Allow functional limits in continuous scales

[econandrew]

An implementation of the idea in #2307, for consideration.

Motivation is to allow programmatic, data-agnostic control of scale limits, taking advantage of rather than having to replicate ggplot's aesthetic calculations. For example, suppose you want limits to always be a multiple of 50:

library(ggplot2)
library(dplyr)

# Example 1: stacked position adjustment

p1 <- ggplot(mpg, aes(class)) + geom_bar(aes(fill = drv))

# define a function to determine limits
limiter <- function(by) {
  function(limits) {
    low <- floor(limits[1]/by)*by
    high <- ceiling(limits[2]/by)*by
    c(low, high)
  }
}

# with PR, you can write something like...
p1 + scale_y_continuous(limits = limiter(50))

# ...current alternative would have to be
themax <- mpg %>% group_by(class) %>% summarise(count = n()) %>% pull(count) %>% max
p1 + scale_y_continuous(limits = limiter(50)(c(0, themax)))

[hadley]

This is a small change and it seems useful, so lets do it! A few suggestions in the comments.

Also needs one or two unit tests. Can you please have a look at the existing tests and think about where they should go?

[hadley]

Since we're adding another condition here, can you please reframe the if statement along these lines:

if (is.null()) {

} else if (is.function()) {

} else if (is.numeric()) {
  
} else {
  stop("Informative error message", call. = FALSE) 
}

[dpseidel]

I did rework this conditional to be more intuitive but since get_limits() is called by both discrete and continuous scales and can be passed character, numeric, POSIXt and now function classes, I ultimately did not add an explicit is.numeric conditional nor an error message. Happy to update further if necessary.

[dpseidel]

Hey @econandrew if you're still interested in this PR, I'm around (:wave: summer intern!) and can help you get this over the finish line if you'd like. Let me know if I can help in anyway.

[thomasp85]

@dpseidel would you consider finishing this off?

[dpseidel]

Yep happy to.

[hadley]

The diff seems to include spurious changes which makes it hard to review.

[dpseidel]

Sorry, yes, I got a little nit picky on the NEWS doc, clearly that should be in a different PR. All cleaned up now.

[thomasp85]

LGTM - thanks for finishing it of Dana

[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/