Allow empty annotations.

[clauswilke]

This fixes #3317, and likely some other current issues.

library(ggplot2)
ggplot() +
  annotate("point", x = numeric(0), y = numeric(0), colour = "red")

^{Created on 2019-05-08 by the reprex package (v0.2.1)}

[clauswilke]

@hadley @yutannihilation @thomasp85 I would need a formal approval before I can merge this. It was Ok'd informally by Thomas here: #3317 (comment)

[hadley]

Can you please use our standard sectioning convention?

# modify_list() ----------------------------------------

[hadley]

Why does it fail?

[clauswilke]

Maybe a poorly formulated comment. The purpose of line 49 is to find the unique lengths that are not 1L. However, if lengths == c(1L), then we get n <- numeric(0) in line 49, and we need n <- 1L in this case.

[hadley]

Maybe this would be a clearer approach?

ulength <- unique(lengths)
unequal <- length(ulength) > 1 || ulength != 1

[clauswilke]

No, that doesn't capture it. unequal should only be true if there is more than one length excluding 1. So if lengths <- c(1, 2), we should have unequal <- FALSE, but if lengths <- c(2, 3), we should have unequal <- TRUE. I'll see if I can find a better way to code this.

The original statement to calculate unequal was exactly right:

unequal <- length(unique(setdiff(lengths, 1L))) > 1L

But now I'm also trying to capture the length n in addition, that required the additional code.

[clauswilke]

I've thought about it some more, and I can't come up with something that would be clearer and simpler than the code in the current PR. Again, we also need to capture n, since it's used in line 64:

ggplot2/R/annotation.r

Line 64 in 4ff8243

data <- new_data_frame(position, n = n)

[clauswilke]

When I woke up this morning, I realized that using setdiff() only when needed may make the code easier to understand. I pushed a new version.

[hadley]

That's much easier to understand — thanks!

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/