ggplot() scaling with scale::percent_format() producing strange results

Question

library(tidyverse)
mtcars %>% 
  count(cyl) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(x = cyl, y = prop)) + 
  geom_point() + 
  scale_y_continuous(labels = scales::percent_format(accuracy = 5L))

If I use scales::percent() above instead of scales::percent_format(accuracy = 5L) I get decimal places in my percentage labels, which I don't want.

The question - what does 5L do in my example above? Why do I need to use the integer 5L instead of 5? And why does 6L change the highest y-value from 40% to 42%? That's just plain strange.

Answer 1

For 5 digits after comma

library(ggplot2)

library(tidyverse)

mtcars %>% 
  count(cyl) %>% 
  mutate(prop = n / sum(n)) %>% 
  ggplot(aes(x = cyl, y = prop)) + 
  geom_point() + 
  scale_y_continuous(labels = scales::percent_format(accuracy=.00001))

Answer 2

First, it doesn't need to be precisely specified as an integer (i.e. 5 works just fine).

Second, you can do ?scales::percent_format at any time in an R console (it's free!). Doing so tells you this about the function:

percent_format(
  accuracy = NULL, scale = 100, prefix = "", suffix = "%",
  big.mark = " ", decimal.mark = ".", trim = TRUE, ...
)

So, it takes many possible parameters all of which have defaults and some are options (via ...).

The default for the accuracy parameter is NULL. If we scroll down just a bit on the help page for the function we see:

• accuracy: Number to round to, NULL for automatic guess.

If we type the function name without parens or a ? prefix, we can see the entire source. Doing so shows that it ultimately calls scales::number() which is defined as:

function (x, accuracy = 1, scale = 1, prefix = "", suffix = "", 
          big.mark = " ", decimal.mark = ".", trim = TRUE, ...) {
  if (length(x) == 0) return(character())
  accuracy <- accuracy %||% precision(x)
  x <- round_any(x, accuracy/scale)
  nsmall <- -floor(log10(accuracy))
  nsmall <- min(max(nsmall, 0), 20)
  ret <- format(scale * x, big.mark = big.mark, decimal.mark = decimal.mark, 
                trim = trim, nsmall = nsmall, scientific = FALSE, ...)
  ret <- paste0(prefix, ret, suffix)
  ret[is.infinite(x)] <- as.character(x[is.infinite(x)])
  ret[is.na(x)] <- NA
  ret
}

This:

accuracy <- accuracy %||% precision(x)

says if accuracy is not NULL use it otherwise guess by using the precision() function.

The next line after that is the ultimate answer to your question.