support sec.axis for scale_discrete

[maxheld83]

scale_y_discrete() and scale_x_discrete() currently appear not to support the sec.axis argument of their scale_continuous() brethren.

sec.axis would be nice for discrete scales as well, such as when you have various ways to refer to factor levels (short and long) and the like.

[thiagoveloso]

I second this request, and for the same reasons as @maxheld83 mentioned. This would be a really nice feature, which in the current ggplot2 version cannot be achieved even with hacks.

[topialla]

Why was this closed? There still seems to be only this workaround available:
https://stackoverflow.com/a/45362497/1842673
Could this be implemented?

[clauswilke]

The issue is still open, as far as I can tell. If somebody would like to contribute a PR, we'd certainly review it.

It didn't make sense to tackle this feature before @paleolimbot had completed his rewrite of position scales, but since that has happened and is merged I'd assume implementing discrete secondary axes shouldn't be too hard.

[paleolimbot]

Famous last words!

I think a guide extension is the way to go about this (actually, I think regular second axes should use this as well). It's not future proof since we're about to redo guides (post patch release), but you could hack this with the current system:

library(ggplot2)

guide_axis_label_trans <- function(label_trans = identity, ...) {
  axis_guide <- guide_axis(...)
  axis_guide$label_trans <- rlang::as_function(label_trans)
  class(axis_guide) <- c("guide_axis_trans", class(axis_guide))
  axis_guide
}

guide_train.guide_axis_trans <- function(x, ...) {
  trained <- NextMethod()
  trained$key$.label <- x$label_trans(trained$key$.label)
  trained
}

ggplot(mpg, aes(class, hwy)) +
  geom_boxplot() +
  guides(x.sec = guide_axis_label_trans(~paste("Ha!", .x)))

^{Created on 2020-04-15 by the reprex package (v0.3.0)}

[rsaporta]

was searching for something similar and found the SO thread linking to here. +1 this useful feature request. Thanks @paleolimbot for the hack

[EnricGTorrents]

I also came here from SO looking for a way to duplicate the discrete y scale. +1 this would be a nice feature to see in future versions. Thanks!

[matteodefelice]

I think it would be nice to have this feature.

I would like to use sec.axis for scale_y_discrete for plotting a second geom: geom_density_ridges. The y-axis for the geom_histogram plot is continuous, while the desired y-axis for the geom_density_ridges plot is discrete.

Example:

p1 <- ggplot()
# scale for geom_histogram 
p1 <- p1 + geom_histogram(data=df, aes(x=xVar, y=..density.., fill=..x.., color=..x..), alpha=0.4, bins=25, position = "identity")
p1 <- p1 + geom_rug(data=df, aes(x=xVar, color=..x..), alpha = 0.15, position = "identity")
p1 <- p1 + scale_fill_distiller(palette="YlGnBu", direction=1)
p1 <- p1 + scale_color_distiller(palette="YlGnBu", direction=1)
# new scale for geom_density_ridges
p1 <- p1 + new_scale_color()
p1 <- p1 + geom_density_ridges(data=df, aes(x=xVar, y=..density.., group=factorVar), stat = "identity")
p1 <- p1 + scale_color_brewer(palette="YlGnBu")
# what should I do below?
p1 <- p1 + scale_y_discrete(sec.axis = sec_axis())
p1 %>% print()

Take care and many thanks

[saul-torres]

I have just arrived here from SO.
Please, it would be great to have this option.

[marton-balazs-kovacs]

I would love this!

[smped]

Yes, also a great idea from my perspective!

[rocanja]

I also support this feature request ... even if the xaxis label could just be duplicated without any transformation ... this would be very useful for large / long plots to have the same label on the top and bottom (or left and right, respectively).