Skip to content

labeller

Mara Averick edited this page Nov 2, 2018 · 12 revisions

Note: The ggplot2 wiki is no longer maintained, please use the ggplot2 website instead!

NOTE: the labeller API has changed. Labellers taking variableand value arguments, like the examples on this page, are now deprecated. See Generic labeller function for more information.


facet_grid has an option, labeller, which is not well documented as to what it does and how it is to be used. This page is meant to try and document it. facet_wrap may gain this option as well. This documentation is descriptive, not proscriptive. It is based on the structure of the source code and functions and is not guaranteed to work or be correct in future versions.

Purpose

The purpose of the labeller argument is to give a way to transform the levels of a factor which a plot is being faceted on before they appear on the plot itself.

Built in acceptable values

  • label_value: The default labeller. This labeller uses the direct character representation of the levels of the factor as what appears in the facet label.
  • label_both: This labeller prepends the name of the variable being faceted on to the factor level, separated by a colon and a space.
  • label_parsed: This labeller takes the character representation of the facet levels and passes them to parse so that they may be typeset as mathematics as per plotmath (see help(plotmath)).
  • label_bquote: This labeller is different that the others in that it is not itself a labeller, but a function that creates a valid labeller. The purpose it to allow the levels of the factor to be substituted into a given expression (which is provided to the generator).

Examples

For examples, we will use the same mpg2 data set as is used in the examples in 7.2 of the ggplot book as well as the same base plot.

mpg2 <- subset(mpg, cyl != 5 & drv %in% c("4", "f")) 
p <- qplot(cty, hwy, data = mpg2)

A simple faceting in both drv and cyl

p + facet_grid(drv ~ cyl)

Base graphic

is the same as if the default labeller is set

p + facet_grid(drv ~ cyl, labeller=label_value)

Labeling with label_value

Note that the argument to labeller can also be the quoted name of a function which is a valid labeller

p + facet_grid(drv ~ cyl, labeller="label_value")

If a different labeller is used, the facet labels change

p + facet_grid(drv ~ cyl, labeller=label_both)

Labeling with label_both

Now instead of the facet labels at the top being "4", "6", and "8", they are "cyl: 4", "cyl: 6", and "cyl: 8". Similarly, the vertical labels are "drv: 4" and "drv: f". There is not a way to set the labels for the horizontal and vertical separately.

For this example, label_parsed does not do anything interesting since the values of the facet variables are not plotmath expressions that would get formatted any differently.

mpg3 <- transform(mpg2,
    cyl = factor(cyl, levels=c(4,6,8), labels=c("alpha", "sqrt(x, y)", "sum(x[i], i==1, n)")))
(p %+% mpg3) + facet_grid(drv ~ cyl, labeller=label_parsed)

Labeling with label_parsed

Here, the (top) facet labels are the greek letter alpha, the y-th root of x, and the sum over i from 1 to n of x sub i.

p + facet_grid(drv ~ cyl, labeller=label_bquote())

Labeling with label_bquote()

First, note that the argument to labeller is not the function label_bquote, but rather the function that is returned when calling label_bquote (with the default arguments). This is because label_bquote is a function which returns a labeller rather than being a labeller itself. The default expression is a greek beta raised to the power of the levels of the factor being faceted on. This expression can be changed with the expr arugment to label_bquote. In the expression, x will be substituted for the value of the level of the factor. For example, to make the facet labels an alpha with the level of the factor as a subscript:

p + facet_grid(drv ~ cyl, labeller=label_bquote(expr = alpha[.(x)]))

Labeling with label_bquote() with a non default expression

Writing new labellers

The structure of a labeller is a function that takes two arguments: variable and value. variable is a length 1 character vector with the name of the variable which is being faceted on. value is the ordered (not necessarily unique) values of the facets. If the faceting variable is a factor, this is a factor including the used levels (note that there may be additional levels to the factor which are not part of the faceting, but were levels of the factor to begin with; thus it is the character representation of the factor which is relevant, not the levels). If the faceting variable is a character variable or a numeric variable, then value is a simple vector of the (not necessarily unique, but ordered) values of the faceting variable. What gets passed as value need not be unique because levels can be repeated in the case of nested faceting; see the last example.

The output of labeller varies quite a lot. All four built in labellers return different structures, all of which are apparently acceptable.

  • A factor whose character representation is the labels
  • A character vector that is the labels
  • A list of expressions, each element of which is one label
  • A list of calls, each element of which is one label

The following is an example of a labeller which wraps long labels at a fixed character count so that they take up less space. This can be useful since the space for facets is not expanded to guarantee that the entire label is visible. (This example is inspired by and derived from this thread on the mailing list.)

label_wrap <- function(variable, value) {
  lapply(strwrap(as.character(value), width=25, simplify=FALSE), 
        paste, collapse="\n")
}  

Alternatively, a label_wrap generator can be made which allows parametrization of the width of wrapping. I prefer to name it differently to differentiate that it is a generator of labellers, not a labeller itself.

label_wrap_gen <- function(width = 25) {
    function(variable, value) {
      lapply(strwrap(as.character(value), width=width, simplify=FALSE), 
            paste, collapse="\n")
    }
}

An example of the usage of these would be

mpg4 <- transform(mpg2, 
    cyl = factor(cyl, levels=c(4,6,8), 
        labels=c("These are cars that have 4 cylinders", 
            "These are other cars that have 6 cylinders", 
            "Here are some 8 cylinder cars if you want that many cylinders")))
    
(p %+% mpg4) + facet_grid(drv ~ cyl, labeller=label_wrap)

Labeling with label_wrap()

(p %+% mpg4) + facet_grid(drv ~ cyl, labeller=label_wrap_gen(width=15))

Labeling with label_wrap_gen() with non-default width

(p %+% mpg4) + facet_grid(~ drv + cyl, labeller=label_wrap_gen(width=15))

Labeling with label_wrap_gen() with non-default width where faceting on two variables in the same direction

Software Versions

This code version of the code was rendered with

  • R version 2.15.0 (2012-03-30)
  • Platform: i386-pc-mingw32/i386 (32-bit)

With the packages

  • ggplot2_0.9.1
  • scales_0.2.1
  • reshape2_1.2.1
  • plyr_1.7.1
  • proto_0.3-9.2

Note: The ggplot2 wiki is no longer maintained, please use the ggplot2 website instead!

Clone this wiki locally