geom_label locations affected by "fill" argument

[emjonaitis]

I found a problem with the location of labels created by geom_label in the context of polar coordinates.

When I do not use the "fill" argument in geom_label, the labels are located correctly. However, when I set "fill" to a color, the labels appear to swap positions, as though theta had been set to minus theta. I expected fill color and location to be able to be changed independently.

Here is the code to reproduce the bug:

penguin_sum <- penguins %>%
  group_by(species) %>%
  summarize(y=n(), .groups="drop") %>%
  merge(summarize(., N=sum(y))) %>%
  mutate(pct=round(100*y/N),
         label=paste0("N=", y, "\n(", pct, "%)"))

penguin_sum
# Counts:
# Adelie 152
# Chinstrap 68
# Gentoo 124

# Labels display on correct segment, with Adelie N=152
ggplot(penguin_sum, aes(x="", y=y, fill=species)) +
  geom_col(color="white") +
  geom_label(aes(label=label), position=position_stack(vjust=0.5)) +
  coord_polar(theta="y") 

# Labels print in mirror, so it looks like Adelie N=124
ggplot(penguin_sum, aes(x="", y=y, fill=species)) +
  geom_col(color="white") +
  geom_label(aes(label=label), fill="white", position=position_stack(vjust=0.5)) +
  coord_polar(theta="y")

reprex::reprex()

[teunbrand]

Thanks for the report! I believe this is because fill = "white" overrules the role that fill = species plays in automatically detecting groups. If you set aes(label = label, group = species), the stacking should work the same as in your correct plot.

[EvaMaeRey]

I don't think this is a bug, but is kind of hard to reason about.

Using layer_data() you can see the fill = "white" dropping the group info.

library(palmerpenguins)
library(tidyverse)
#> Warning: package 'ggplot2' was built under R version 4.4.1
penguin_sum <- penguins %>%
  group_by(species) %>%
  summarize(y=n(), .groups="drop") %>%
  merge(summarize(., N=sum(y))) %>%
  mutate(pct=round(100*y/N),
         label=paste0("N=", y, "\n(", pct, "%)"))

penguin_sum
#>     species   y   N pct        label
#> 1    Adelie 152 344  44 N=152\n(44%)
#> 2 Chinstrap  68 344  20  N=68\n(20%)
#> 3    Gentoo 124 344  36 N=124\n(36%)
# Counts:
# Adelie 152
# Chinstrap 68
# Gentoo 124

# Labels display on correct segment, with Adelie N=152
ggplot(penguin_sum, 
       aes(x="", y=y, fill=species)) +
  geom_col(color="white") +
  geom_label(aes(label=label), 
             position = position_stack(vjust=0.5))# +
            # coord_polar(theta="y") 

 

layer_data(i = 2)
#>          label x   y    fill PANEL group ymax xmin xmax ymin colour family
#> 1 N=152\n(44%) 1 268 #F8766D     1     1  344    1    1  192  black       
#> 2  N=68\n(20%) 1 158 #00BA38     1     2  192    1    1  124  black       
#> 3 N=124\n(36%) 1  62 #619CFF     1     3  124    1    1    0  black       
#>       size angle hjust vjust alpha fontface lineheight linewidth linetype
#> 1 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1
#> 2 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1
#> 3 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1

Groups for geom_label are 1,2,3

# Labels print in mirror, so it looks like Adelie N=124
ggplot(penguin_sum, aes(x="", y=y, fill=species)) +
  geom_col(color="white") +
  geom_label(aes(label=label), fill="white", position=position_stack(vjust=0.5))# +
  # coord_polar(theta="y")

layer_data(i = 2)
#>          label x   y PANEL group ymax xmin xmax ymin colour  fill family
#> 1 N=152\n(44%) 1  76     1     1  152    1    1    0  black white       
#> 2  N=68\n(20%) 1 186     1     1  220    1    1  152  black white       
#> 3 N=124\n(36%) 1 282     1     1  344    1    1  220  black white       
#>       size angle hjust vjust alpha fontface lineheight linewidth linetype
#> 1 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1
#> 2 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1
#> 3 3.866058     0   0.5   0.5    NA        1        1.2      0.25        1

Groups for geom_label, above are 1,1,1, so the species info is lost for the layer, w fill = "white".

So you can add it back.

last_plot() +
  aes(group = species)

And back to coord_polar()

last_plot() + 
  coord_polar(theta="y")

^{Created on 2025-12-01 with reprex v2.1.1}

[teunbrand]

I don't think this is a bug

I agree with this: it is working as intended and requires no adaptation in ggplot2.