Accessible colour contrasts with {coloratio}


December 30, 2020

The word 'coloratio' in green text on a black background and vice versa to its right.

This blog’s original theme: insufficient contrast!


I made a small R package called {coloratio} to evaluate colour-contrast ratios for accessibility. Then I found out that {savonliquide} already exists to do this.

Accessible charts

The UK government’s website, GOV.UK, was developed with user needs and accessibility in mind. I’ve been using {ggplot2} to recreate the simple, accessible chart styles suggested for use on GOV.UK by the Government Statistical Service.

But I wondered: is it possible to programmatically select a high-contrast text colour to overlay the fill colours of a {ggplot2} barplot? You would want black text over white and vice versa, for example.

What is ‘high contrast’ anyway? GOV.UK’s Design System refers to W3C’s contrast guidance from WCAG 2.1, which suggests a ratio of 4.5:1 for regular text on a block-coloured background.

It isn’t a big deal to program this ‘manually’, but that’s not fun.

Ratio calculation

Is the contrast accessible?

How about a small package with some functions to derive colour contrast ratios? Introducing {coloratio}.

install.packages("remotes")  # if not yet installed

Pass two colours to cr_get_ratio() as hex values or named colours—see colors()—and it performs the necessary calculations to derive relative luminance and return a colour contrast ratio.

library(coloratio)  # attach package

  "papayawhip", "#000000",  # colours to compare
  view = TRUE  # optional demo of colours 

[1] 18.55942

This contrast value is above the 4.5 threshold, so we’re good to go. You’ll get a warning if the contrast is insufficient.

cr_get_ratio("olivedrab", "olivedrab2")
Warning in cr_get_ratio("olivedrab", "olivedrab2"): Aim for a value of 4.5 or higher.
[1] 2.755693

Surprise: as stunning as an all-olivedrab palette might be, these colours aren’t distinct enough to be accessible.

Black or white?

cr_get_ratio() in turn powers the function cr_choose_bw(), which returns black or white depending on the greatest contrast with a supplied background colour.

[1] "black"
[1] "white"

To demonstrate better, let’s create a grouped barplot with lighter (lemonchiffon3) and darker (hotpink4) fill colours, then use cr_choose_bw() to choose black or white for the overlaying text.

library(tidyverse)  # for data manipulation

# Example data
d <- data.frame(
  x_val = c("A", "A", "B", "B"),
  y_val = c(3, 6, 4, 10),
  z_val = c("a", "b", "a", "b")
) %>% 
  mutate(  # add colour columns
    fill_col = rep(c("hotpink4", "lemonchiffon3"), 2),
    text_col = map_chr(fill_col, cr_choose_bw)

d  # preview
  x_val y_val z_val      fill_col text_col
1     A     3     a      hotpink4    white
2     A     6     b lemonchiffon3    black
3     B     4     a      hotpink4    white
4     B    10     b lemonchiffon3    black

No surprise: white was returned for the darker fill and black for the lighter fill.

We can now refer to this information in the colour argument of geom_text().

ggplot(d, aes(x_val, y_val, fill = z_val)) +
  geom_bar(position = "dodge", stat = "identity") +
  scale_fill_manual(values = d$fill_col) +    # fill colour
  geom_text(aes(y = 0.5, label = y_val), 
            position = position_dodge(0.9), 
            size = 5, colour = d$text_col) +  # text colour 
  coord_flip() + 
  theme_minimal(base_size = 16) +  # clean up the theme
  theme(axis.text.x = element_blank(), axis.title = element_blank(), 
        legend.title = element_blank(), panel.grid = element_blank())

As desired: black on the lighter fill; white on the darker fill. The default would be black text, which would provide insufficient contrast for darker fills.

Aside: cr_choose_bw() in geom_text()?

Originally I wanted geom_text() to choose text colours on the fly, rather than adding them to the input data. This roundabout solution—which outputs a similar plot to the one above—requires you to build the plot object, then interrogate it with ggplot_build() to identify the bar-fill colours.

