Error using ggsave with purrr::map; Working in release 3.3.3, but not >= 3.3.4

[phdjsep]

---

I'm using purrr::map and ggsave to write out multiple plots to disk and the following code works in ggplot2 release = 3.3.3, but not >= 3.3.4. Can't pinpoint what's changed that prevents this from working.

ggplot2 v3.3.3 (working)

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(ggplot2)
library(purrr)

# Working with ggplot 3.3.3, but not 3.3.4
diamonds %>%
  group_by(cut) %>%
  group_split() %>%
  purrr::map(~ ggsave(filename = paste("diamonds", .x$cut, .x$color, sep = "_"),
                      plot = ggplot(.x, aes(x = carat, y = depth)) +
                        geom_point(),
                      device = 'jpeg',
                      path = "~/Desktop/empty/"))
#> Saving 7 x 5 in image
#> Saving 7 x 5 in image
#> Saving 7 x 5 in image
#> Saving 7 x 5 in image
#> Saving 7 x 5 in image
#> [[1]]
#> NULL
#> 
#> [[2]]
#> NULL
#> 
#> [[3]]
#> NULL
#> 
#> [[4]]
#> NULL
#> 
#> [[5]]
#> NULL

ggplot2 v3.3.4 (broken)

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(ggplot2)
library(purrr)

# Working with ggplot 3.3.3, but not 3.3.4
diamonds %>%
  group_by(cut) %>%
  group_split() %>%
  purrr::map(~ ggsave(filename = paste("diamonds", .x$cut, .x$color, sep = "_"),
                      plot = ggplot(.x, aes(x = carat, y = depth)) +
                        geom_point(),
                      device = 'jpeg',
                      path = "~/Desktop/empty/"))
#> Saving 7 x 5 in image
#> Error in if (!dir.exists(dir)) {: the condition has length > 1

[teunbrand]

As far as I can tell, you're giving multiple filenames to ggsave(). Giving just a single filename per plot should help run the code again, e.g. filename = paste("diamonds", .x$cut[1], .x$color[1], sep = "_"). Maybe ggsave() should throw a warning when multiple filenames are given?

[phdjsep]

Thanks for the reply @teunbrand. Unfortunately that would defeat the purpose of what I'm after. I don't want to write 30+ ggsave calls to name files individually. Also, the code above works with ggplot2 v3.3.3, but not ggplot2 v3.3.4 or greater so it wasn't an issue with ggsave previously.

[teunbrand]

Yes I understand that this used to work, but I'm trying to find out whether this issue is a true regression or off-label use that has broken.

Perhaps I have caused a little bit of confusion. In every purrr::map() iteration, the ggsave() function receives a vector of however many rows were in that group as the filename argument. In this case, every plot that you'd want to save has multiple filenames associated with it.

I'd agree with you that the error is quite uninstructive, so perhaps that could be improved.

[phdjsep]

Ah. I see. That’s a helpful link, thank you. I will try and dig in more then. 🙏🏻

[phdjsep]

Thank you for your helpful advice @teunbrand! I think this was a combination of the "off-label" issue you indicated as well as my oversight of that filename parameter. You were correct in that I was passing in more than one filename, so ggsave was choking on that (as the error indicated).

So the 3.3.4 release did change things in that it was previously lenient with my passing it multiple filenames, but also happy to say that it was an easier fix and I can still use the purrr approach to generating the plots using ggsave.

For those interested, here's the updated example which works with ggplot 3.4.0. The changes are in using the unique() function to pull a single value from the df as well as removing the .x$color as that results in multiple values again since I'm not grouping by it, only cut.

library(dplyr)
library(ggplot2)
library(purrr)

diamonds %>%
  group_by(cut) %>%
  group_split() %>%
  purrr::map(~ ggsave(filename = paste("diamonds", unique(.x$cut), sep = "_"), 
                      plot = ggplot(.x, aes(x = carat, y = depth)) +
                        geom_point(),
                      device = 'jpeg',
                      path = "~/Desktop/empty/"))