scale_colour_manual does not drop unused levels.

[paulMT]

Dear all,

Before aking my question, I went through all the web discussions about scale_colour_manual(). My impression is that the problem is related to the new version of ggplot (version 3.5.5) as before it worked for me.

As far as I understand from the help function (and from the discussions on the web), with scale_colour_manual() the unused factor levels are omitted from the legend.

This does not work (any more) for me, even if I set drop = TRUE as illustrated in the example below. I am using a recent version of ggplot2 (3.3.5).

In this example the legend still contains a key for "Y" (red) although the factor (the variable Extra) in the dataset only contains "N".

library(ggplot2); library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
theme_set(theme_bw(base_size = 15))

## a local copy of the BOD dataset
## time points larger than five days are labelled as extra
tib_BOD <- BOD %>%
  as_tibble() %>%
  mutate(Extra = factor(ifelse(Time > 5, "Y", "N")))
tib_BOD
#> # A tibble: 6 x 3
#>    Time demand Extra
#>   <dbl>  <dbl> <fct>
#> 1     1    8.3 N    
#> 2     2   10.3 N    
#> 3     3   19   N    
#> 4     4   16   N    
#> 5     5   15.6 N    
#> 6     7   19.8 Y

## a subset without Extra == "Y"
tib_BOD_noExtra <- tib_BOD %>% filter(Extra == "N")

## named vector to define the colours
fcol_Extra <- c(N = "blue", Y = "red")

## the legend contains the "Y" although it is not used
## drop = TRUE is not required as this is the default
ggplot(tib_BOD_noExtra) +
  geom_point(aes(x = Time, y = demand, colour = Extra), size = 5) +
  scale_colour_manual(values = fcol_Extra, drop = TRUE) 

^{Created on 2021-07-03 by the reprex package (v2.0.0)}

[DavoWW]

Hi @paulMT,
Thanks for posting a great reproducible example.
You had a small error in your character vector of colours; now it works.

suppressPackageStartupMessages(library(tidyverse))

tib_BOD <- read.table(header=TRUE, text ="
Time demand Extra
1    8.3 N    
2   10.3 N    
3   19   N    
4   16   N    
5   15.6 N    
7   19.8 Y
")

tib_BOD
#>   Time demand Extra
#> 1    1    8.3     N
#> 2    2   10.3     N
#> 3    3   19.0     N
#> 4    4   16.0     N
#> 5    5   15.6     N
#> 6    7   19.8     Y

## a subset without Extra == "Y"
tib_BOD_noExtra <- tib_BOD %>% filter(Extra == "N")

## named vector to define the colours
#fcol_Extra <- c(N = "blue", Y = "red")  # This line was incorrect

fcol_Extra <- c("blue","red")

ggplot(tib_BOD_noExtra) +
  geom_point(aes(x = Time, y = demand, colour = Extra), size = 5) +
  scale_colour_manual(values = fcol_Extra) 

ggplot(tib_BOD) +
  geom_point(aes(x = Time, y = demand, colour = Extra), size = 5) +
  scale_colour_manual(values = fcol_Extra) 

^{Created on 2021-07-04 by the reprex package (v2.0.0)}

[paulMT]

Thank you for your reply. It solves the problem in my small example, but not generally. I really need to work with named colour vectors as recommended in one of the help examples of scale_colour_manual(). If I run the (second) help example with the named vector (cols), I get 4 keys in the legend (ggplot2 version 3.3.5), while a colleague got 3 keys only with ggplot2 version 3.3.3. There, the orange key is dropped, as it is not used in the graph. So, as I far as I can control, something has changed in comparison to previous versions. All of the sudden a lot of code does not run any more in a proper way.

[smouksassi]

values vector in scale_color_manual

[paulMT]

Thank you very much, this solves my problem.
I will do some further research, but I think that something similar holds for all scale__manual.() functions.