No ticks when the limit of Coord is in the opposite order to the Scale

[yutannihilation]

(Originally reported at #3948)

Should the p1 below be errored? Or, should the ticks be drawn?

library(ggplot2)
library(patchwork)

d <- data.frame(
  x = 1:4,
  y = 1:4
)

p <- ggplot(d, aes(x, y)) +
  geom_point() +
  scale_y_reverse()

p1 <- p + coord_cartesian(ylim = c(1, 10))
p2 <- p + coord_cartesian(ylim = c(10, 1))
p1 / p2

^{Created on 2020-04-19 by the reprex package (v0.3.0)}

[clauswilke]

I think axis ticks should be drawn. Note that the problem exists even without scale_y_reverse().

library(ggplot2)
library(patchwork)

d <- data.frame(
  x = 1:4,
  y = 1:4
)

p <- ggplot(d, aes(x, y)) +
  geom_point()

p1 <- p + coord_cartesian(ylim = c(1, 10))
p2 <- p + coord_cartesian(ylim = c(10, 1))
p1 / p2

^{Created on 2020-04-18 by the reprex package (v0.3.0)}

[yutannihilation]

Oh, I didn't notice it... Thanks.

[yutannihilation]

Under the hood, this seems a problem of scale. Here, scale$get_breaks(continuous_range) returns NAs if the continuous_range is in the opposite order.

ggplot2/R/scale-view.r

Line 19 in d3d47be

breaks <- scale$get_breaks(continuous_range)

library(ggplot2)

scale <- scale_x_continuous()
scale$train(c(1, 3, 5))
scale$get_breaks(c(1, 5))
#> [1] 1 2 3 4 5
scale$get_breaks(c(5, 1))
#> [1] NA NA NA NA NA

In addition, I have no idea what is the expected behaviour in the case below. Should this be errored...?

d <- base::data.frame(
  x = 1:4,
  y = 1:4
)

p <- ggplot(d, aes(x, y)) +
  geom_point()

p + scale_y_continuous(limits = c(10, 1))
#> Warning: Removed 4 rows containing missing values (geom_point).

^{Created on 2020-04-22 by the reprex package (v0.3.0)}