Dashed segments cannot align visible endpoints because dash phase is fixed at start

[Hide-Fun]

Thanks to the ggplot2 team and community for the great work on this package.

When I draw multiple dashed segments with geom_segment() in ggplot2, segments with different start positions but the same end position have visibly misaligned dash endpoints. It seems the dash pattern is anchored at each segment's start, so the visible ends don't line up even when the end coordinates are identical.

library(ggplot2)
packageVersion("ggplot2")
#> [1] '4.0.0'

df <- data.frame(
  x = c(1, 2, 3),
  x_end = c(4, 4, 4),
  y = c(1, 2, 3)
)

ggplot() +
  geom_segment(
    data = df,
    aes(x = x, y = y , xend = x_end, yend = y),
    linetype = "dashed",
    colour = "steelblue"
  )

^{Created on 2025-11-25 with reprex v2.1.1}

[smouksassi]

if you use geom_path you can control the direction and where the line "starts" you will still have it not end with the same pattern given the different length

library(ggplot2)

df <- data.frame(
  x = c(1, 2, 3),
  x_end = c(4, 4, 4),
  y = c(1, 2, 3)
)
df2 <- data.frame(
  x = c(4,3,2,1,
        4,3,2,
        4,3),
  y = c(1,1,1,1,
        2,2,2,
        3,3)
)

ggplot() +
  geom_segment(
    data = df,
    aes(x = x, y = y , xend = x_end, yend = y),
    linetype = "dashed",
    linewidth = 2,
    colour = "steelblue"
  )+
  geom_path(
    data = df2,
    aes(x = x, y = y-0.1,group=y),
    linetype = "dashed",
    linewidth = 2,
    colour = "red")

^{Created on 2025-11-24 with reprex v2.1.1}

[teunbrand]

In a similar vein, you can just swap the xend with x and yend with y and the start and end will be swapped.

library(ggplot2)

df <- data.frame(
  x = c(1, 2, 3),
  x_end = c(4, 4, 4),
  y = c(1, 2, 3)
)

ggplot() +
  geom_segment(
    data = df,
    aes(x = x_end, y = y , xend = x, yend = y),
    linetype = "dashed",
    colour = "steelblue"
  )

^{Created on 2025-11-25 with reprex v2.1.1}

The intricacies of how linetypes are rendered isn't under ggplot2's control, but is in the grid package's domain. For that reason, I'm positing that there isn't a meaningful change to be implemented and this issue can be closed.