# Build simple grouped barplot again
p <- ggplot(d, aes(x_val, y_val, fill = z_val)) +
  geom_bar(position = "dodge", stat = "identity") +
  scale_fill_manual(values = c("hotpink4", "lemonchiffon3")) +

# Extract the p-object fills and choose text overlay colour
p + geom_text(
  aes(y = 0.5, label = y_val), position = position_dodge(0.9), size = 5,
  colour = map_chr(  # make text colour dependent on bar colour
    ggplot_build(p)[[1]][[1]]$fill,  # access p-object fills
    coloratio::cr_choose_bw   # choose black/white text based on fill

I put this to the RStudio Community with no answer to date. Let me know if you have any ideas.

A soapy slip-up

Having addressed my need, I was suspicious. Surely this has been done in R before?

Whoops. {savonliquide} by Ihaddaden M. EL Fodil can query the WebAIM contrast checker API to get the contrast ratio for two colours. And it’s on CRAN.


Maybe I missed it because of the name, which translates to ‘liquid soap’?

Anyway, like coloratio::cr_get_ratio(), you can pass two hex values or named colours to {savonliquide}’s check_contrast() function.

savonliquide::check_contrast("blanchedalmond", "bisque2")

* The Contrast Ratio is 1.03

* The result for the AA check is : FAIL

* The result for the AALarge check is : FAIL

* The result for the AAA check is : FAIL

* The result for the AAALarge check is : FAIL

The output is richer than coloratio::cr_get_ratio(). You can see here that the supplied colours fail additional accessibility checks from WCAG 2.1 that involve large text and more stringent contrast thresholds.

Handily, there’s also the savonliquide::check_contrast_raw() variant that returns a list with each result as an element.


So… should you wash your hands of {coloratio}?1 Well, it fills the micro-niche of an R package that doesn’t require an internet connection to fetch colour contrast ratios. But it’s probably never going to go on CRAN, so you should use {savonliquide}.

I certainly learnt a lesson about due diligence during package development. Especially because I also discovered recently that I had also somehow managed to reinvent the {badger} package with my own {badgr} package.2 Whoops again.

At worst, I got to learn more about accessibility, practice some package building, and solve my initial problem (kinda).

I also got to admire the creativity of the names in the named-colour set. ‘Papayawhip’ sounds really appealing. Or perhaps painful. Just like package development.3


Session info
Last rendered: 2024-07-14 20:28:01 BST
R version 4.4.0 (2024-04-24)
Platform: aarch64-apple-darwin20
Running under: macOS Ventura 13.2.1

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/lib/libRblas.0.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.4-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.12.0

[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: Europe/London
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] lubridate_1.9.3      forcats_1.0.0        stringr_1.5.1       
 [4] dplyr_1.1.4          purrr_1.0.2          readr_2.1.5         
 [7] tidyr_1.3.1          tibble_3.2.1         ggplot2_3.5.1       
[10] tidyverse_2.0.0      coloratio_0.0.0.9004

loaded via a namespace (and not attached):
 [1] gtable_0.3.5       jsonlite_1.8.8     crayon_1.5.3.9000  compiler_4.4.0    
 [5] tidyselect_1.2.1   scales_1.3.0       yaml_2.3.8         fastmap_1.2.0     
 [9] R6_2.5.1           labeling_0.4.3     generics_0.1.3     curl_5.2.1        
[13] knitr_1.46         htmlwidgets_1.6.4  munsell_0.5.1      pillar_1.9.0      
[17] tzdb_0.4.0         rlang_1.1.4.9000   utf8_1.2.4         savonliquide_0.2.0
[21] stringi_1.8.3      xfun_0.43          timechange_0.3.0   cli_3.6.3.9000    
[25] withr_3.0.0        magrittr_2.0.3     digest_0.6.35      grid_4.4.0        
[29] rstudioapi_0.16.0  hms_1.1.3          lifecycle_1.0.4    vctrs_0.6.5       
[33] evaluate_0.23      glue_1.7.0         farver_2.1.1       fansi_1.0.6       
[37] colorspace_2.1-0   httr_1.4.7         rmarkdown_2.26     tools_4.4.0       
[41] pkgconfig_2.0.3    htmltools_0.5.8.1 


  1. I assure you this is an excellent savon liquide pun.↩︎

  2. {badger} has functions to generate a bunch of badges you’re likely to want. {badgr} focuses only on custom badges and has some extra options relative to badger::badge_custom(), like the ability to add an icon. But wow, how did I miss this?↩︎

  3. #deep↩︎