Regression: ggsave() fails if `device` is function without `...`

[bwiernik]

Since a9b3b17, if a device is passed to ggsave() and this device lacks a ... argument (at least if it also uses a file argument instead of filename), it produces an "unused argument" error.

ggplot2::ggsave("test.emf", ggplot2::ggplot(), device = devEMF::emf)
#> Saving 7 x 5 in image
#> Error in (function (file = "Rplot.emf", width = 7, height = 7, bg = "transparent", : unused argument (filename = "test.emf")

^{Created on 2021-07-02 by the reprex package (v2.0.0)}

[clauswilke]

Is this different from #4521?

[yutannihilation]

It's a different package.

https://cran.r-project.org/package=devEMF

Looking at the diff of a9b3b17, I guess the problem is that we remap filename to file but don't remove the original one from the arguments, so probably both "file" and "filename" are passed?

ggplot2/R/save.r

Lines 161 to 172 in acaa98e

	args <- formals(device)
	call_args <- list()
	if ("file" %in% names(args)) {
	call_args$file <- filename
	}
	if ("res" %in% names(args)) {
	call_args$res <- dpi
	}
	if ("units" %in% names(args)) {
	call_args$units <- 'in'
	}
	dev <- function(...) do.call(device, modify_list(list(...), call_args))

[bwiernik]

Yes, that’s the issue.