven't gotten around to discrete ones yet. — teunbrand, Jun 01 '22 at 16:16
you can also [build the guide as a separate plot](https://stackoverflow.com/a/50540633/7941188) — tjebo, Jun 01 '22 at 16:23

score 5 · Accepted Answer · answered Jun 01 '22 at 16:34

guide_colorsteps() can be used in combination with the breaks = argument of scale_fill_viridis_c() to print a discrete colorbar with the continuous fill. We can modify the code linked in the comments to add the triangles to that guide. Since you have a horizontal bar, it requires moving and pointing the triangles horizontally instead of vertically.

library(ggplot2)
library(gtable)
library(grid)

bias.df2 <- read.csv("~/Downloads/data.csv")

my_triangle_colorsteps <- function(...) {
  guide <- guide_colorsteps(...)
  class(guide) <- c("my_triangle_colorsteps", class(guide))
  guide
}

guide_gengrob.my_triangle_colorsteps <- function(...) {
  # First draw normal colorsteps
  guide <- NextMethod()
  # Extract bar / colours
  is_bar <- grep("^bar$", guide$layout$name)
  bar <- guide$grobs[[is_bar]]
  extremes <- c(bar$gp$fill[1], bar$gp$fill[length(bar$gp$fill)])
  # Extract size
  width  <- guide$widths[guide$layout$l[is_bar]]
  height <- guide$heights[guide$layout$t[is_bar]]
  short  <- min(convertUnit(width, "cm",  valueOnly = TRUE),
                convertUnit(height, "cm", valueOnly = TRUE))
  # Make space for triangles
  guide <- gtable_add_cols(guide, unit(short, "cm"),
                           guide$layout$t[is_bar] - 1)
  guide <- gtable_add_cols(guide, unit(short, "cm"),
                           guide$layout$t[is_bar])

  left <- polygonGrob(
    x = unit(c(0, 1, 1), "npc"),
    y = unit(c(0.5, 1, 0), "npc"),
    gp = gpar(fill = extremes[1], col = NA)
  )
  right <- polygonGrob(
    x = unit(c(0, 1, 0), "npc"),
    y = unit(c(0, 0.5, 1), "npc"),
    gp = gpar(fill = extremes[2], col = NA)
  )
  # Add triangles to guide
  guide <- gtable_add_grob(
    guide, left, 
    t = guide$layout$t[is_bar],
    l = guide$layout$l[is_bar] - 1
  )
  guide <- gtable_add_grob(
    guide, right,
    t = guide$layout$t[is_bar],
    l = guide$layout$l[is_bar] + 1
  )
  
  return(guide)
}

ggplot(bias.df2) +  
  geom_tile(aes(x=x, y=y, fill=pr), alpha=0.8) + 
  scale_fill_viridis_c(na.value="white", limits = c(-10, 10),
                       breaks = c(-10, -7, -4, -1, -.5, .5, 1, 4, 7, 10), 
                       guide = my_triangle_colorsteps(show.limits = TRUE)) +
  coord_equal() +
  theme(legend.position="bottom") +
  theme(legend.key.width=unit(2, "cm"))

^{Created on 2022-06-01 by the reprex package (v2.0.1)}

How to modify a continous ggplot legend in R

1 Answers1