transform = "reverse" does not work in scale_size_area() but does in every other size scaling function

[Lucielclr]

The transform = "reverse" argument does not work for scale_size_area() specifically.

Examples :

1. scale_size()
   p <- ggplot(CO2, aes(Plant, conc, size = uptake)) + geom_point(shape = 21) + scale_size(transform = "reverse", max_size = 5)

2. scale_size_area()
   p <- ggplot(CO2, aes(Plant, conc, size = uptake)) + geom_point(shape = 21) + scale_size_area(transform = "reverse")

[teunbrand]

Thanks for the report! The transform is doing as one expects, but because the data is rescaled to the maximum (which becomes negative due to the transform), we get right back to positive numbers again.

scales::rescale_max(scales::transform_reverse()$transform(1:5))
#> [1] 1 2 3 4 5

^{Created on 2025-03-06 with reprex v2.1.1}

[teunbrand]

Giving this a second thought, I don't really know what the expected outcome here would be.
Combing a reverse transform with some absolute palette is essentially undoing the transform.
It'd be better to tackle this issue out the output range side, by flipping the range parameter.

library(ggplot2)

ggplot(CO2, aes(Plant, conc, size = uptake)) + 
  geom_point(shape = 21) + 
  scale_size(range = c(6, 0))

^{Created on 2025-03-18 with reprex v2.1.1}

In theory, we could intercept this particular combination of transformation/aesthetic/palette and adjust it to agree with intuition. However, I don't think it is worth making an exception here because it could throw off assumptions people have made about scales. For that reason, I propose adjusting the output range is the solution to this issue and it requires no further changes to ggplot2's code.