Feature request: could functionality be provided to guess ggplot height and width?

[davidhodge931]

Hi @teunbrand, seeing as this functionality is not going to be provided in {ggplot2}, maybe it could be added into {ggh4x}?

An educated guess at dimensions would be useful. If it works, great. If it doesn't, you've at least got a starting point to tweak from. Otherwise, it is going to be a super pain to save any plots with a specified panel height.

tidyverse/ggplot2#6367

[teunbrand]

Hi David, yeah I'm not opposed to that. It fits into the 'hacks' theme of ggh4x as a 95% solution of a common problem. Not sure when I'll have the time/motivation to add it to ggh4x, but probably before new ggplot2 version hits the shelves.

[davidhodge931]

Thanks @teunbrand - would be super useful

[pormr]

Maybe we could implement something like this (inspired by egg::set_panel_size¹²):

ggsave_auto <- function(filename, plot = ggplot2::last_plot(),
                        margin = unit(1, "mm"), ...) {
  if (!any(grepl("^Forced", class(plot$facet)))) {
        warning("This ggplot object doesn't have fixed size panel. ",
                "Please apply `ggh4x::force_panelsizes()` first.")
  }
  
  g <- ggplot2::ggplotGrob(plot)
  w <- grid::convertWidth(sum(g$widths) + margin, unitTo = "in", valueOnly = TRUE)
  h <- grid::convertHeight(sum(g$heights) + margin, unitTo = "in", valueOnly = TRUE)
  
  ggplot2::ggsave(filename, plot, width = w, height = h, ...)
}

Usage:

p <- ggplot(mtcars, aes(x = mpg, y = wt, colour = factor(cyl))) +
  geom_point() +
  facet_wrap(~ carb, nrow = 1) +
  ggh4x::force_panelsizes(rows = unit(2, "cm"),
                          cols = unit(1, "in"))

ggsave_auto("mtcars.png", p)

Footnotes

1. https://github.com/baptiste/egg/tree/master?tab=readme-ov-file#setting-panel-size ↩

2. https://github.com/baptiste/egg/blob/c4dac61c847f036224a82cd0be06ab10fc02fa87/R/set_panel_size.r#L20-L38 ↩

[teunbrand]

Fixed sizes will become part of base ggplot2, so it should be independent of ggh4x::force_panelsizes().
Other than that, implementation will be pretty close to what I had in tidyverse/ggplot2#6371.