Duplicated aes warning with multiple modified aesthetics

[teunbrand]

This is a simple fix for some warnings that are thrown when multiple modified aesthetics are used.
The warnings could be triggered with the reprex below. With this PR, these warnings won't be triggered.

# Reprexed before PR
library(ggplot2)
library(colorspace)

df <- data.frame(x = 1, y = 1, lab = "test")

ggplot(df, aes(x, y, label = lab)) +
  geom_label(
    aes(colour = stage(lab, after_scale = darken(colour, 0.3)),
        fill   = after_scale(lighten(colour, 0.5)))
  )
#> Warning: Duplicated aesthetics after name standardisation: NA
#> Warning: Duplicated aesthetics after name standardisation: NA

Effectively, the two chunks of code below take the names of the (modified) aesthetics twice.

ggplot2/R/geom-.r

Line 146 in c89c265

names(modified_aes) <- rename_aes(names(modifiers))

ggplot2/R/aes.r

Lines 168 to 169 in 759c63c

	rename_aes <- function(x) {
	names(x) <- standardise_aes_names(names(x))

This surprisingly works well when there is only a single modified aesthetic, but not so much with multiple ones.

The following erroneous code (duplicated colour/color):

ggplot(df, aes(x, y, label = lab)) +
  geom_label(
    aes(colour = stage(lab, after_scale = darken(colour, 0.3)),
        color  = after_scale(lighten(color, 0.5)))
  )

Used to throw these warnings:

#> Warning: Duplicated aesthetics after name standardisation: colour
#> Warning: Duplicated aesthetics after name standardisation: NA
#> Warning: Duplicated aesthetics after name standardisation: NA

And now throws the following warnings:

#> Warning: Duplicated aesthetics after name standardisation: colour
#> Warning: Duplicated aesthetics after name standardisation: colour
#> Warning: Duplicated aesthetics after name standardisation: colour

[hadley]

Could you please add a simple unit test that uses expect_snapshot_warning() to capture the warning messages?

[teunbrand]

Sure, are you OK with bumping the {testthat} requirement to >=3.1.2 for this test?

[hadley]

Yes, that's fine.

[teunbrand]

Thanks, I also bumped rlang to >=1.0.0 due to a recently added use of on_load().

[hadley]

Since I don't think you're actually doing any substitution, quote() would be more appropriate here. But I think you could also just do p <- ggplot() + ... in the expect_snapshot_warning().

[teunbrand]

Good points, I wasn't aware of this, so thanks!

[hadley]

I only see one warning in the test?

[teunbrand]

Ah indeed! The second of those warnings only occurred when ggplot_build(p) was evaluated in the console, due to automatically printing the result. I've now disabled the legend, which caused that extra warning.

[hadley]

Thanks! The last thing is to add a bullet to the top of NEWS.md. It should briefly describe the change and end with (@yourname, #issuenumber).

[hadley]

Thanks!