`geom_spoke` uses the wrong angle if radius is `Inf` in some cases

[eliocamp]

Simple segment at a 45° angle:

library(ggplot2)

ggplot() +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = 10),
    aes(0, 0, angle = angle, radius = radius),
  )  +
  annotate("point", x = 2, y = 2)

The angle is correct (passes through the x = y point)

I can use radius = Inf to extend it to the limits of the plots and still works.

ggplot() +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = Inf),
    aes(0, 0, angle = angle, radius = radius),
  )  +
  annotate("point", x = 2, y = 2)

However, when there is other data, infinite radius is draw with the wrong angle

data.frame(x = rnorm(100, sd = 2)) |> 
  transform(y = x^2 + rnorm(100)) |> 
  ggplot(aes(x, y)) +
  geom_point() +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = Inf),
    aes(0, 0, angle = angle, radius = radius),
  ) +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = 10),
    color = "red",
    aes(0, 0, angle = angle, radius = radius),
  )  +
  annotate("point", x = 10, y = 10, color = "red")

It seems to be pointing to the upper right corner of the plot

data.frame(x = rnorm(100, sd = 2)) |> 
  transform(y = x^2 + rnorm(100)) |> 
  ggplot(aes(x, y)) +
  geom_point() +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = Inf),
    aes(0, 0, angle = angle, radius = radius),
  ) +
  geom_spoke(
    data = data.frame(angle = pi / 4, radius = 10),
    color = "red",
    aes(0, 0, angle = angle, radius = radius),
  )  +
  annotate("point", x = 10, y = 10, color = "red") +
  lims(y = c(NA, 30))

[teunbrand]

Thanks for the report Elio! This is because x + cos(pi / 4) * Inf evaluates to Inf, and so does y + sin(pi / 4) * Inf. The spoke's end is then placed at c(Inf, Inf), which gets squished to the limit, placing it in the upper right corner. This placement won't change because of data. I'm not sure ggplot2 should promise to handle edgecases like infinite radii.

[eliocamp]

ggplot2 does handle infinite xmin and xmax in geom_rect, segment and friends, so I assumed that it would also work in spoke. If it doesn't work, maybe it should drop it with a warning?

[teunbrand]

It doesn't handle infinites well, it just works out if the lines are perfectly horizontal or vertical.
In the plot below, you'd expect the line to never reach y = 4, because the second x is infinity away.
Yet, it does so within the range of the plot anyway.

library(ggplot2)

df <- data.frame(
  x = c(0, Inf),
  y = c(0, 4)
)

ggplot(df, aes(x, y)) +
  geom_path()

^{Created on 2025-10-02 with reprex v2.1.1}

Throwing a warning about infinites sounds reasonable to me.