Save plot exactly as previewed in the "Plots" panel

Question

I know that a very similar question already exists here, but the provided answer did not work for me.

This is my usual workflow: I generate a plot and adjust the size of the plot in the "Plots" Panel of RStudio until I am satisfied. I then call dev.size() to get the exact size. Afterwards, I save the plot with ggsave(...,dpi=300) and specify the previously determined width and height. The problem is, that after saving, the plot looks completely different, especially the text sizes.

If I use the "Export" option from RStudio the plot looks exactly as it does in the preview, but the quality is quite bad and doing this manually is tedious.

Here is the picture, that hopefully illustrates what I mean:

The code I use to save the plot looks like this:

library(ggplot2)

ggplot(mtcars, aes(x = mpg, y = wt)) + 
  geom_point() + 
  facet_grid(vs + am ~ gear, margins = "vs") +
  theme_Publication()
ggsave("plot.png", width=4, height=3.2, dpi=300)

I would love to know, if there is an option to "programmatically" save a plot which exactly recreates the "Plots" preview in high quality.

Thanks a lot in advance!

Answer 1

1. In RStudio, click on "Zoom", and then zoom the plot to the size and aspect ratio that satisfy you;
2. Right click on the plot, then "Copy Image Address", paste the address somewhere, you get the width (www) and height (hhh) information in the address;
3. ggsave("plot.png", width = www/90, height = hhh/90, dpi = 900) ggsave("plot.pdf", width = www/90, height = hhh/90)

Answer 2

You can use dev.print:

dev.print(pdf, 'filename.pdf')

This will save the plot as it appears in the RStudio plot tab. pdf can be replaced with other file formats.
To obtain a .svg file that is easy to import into word processors use

dev.print(svg, 'filename.svg')

An SVG file does not consist of fixed pixels but can be scaled to any height or width.

png/jpeg/bmp

In part 2 of this answer it is claimed that the filetype for dev.print can also be png, jpeg and bmp. But in my experience simply using dev.print(png, 'filename.png') will result in a plot with too small text. Instead, if the output must be a fixed-pixel format such as png, using the package 'rsvg' produces better results:

#install.packages("rsvg")
library(rsvg)

#Create a temporary file
tmp <- tempfile()

#Put the current plot into the tempfile in the svg format
dev.print(svg,tmp)

#Convert the svg temp file to png and store it in plot.png
rsvg_png(tmp, "plot.png",height=2000)

Only setting the height or width (could replace height with width above) preserves the aspect ratio of the plot as it shown in RStudio. Increase the number of pixels (2000 for height in the above) to improve the resolution of the image.

Answer 3

1. I found this resource very helpful for understanding resolution and text sizes in ggplot2: https://www.christophenicault.com/post/understand_size_dimension_ggplot2/

2. It sounds like you'd benefit from using the camcorder package, which offers an alternate workflow. https://thebioengineer.github.io/camcorder/index.html Instead of showing each plot tweak in the Plots window with a fidgety resolution and size that might not match your ultimate destination, camcorder::gg_record will render the plot to your dimensional specifications and show it in the Viewer pane as a final plot. It's equivalent to saving each plot with ggsave and opening each result -- but all automatic and within the RStudio IDE.

Here's an example:

library(ggplot2)
library(camcorder)
camcorder::gg_record() 

This will now overwrite the plotting device so each plot will be saved to a temporary directory and displayed in the Viewer.

p <- ggplot(mtcars, aes(x = mpg, y = wt)) + 
       geom_point() + 
       facet_grid(vs + am ~ gear, margins = "vs")
p

I can now see that plot in the Viewer. Maybe I don't like the dimensions, so I run:

gg_resize_film(
  height = 4,
  width = 7,
  units = "in",
  dpi = 300
)

That changes the output from now on to 7x4". But now the text is too large for what I want.

So I can tweak that:

p + theme_gray(base_size = 8)

Good, I like that. If I navigate to my tempdir(), I see the files I made. If I want to keep any of those, I should move them into a real folder, otherwise they'll be deleted after this R session. Or I could initially specify a non-temp folder for all the outputs to go with gg_record(dir = "my_dir").

Camcorder also has a neat camcorder::gg_playback() function which makes a gif out of all the drafts -- neat for documenting your process visually.

Answer 4

You can export a graph in a programmatic way to various formats (png, ppt, pdf, svg, doc) using my export package, including in the size at which the graph is shown on-screen.

Plot in RStudio Plots window after sizing window to a good size :

Note: if you would like to have your plot in a separate window that you manually resize you can call x11() first (or windows(), but x11() works on both Windows & Mac+Ubuntu, windows() only on Windows) :

x11()

Make plot :

library(ggplot2)
library(export)
# theme_Publication() sourced from rpubs.com/Koundy/71792

ggplot(mtcars, aes(x = mpg, y = wt)) + 
     geom_point() + 
     facet_grid(vs + am ~ gear, margins = "vs") + theme_Publication()

Resize plot window to appropriate size.

Export plot in size of plot window to png & Powerpoint :

# export to PNG at current size:
graph2png(file="plot.png", width=dev.size(units="px")[[1]]/90, height=dev.size(units="px")[[2]]/90, dpi=900)
# note: this would be equivalent to ggsave(file="plot.png", width=dev.size(units="px")[[1]]/90, height=dev.size(units="px")[[2]]/90, dpi=900)

# export to Powerpoint (in editable vector format) at current size:
graph2ppt(file="plot.pptx", width=dev.size(units="px")[[1]]/90, height=dev.size(units="px")[[2]]/90)

# export to pdf (but you need pdf editor to edit) at current size:
graph2pdf(file="plot.pdf", width=dev.size(units="px")[[1]]/90, height=dev.size(units="px")[[2]]/90)

# export to svg (vector format, you can use InkScape to edit) at current size:
graph2svg(file="plot.svg", width=dev.size(units="px")[[1]]/90, height=dev.size(units="px")[[2]]/90)

For publication, I typically use either Powerpoint format (can be edited in PPT after right clicking & pressing ungroup, print to PDF once you are happy with the result) or PDF or SVG, as they are vector format and print much better than a bitmap format like PNG.

To close the plot window you can call

dev.off()