The shades package allows colours to be manipulated easily in R. Properties such as brightness and saturation can be quickly queried, changed or varied, and perceptually uniform colour gradients can be constructed. It plays nicely with the pipe operator from the popular magrittr package, and fits naturally into that paradigm. It can also be used with ggplot2 scales.
The package is available on CRAN. You can also install the current development version from GitHub using devtools:
Feedback on the package or suggestions are welcome, either by filing an issue or by email.
Colours are represented in R using CSS-style hex strings, but there is also a dictionary of predefined named colours such as "red" and "blue". Either of these may be passed to most graphics functions, but creating variations on a particular colour can be awkward.
The shades package defines a simple class, shade, which uses exactly this same convention and is entirely compatible with built-in colours, but it also stores information about the coordinates of the colours in a particular colour space.
library(shades)
red <- shade("red")
print(unclass(red))
## [1] "red"
## attr(,"space")
## [1] "sRGB"
## attr(,"coords")
## R G B
## [1,] 1 0 0From here, the package switches between colour spaces as required, allowing various kinds of colour manipulation to be performed straightforwardly. For example, letâs find the saturation level of a few built-in colours.
Now letâs consider a colour gradient stepping through two different colour spaces, which we might want to use as a palette or colour scale.
Here, we are using the swatch function to visualise a set of colours as a series of squares. Notice the more uniform appearance of the gradient when it traverses through the Lab colour space.
Similarly, we can create a set of new colours by changing the brightness and saturation levels of some base colours, and make the code more readable by using the magrittr pipe operator.
library(shades); library(magrittr)
c("red","blue") %>% brightness(0.6) %>% saturation(seq(0,1,0.25)) %>% swatch
This operation takes the original two colours, reduces their brightness to 60%, assigns a whole series of saturation levels to the result, and then passes it to swatch for visualisation. Notice that the pipeline is combinative (like the base function outer), returning each combination of parameters in a multidimensional array. The final shades are arranged in two rows by swatch, for convenience.
Note that NA can be used as a pass-through value:
Any of these gradients can be directly passed to a standard graphical function, to be used as a colour scale. However, when choosing a colour scale, it is helpful to bear in mind that some viewers may have a colour vision deficiency (colour blindness), making it harder for them to distinguish certain colours and therefore to see a continuous scale. The dichromat function can be used to simulate this.
Here we are using the built-in âviridisâ colour map, developed for Pythonâs matplotlib, which was specifically designed to appear continuous under as many conditions as possible. When shown with simulated red-blindness, the default for dichromat, it is clearly much more interpretable than a typical rainbow palette generated by Râs built-in graphics functions.
The package also supports colour mixing, either additively (as with light) or subtractively (as with paint). For example, consider additive mixtures of the three primary RGB colours.
Similarly, we can subtractively combine the three secondary colours.
A âlight mixtureâ infix operator, %.)%, and a âpaint mixtureâ infix operator, %_/%, are also available.
Finally, you can calculate perceptual distances to a reference colour, as in
The shades package can be used with the popular ggplot2 graphics library in different ways, with different levels of integration. Firstly, gradients from this package can be used as ggplot2 colour scales through the manual scale functions; for example,
library(shades); library(ggplot2)
mtcars$cyl<- factor(mtcars$cyl)
ggplot(mtcars, aes(cyl,mpg,fill=cyl)) + geom_boxplot() + scale_fill_manual(values=gradient("viridis",3))
This does not require the two packages to know anything about each other, and is flexible and powerful, but it doesnât easily allow existing ggplot2 scales to be modified using the colour manipulation functions from shades. As of shades version 1.3.0, it is also possible to call the packageâs colour property functions directly on palette functions and scales, so that (for example), we can darken all colours in an existing scale slightly:
Notice here that we have chosen to use the delta() function, which is available in all colour property functions, to request a relative reduction of 20 to the original lightness of each colour in the scale. We could also have given a literal value to fix the lightness of all colours to a certain level.
The shades package aims to bring together a range of colour manipulation tools and make them easy to use. However, there are several other packages available that can do similar things, sometimes in slightly different ways. These include
grDevices package, which is shipped with R and used as the basis for shades;colorspace package, which provides formal colour classes and transformations between spaces;munsell, which interprets colours in Munsell notation and does some colour manipulation;viridis and RColorBrewer, which provide the colour scales from matplotlib and ColorBrewer;dichromat, which provides another implementation of the dichromat function (a duplication which I didnât discover until after writing this packageâs version!); andcolorblindr, which provides alternative tools for simulating colour blindness in figures.This package was also partly influenced by Colors.jl, a colour manipulation package for Julia.