Consistent colouring in `ggplot2` when factor levels are (sometimes) missing using `scale_colour_manual` or an alternative

Question

I have a df and a colour palette cp:

df <- data.frame(x = c(1,2,3), y = c(1,1,1), col = c(1,2,3), 
        label = c("one", "two", "three"))
cp <- c("#9986A5", "#79402E", "#0F0D0E")

I can plot the labels with the corresponding colour from my cp with:

library(ggplot2)
ggplot(df, aes(x, y)) + geom_text(aes(label = label, color = as.factor(col)))
  + scale_colour_manual(values = cp)

However, when there are factor levels missing, the colouring gets inconsistent:

df$col <- c(1,1,3)
ggplot(df, aes(x, y)) + geom_text(aes(label = label, color = as.factor(col))) 
  + scale_colour_manual(values = cp)

How can I achieve a consistent colouring that is independent from missing factor levels?

Answer 1

I found the use of drop = FALSE in scale ... manual calls to be really helpful with the problem of unrepresented factor levels leading to inconsistent colour scales.

Answer 2

If you use a named vector to specify your colour palette, levels will be matched up with colours consistently:

cp <- c(`1` = "#9986A5", `2` = "#79402E", `3` = "#0F0D0E")

You have to quote the numbers to get them to work as names for the vector.