Show proper error when both x and y are specified on stat_density() and stat_ecdf()

[yutannihilation]

Fix #4776

[thomasp85]

We are now doing unit tests for all errors so can I get you to add tests for this

[yutannihilation]

Sure

[yutannihilation]

It seems this existing code will be broken by this change. I'm still thinking, but this might mean the same case can be found with stat_bin() and stat_count()...?

library(ggplot2)

do_plot <- function() {
  ggplot(mtcars, aes(mpg, factor(cyl))) +
    stat_density(aes(fill = after_stat(density)), geom = "raster", position = "identity")
}

do_plot()

devtools::load_all("~/GitHub/ggplot2/")
#> ℹ Loading ggplot2
do_plot()
#> Error in `stat_density(aes(fill = after_stat(density)), geom = "raster", position =
#> "identity")`:
#> ! Problem while computing stat.
#> ℹ Error occurred in the 1st layer.
#> Caused by error in `setup_params()` at ggplot2/R/ggproto.r:182:16:
#> ! `stat_density()` must only have an x or y aesthetic.

^{Created on 2022-07-27 by the reprex package (v2.0.1)}

[clauswilke]

@yutannihilation The example you found that breaks is an important use case I think. Fundamentally this highlights the core problem, which is that we don't explicitly distinguish between aesthetics given to the stat and to the geom. We may have to accept the confusing error message.

[yutannihilation]

Thanks. Yes, I'm almost giving up this pull request. Rather, my current concern is if stat_bin() and stat_count() might be also wrong. The below example won't work even if I remove "must only have an x or y" check, but I don't figure out if this should work like the example of stat_density() above or not...

ggplot(mtcars, aes(mpg, factor(cyl))) +
  stat_bin(aes(fill = after_stat(count)), geom = "raster", position = "identity")

[yutannihilation]

As this pull request seems impossible, I'm giving up this. Sorry that I didn't come up with the usage.