Feature Request: Allow reversing discrete scales

[billdenney]

In many scenarios across aesthetics (x, y, fill, colour, and probably anything discrete), it would help to have the option to reverse factor levels on a discrete axis.

What would you think of a PR for one of the options below?

However this is implemented, it would also propagate to a change in the scales package. I think that the simplest change would be to make a scales::reverse_trans_discrete() function, but I've not fully dived in to confirm that.

I think that the simplest and least invasive change would be to add a reverse argument to all the scale_*_discrete() and similar categorical functions (e.g. scale_colour_hue() and scale_colour_viridis_d()). With reverse=TRUE, scales::reverse_trans_discrete() would be used.

library(ggplot2)
d <-
  data.frame(
    x_fac=factor(LETTERS[1:3]),
    y=1:3
  )
ggplot(d, aes(x=x_fac, y=y)) +
  geom_point()

ggplot(d, aes(x=x_fac, y=y)) +
  geom_point() +
  scale_x_reverse()
#> Warning in Ops.factor(x): '-' not meaningful for factors
#> Warning: Transformation introduced infinite values in continuous x-axis
#> Error: Discrete value supplied to continuous scale
# Thanks @jennybc! (from https://gist.github.com/jennybc/6f3fa527b915b920fdd5)
ggplot(d, aes(x=x_fac, y=y)) +
  geom_point() +
  scale_x_discrete(limits=rev(levels(d$x_fac)))

# Implementation option: scale_x_discrete and similar gains a `reverse`
# argument which defaults to `FALSE`
ggplot(d, aes(x=x_fac, y=y)) +
  geom_point() +
  scale_x_discrete(reverse=TRUE)
#> Error in discrete_scale(c("x", "xmin", "xmax", "xend"), "position_d", : unused argument (reverse = TRUE)

Created on 2019-02-03 by the reprex package (v0.2.0).

[clauswilke]

Maybe one option would be to allow functions as limit arguments, just like we do for labels?

library(ggplot2)

ggplot(iris, aes(Species, Sepal.Length)) + geom_boxplot() +
  scale_x_discrete()

# works, reverses labels (but not data)
ggplot(iris, aes(Species, Sepal.Length)) + geom_boxplot() +
  scale_x_discrete(labels = rev)

# doesn't work
ggplot(iris, aes(Species, Sepal.Length)) + geom_boxplot() +
  scale_x_discrete(limits = rev)
#> Error in match(as.character(x), limits): 'match' requires vector arguments

^{Created on 2019-02-03 by the reprex package (v0.2.1)}

[clauswilke]

There's an old (abandoned?) PR about a similar topic for continuous scales: #2334

[billdenney]

The PR seems like it would be a good fit if applied to discrete scales (and it would help to add that to the documentation with a reversing example, too).

[hadley]

Note that the colour scales have a direction argument which this could replace.

[paleolimbot]

Having discrete position scales support functions as the limits argument seems like a reasonable solution to this. All other scales allow this (discrete and continuous). Then you could do limits = rev:

library(ggplot2)

ScaleDiscretePositionFunc <- ggproto(
  "ScaleDiscretePositionReversed", ScaleDiscretePosition,
  get_limits = function(self) {
    if (self$is_empty()) {
      c(0, 1)
    } else if (is.null(self$limits)) {
      self$range$range
    } else if (is.function(self$limits)) {
      self$limits(self$range$range)
    } else {
      integer(0)
    }
  }
)

scale_x_discrete2 <- function(..., expand = waiver(), position = "bottom") {
  sc <- discrete_scale(c("x", "xmin", "xmax", "xend"), "position_d", identity, ...,
                       expand = expand, guide = "none", position = position, super = ScaleDiscretePositionFunc)
  
  sc$range_c <- ggplot2:::continuous_range()
  sc
}

ggplot(mpg, aes(class, cty)) +
  geom_boxplot() +
  scale_x_discrete2(limits = rev)

^{Created on 2019-06-25 by the reprex package (v0.2.1)}

The code that needs to be updated is here:

ggplot2/R/scale-discrete-.r

Lines 81 to 85 in e2bdf85

	get_limits = function(self) {
	if (self$is_empty()) return(c(0, 1))
	self$limits %||% self$range$range %||% integer()
	},

[billdenney]

@paleolimbot, That seems like a great solution that would allow simple generalization across this and other solutions.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/