Skip to content
Prev 398046 / 398502 Next
Original Raw

ggplot with arrows

Ebert,Timothy Aaron 
That was good in that the arrows look like part of the figure.
In this problem I have two choices. The easy one is that this is a one-off task. The easy solution (for me) is to to offload the figure to another program and add the arrows manually. They will look like I want and I will be done in 5 to 10 minutes. It might take a few minutes longer if I find that PowerPoint is not quite working out and I need a program like Photoshop. I can copy-paste, or change the output file type. The difficult problem is where this is one case in a broader application where I need to graph a function within the specified range and add arrows to the ends of the function if it continues beyond the plotted limits as expected from the plotted portion. If this is for a classroom it may use methods far beyond the abilities of the students present (programming, and mathematically).

Arrows for sin(x) or tan(x) do not make as much sense as for the x^2 function provided. The program does not catch cases where the function may be undefined or change directions outside the printed range. The arrow gives the impression that the figure will continue in the same direction for all x outside of the printed range.
Calculus gives us a way to check function performance. Something like this, though I suspect this is a start rather than a final solution:
library(ggplot2)
library(grid)
library(numDeriv)

Min_x <- -10
Max_x <- 1
my_function <- "x^2 / (1 - x)"

f_expr <- parse(text = my_function)
f <- function(x) eval(f_expr[[1]], envir = list(x = x))
detect_discontinuity <- function(f, Min_x, Max_x, buffer = 2, n = 1000, jump_tol = 100) {
  x_seq <- seq(Min_x - buffer, Max_x + buffer, length.out = n)
  y_seq <- tryCatch(f(x_seq), error = function(e) rep(NA, length(x_seq)))
  bad_vals <- !is.finite(y_seq)
  jumps <- abs(diff(y_seq))
  big_jump <- which(jumps > jump_tol)
  discont_locs <- sort(unique(c(
    x_seq[bad_vals],
    x_seq[big_jump]
  )))

  list(
    discontinuities = discont_locs,
    has_left = any(discont_locs < Min_x),
    has_right = any(discont_locs > Max_x)
  )
}
plot_with_discontinuity_check <- function(f, Min_x, Max_x) {
  check <- detect_discontinuity(f, Min_x, Max_x)

  p <- ggplot() +
    stat_function(
      fun = f,
      color = "blue", linewidth = 1.25,
      xlim = c(Min_x, Max_x)
    ) +
    theme_linedraw()
  print(p)
  grid.force()
  if (!check$has_left && !check$has_right) {
    grid.edit("GRID.polyline", grep = TRUE, arrow = arrow(ends = "both"))
  } else if (!check$has_left) {
    grid.edit("GRID.polyline", grep = TRUE, arrow = arrow(ends = "first"))
  } else if (!check$has_right) {
    grid.edit("GRID.polyline", grep = TRUE, arrow = arrow(ends = "last"))
  }
  if (check$has_left) {
    grid.text("Left: Discontinuity prevents extension",
              x = unit(0.05, "npc"), y = unit(0.95, "npc"),
              just = "left", gp = gpar(col = "black", fontsize = 8))
  }
  if (check$has_right) {
    grid.text("Right: Discontinuity prevents extension",
              x = unit(0.95, "npc"), y = unit(0.95, "npc"),
              just = "right", gp = gpar(col = "black", fontsize = 8))
  }
}
plot_with_discontinuity_check(f, Min_x, Max_x)

I suspect there are better approaches, but some form of error checking is essential to ensure that arrows are appropriate at both ends of the function as graphed. Graphs should not lie, but that is a difficult task in the generic problem. Prove that the function does not change direction or have a discontinuity outside the printed range from -infinity to +infinity. It is also possible that the graph is well behaved on one side but not the other.

Is there a better solution? Are there other ways to check for conditions that might make arrows inappropriate?
One could check and see if the graph ends at a discontinuity and add a bar at the end of the line. One might check to see if a discontinuity occurs within the figure and add a bar at the end of each line at the point of the discontinuity. I think the current graph glosses over the problem. How broadly does one define a function? Is this a function with respect to this problem: Ifelse(x>4,x^2,3x*x^0.5)?

Tim

-----Original Message-----
From: Paul Murrell <paul at stat.auckland.ac.nz>
Sent: Sunday, August 3, 2025 5:01 PM
To: Rui Barradas <ruipbarradas at sapo.pt>; Ebert,Timothy Aaron <tebert at ufl.edu>; Thomas Subia <tgs77m at yahoo.com>; r-help at r-project.org
Subject: Re: [R] ggplot with arrows

[External Email]

Hi

At the risk of being labelled a "purist", you do not have to leave R to gain access to the underlying components and modify them.

The following code combines Rui's tidy up of the ggplot() call with an example of directly modifying the underlying 'grid' grob to add arrows to the ends.  There are more arguments in arrow() that you can explore if you want to.


library(ggplot2)
library(grid)

ggplot() +
   stat_function(
     fun = function(x) x^2,
     color = "blue", linewidth = 1.25,
     xlim = c(-5, 5)
   ) +
   theme_linedraw()

grid.force()
## Check the name of the line grob
## grid.ls()
grid.edit("GRID.polyline", grep=TRUE,
           arrow=arrow(ends="both"))


Hope that helps.

Paul
On 2/08/2025 3:24 am, Rui Barradas wrote:
--
Dr Paul Murrell (he/him)
Te Kura Tatauranga | Department of Statistics Waipapa Taumata Rau | The University of Auckland Private Bag 92019, Auckland 1142, New Zealand
64 9 3737599 x85392
paul at stat.auckland.ac.nz
http://www.stat.auckland.ac.nz/~paul/

Thread (10 messages)

Thomas Subia ggplot with arrows Aug 1 Ebert,Timothy Aaron ggplot with arrows Aug 1 @vi@e@gross m@iii@g oii gm@ii@com ggplot with arrows Aug 1 Rui Barradas ggplot with arrows Aug 1 Rui Barradas ggplot with arrows Aug 1 Paul Murrell ggplot with arrows Aug 3 Ebert,Timothy Aaron ggplot with arrows Aug 3 Thomas Subia ggplot with arrows Aug 3 Martin Maechler ggplot with arrows Aug 19 Thomas Subia ggplot with arrows Aug 19