quantile vignette example does not work with larger numbers of quantiles

[nickreich]

I tried to reproduce the quantile vignette example but with 10 quantiles instead of 4. The resulting color scale is mis-ordered and mis-labeled.

library(ggplot2)
library(ggridges)
library(viridis)
#> Loading required package: viridisLite
ggplot(iris, aes(x=Sepal.Length, y=Species, fill = factor(stat(quantile)))) +
  stat_density_ridges(
    geom = "density_ridges_gradient", calc_ecdf = TRUE,
    quantiles = 10, quantile_lines = TRUE
  ) +
  scale_fill_viridis_d(name = "Quantiles")
#> Picking joint bandwidth of 0.181

^{Created on 2020-12-28 by the reprex package (v0.3.0)}

[nickreich]

For what it's worth, it does work with 9 quantiles (and lower numbers as well):

library(ggplot2)
library(ggridges)
library(viridis)
#> Loading required package: viridisLite
ggplot(iris, aes(x=Sepal.Length, y=Species, fill = factor(stat(quantile)))) +
  stat_density_ridges(
    geom = "density_ridges_gradient", calc_ecdf = TRUE,
    quantiles = 9, quantile_lines = TRUE
  ) +
  scale_fill_viridis_d(name = "Quantiles")
#> Picking joint bandwidth of 0.181

^{Created on 2020-12-28 by the reprex package (v0.3.0)}

[clauswilke]

I suspect the problem is this line:

ggridges/R/stats.R

Line 290 in ec65844

df_final$quantile <- factor(df_final$quantile)

We're converting the quantiles into a factor separately for each group, and that will go wrong if some groups don't contain all levels. Solutions would be to explicitly set the levels or to do the factor conversion on the entire resulting data frame, not by group.

[clauswilke]

To do the factor conversion on the entire data frame, we could follow the template of StatDensity2d:
https://github.com/tidyverse/ggplot2/blob/feaa4bc341d1df8663ed478b448754b3bca3a375/R/stat-density-2d.r#L127-L132