NA key does not show up in legend when colour limits include NA

[davidhodge931]

library(palmerpenguins)
library(tidyverse)

penguins |>
  ggplot() +
  geom_point(
    aes(x = flipper_length_mm, 
        y = body_mass_g,
        col = species)) +
  scale_colour_manual(values = c("red", "blue"), limits = c("Adelie", "Gentoo"))
#> Warning: Removed 2 rows containing missing values or values outside the scale range
#> (`geom_point()`).

penguins |>
  ggplot() +
  geom_histogram(
    aes(x = flipper_length_mm, 
        fill = species)) +
  scale_fill_manual(values = c("red", "blue"), limits = c("Adelie", "Gentoo"))
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> Warning: Removed 2 rows containing non-finite outside the scale range
#> (`stat_bin()`).

^{Created on 2024-03-06 with reprex v2.1.0}

[teunbrand]

This is to be expected. For discrete scales, limits determine the 'universe' of values and if it doesn't include NA then NA shouldn't be shown.

A bug is that NA isn't shown as a key even if included as limits:

library(palmerpenguins)
library(ggplot2)

penguins |>
  ggplot() +
  geom_point(
    aes(x = flipper_length_mm, 
        y = body_mass_g,
        col = species)) +
  scale_colour_manual(
    values = c("red", "blue"), 
    limits = c("Adelie", "Gentoo", NA)
  )
#> Warning: Removed 2 rows containing missing values or values outside the scale range
#> (`geom_point()`).

^{Created on 2024-03-07 with reprex v2.1.0}

[davidhodge931]

Ah, okay it seems sort of inconsistent to me then, as if you set the limits to be only Adelie and Gentoo, it still plots NA

[teunbrand]

Sorry 'shouldn't be shown' should have been 'shouldn't be shown in the legend'. Apologies for the confusion

[davidhodge931]

Oh okay. As a general rule, I think anything that is plotted should be in the legend. That last plot is weird the way there is no grey dot in the legend

library(palmerpenguins)
library(tidyverse)

p <- penguins |>
  ggplot() +
  geom_point(
    aes(x = flipper_length_mm, 
        y = body_mass_g,
        col = sex))

#I think there should be a legend NA value here if NA is plotted
p +
  scale_colour_manual(values = c("red", "blue"), limits = c("female", "male"))
#> Warning: Removed 2 rows containing missing values or values outside the scale range
#> (`geom_point()`).

#The NA is missing a grey dot in the legend
penguins |>
  ggplot() +
  geom_point(
    aes(x = flipper_length_mm, 
        y = body_mass_g,
        col = species)) +
  scale_colour_manual(
    values = c("red", "blue"), 
    limits = c("Adelie", "Gentoo", NA)
  )
#> Warning: Removed 2 rows containing missing values or values outside the scale range
#> (`geom_point()`).

^{Created on 2024-03-07 with reprex v2.1.0}

[teunbrand]

As a general rule, I think anything that is plotted should be in the legend

Well that is the default behaviour for scales that display a legend, but users can intervene by setting the limits or breaks of the scale.

[davidhodge931]

Is the oob default for scale_colour_manual keep or censor?

[teunbrand]

Neither: there are no bounds to be out of, unlike continuous scales. For discrete scales, the default is to honour the na.translate setting.

[davidhodge931]

Feel free to close, if you think everything is fine here

[teunbrand]

I think the NA-key not showing up when including NA in the limits is a bug, so let's keep this open for now.