Compatibility with custom geoms about ggnewscale HOT 4 CLOSED

Great! This now should be fixed in the dev version you can install with remotes::install_github("eliocamp/ggnewscale@dev"). Let me know if it works :)

Thanks for the report and the example (I added it as a test).

from ggnewscale.

Thanks. Do you have a reproducible example? ggnewscale is supposed to mimic your solution, but it's possible it needs to do more.

from ggnewscale.

Thanks for the quick reply.
Here is an example which throws the error mentioned above:

library(tidyverse)
library(ggnewscale)

# Manipulate `setup_data()` of `StatYdensity` and store the object as `StatYdensity2`.
StatYdensity2 <- ggproto("StatYdensity2", StatYdensity,
                         setup_data = function (data, params) {
                           data <- plyr::ddply(data, c("x", "fill"), within,
                                               # Assign the range of `y` to the newly created column `range`.
                                               assign("range", list(range(y))))
                           data
                         })

set.seed(5)

df <- data.frame(
  x = floor(runif(100, min=1, max=5)),
  y = floor(runif(100, min=1, max=10)),
  gender = c("female", "male")[floor(runif(100, min=1, max=3))],
  fill = floor(runif(100, min=1, max=5))
)

df %>% ggplot(aes(x, y, fill = gender, group = interaction(x, gender))) + 
  # Call `geom_violin()` using the manipulated stat `StatYdensity2`.
  geom_violin(stat = "ydensity2") + 
  new_scale_fill() + 
  geom_point(aes(x, y, fill = fill), shape = 21, inherit.aes = F)

Note that the call of plyr::ddply() does not make that much sense in this example, however as pointed out earlier, this is the source of the error.

Either applying the solution of the original post or simply calling geom_violin() instead of geom_violin(stat = "ydensity2") works like a charm. plyr::ddply() just isn't happy about the renaming of the variables that takes places due to the call of new_scale_fill().

At this point I think it's best to use the solution is already described. I just thought it's not that unusual to use plyr::ddply() in geoms at some point in general.

Looking forward to hearing about your thoughts on this.

from ggnewscale.

Great, thank you very much.
I am quite busy at the moment, but I will give it a try as soon as can in the next days!

Thanks again and all the best!

from ggnewscale.