Feature request: Vary colour/fill along ridgeline

[roaldarbol]

Hey Claus - thanks for some amazing work with ggridges!

I'm using geom_ridgeline, and would like to be able to colour the line by a variable (e.g. the same as height). Although it's possible to colour the fill/colour using geom_density_ridges_gradient, it doesn't seem to be possible in other cases, e.g. for a time series.

Here's a minimal example of what I'd like to be able to do:

lincoln_weather %>% 
  mutate(
         date = as_date(CST),
         day_n = mday(date)) %>% 
  filter(as_date(date) < as_date("2017-01-01")) %>% 
ggplot(aes(x = day_n, y = Month, height = `Mean Temperature [F]`, fill = `Mean Temperature [F]`)) +
  geom_ridgeline(scale = 0.05)

[clauswilke]

I'm not sure what exactly you're asking. There is geom_ridgeline_gradient(). Does that do what you want, or are you asking for something different?

[roaldarbol]

It does almost - works for fill, but not for colour. I was using colour myself, and just assumed they would both be implemented in a similar way. So I guess the request is really about supporting varying colour. Currently, it raises the error:

Caused by error in `draw_group()`:
! These aesthetics can not vary along a ridgeline: color, size, linetype

Then in addition, it might be worth changing the error message of geom_ridgeline. The current error suggests that it's universally impossible to vary aesthetics along ridgelines (! Aesthetics can not vary along a ridgeline). Might be better to point people in the direction, that you should instead use geom_ridgeline_gradient, e.g.

Caused by error in `draw_group()`:
! Aesthetics can not vary along `geom_ridgeline` - use `geom_ridgeline_gradient` instead

[clauswilke]

The limitation of which aesthetics can vary come from the R graphics device. There's little I can do. And I don't want to encourage people to use the gradient versions of the geoms because I'm uncomfortable with how the effect is achieved. It's a bad workaround.

If at some point the R graphics device can vary color and/or size along a line then we can consider this feature.

[roaldarbol]

That's of course up to you. I think it would be useful now that it is implemented anyways.

I'm not sure about whether it's not possible to vary colour along a line. This example works just fine:

lincoln_weather %>% 
  mutate(date = as_date(CST),
         day_n = mday(date)) %>% 
  filter(as_date(date) < as_date("2017-01-01")) %>% 
ggplot(aes(x = day_n, y = `Mean Temperature [F]`, colour = `Mean Temperature [F]`, group = Month)) +
  geom_line()

Is there an underlying limitation you're thinking of?

EDIT: Ah, because the Y axis is essentially categorical in the ridgelines? But I guess if height is implemented, it should still be possible in theory?

[clauswilke]

The code in your latest example doesn't run for me. Please use the reprex package to provide examples: https://reprex.tidyverse.org/articles/magic-reprex.html

[roaldarbol]

Ah, sorry, here you go:

library(ggridges)
library(lubridate)
#> Loading required package: timechange
#> 
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union
library(tidyverse)
lincoln_weather %>% 
  mutate(date = as_date(CST),
         day_n = mday(date)) %>% 
  filter(as_date(date) < as_date("2017-01-01")) %>% 
ggplot(aes(x = day_n, y = `Mean Temperature [F]`, colour = `Mean Temperature [F]`, group = Month)) +
  geom_line()

^{Created on 2023-01-09 with reprex v2.0.2}

[clauswilke]

The thing is this may look fine if you're not looking carefully, but you're not actually getting continuous lines, you're getting individual line segments. This is both ugly and causes a headache for something like ggridges. I can't use the same rendering code I currently have and just change line color along the way.

Increasing the line thickness demonstrates this clearly.

library(tidyverse)
library(ggridges)
library(lubridate)

lincoln_weather %>% 
  mutate(date = as_date(CST),
         day_n = mday(date)) %>% 
  filter(as_date(date) < as_date("2017-01-01")) %>% 
  ggplot(aes(x = day_n, y = `Mean Temperature [F]`, colour = `Mean Temperature [F]`, group = Month)) +
  geom_line(linewidth = 5)

^{Created on 2023-01-09 with reprex v2.0.2}

[roaldarbol]

Oh boy, I didn't know that! That's unfortunate, I guess I'll just have to find a way around it for now then... Do you know if changing this behaviour is on the core R teams radar? Or the Posit team?