Use of length 1 numeric vector for expand gives confusing error message

[alanocallaghan]

The error message when you supply a single numeric value to expand says that a single numeric value is okay, which seems a bit kafkaesque*

library("ggplot2")
ggplot(mtcars) +
  aes(mpg, cyl) +
  geom_point() +
  scale_y_continuous(expand=0)
# Error: `expand` must be a numeric vector with 1 or 2 elements
is.vector(0)
# [1] TRUE
is.numeric(0)
# [1] TRUE
length(0)
# [1] 1

*I confess, I've not read kafka so that's probably wrong, but it reminds me of that Asterix & Obelix cartoon with the forms

[clauswilke]

Yes, the code checks for length 2 or 4 and then produces the error message:

ggplot2/R/scale-expansion.r

Lines 67 to 70 in 885c3c1

	expand_range4 <- function(limits, expand) {
	if (!(is.numeric(expand) && length(expand) %in% c(2,4))) {
	abort("`expand` must be a numeric vector with 1 or 2 elements")
	}

[alanocallaghan]

My understanding based on the docs a few years ago was that expand is multiplicative (ie 0.05 expands by 5%), and if length 1 it expands both sides equally, length 2 each value controls one of left/right or top/bottom. I guess length 2 for each side could mean that each pair is passed to expansion, which seems to be the way things are handled based on some quick testing.

Either way this isn't really clear in the argument documentation - there's no indication of how long the vector should be or what each value corresponds to