In ggplot2 you can pass character arguments inside a user defined function using aes_string. How can you do the same for facet grid which takes a formula, not aes?
FUN <- function(data, x, y, fac1, fac2) {
ggplot(data = data, aes_string(x=x, y=y)) +
geom_point() + facet_grid(as.formula(substitute(fac1 ~ fac2)))
}
FUN(mtcars, 'hp', 'mpg', 'cyl', 'am')
reformulate() seems to work just fine.
FUN <- function(data, x, y, fac1, fac2) {
ggplot(data = data, aes_string(x=x, y=y)) +
geom_point() + facet_grid(reformulate(fac2,fac1))
}
FUN(mtcars, 'hp', 'mpg', 'cyl', 'am')
reformulate.
reformulate(fac1, ".") also works.With rlang magic and new ggplot2 v3.0.0 features you can do :
FUN <- function(data, x, y, fac1, fac2) {
ggplot(data = data, aes(x = !!ensym(x), y = !!ensym(y))) +
geom_point() +
facet_grid(eval(expr(!!ensym(fac1) ~ !!ensym(fac2))))
}
FUN(mtcars, 'hp', 'mpg', 'cyl', 'am')
Note that we don't use aes_string, which is soft deprecated.
Personally in these cases I like to use a function I called glue_formula (referencing package glue) :
glue_formula <- function(.formula, .envir = parent.frame(), ...){
formula_chr <- gsub("\\n\\s*","",as.character(.formula)[c(2,1,3)])
args <- c(as.list(formula_chr), .sep=" ", .envir = .envir)
as.formula(do.call(glue::glue, args),env = .envir)
}
FUN2 <- function(data, x, y, fac1, fac2) {
ggplot(data = data, aes(x = !!ensym(x), y = !!ensym(y))) +
geom_point() + facet_grid(glue_formula({fac1} ~ {fac2}))
}
FUN2(mtcars, 'hp', 'mpg', 'cyl', 'am')
It's not tidyverse approved (see interesting discussion here) but it has served me well.
reformulate(fac2,fac1) seems so much cleaner and easier to reason about than eval(expr(!!ensym(fac1) ~ !!ensym(fac2)))
as.formula(paste(fac1,"~", fac2))