How to extract labels shown by ggrepel

[jttoivon]

Hi,

How can I extract a list of those labels that ggrepel shows in the result figure?
In the below example I create 500 points, and ggrepel shows label for 395 of these points only. However, when I
try to extract the list of these 395 labels, I get all 500 labels. Any idea what I am doing wrong?

library(tidyverse)
library(ggrepel)
library(grid)
dev.size()
create_data <- function(n=10) {
  x <- runif(n)
  y <- runif(n)
  label <- paste0("***",1:n,"***")
  df <- tibble(x=x, y=y, label=label)
  df
}
set.seed(123)
tmp <- create_data(n=500)
g <- tmp %>% ggplot(aes(x, y, label=label)) + geom_text_repel()
g
extract_labels <- function(g, width=9.8, height=7) {
  # Trying to imitate solution to this https://github.com/slowkow/ggrepel/issues/24
  grid.newpage()
  pushViewport(viewport(width = width, height = height))
  gg <- g %>% ggplotGrob() %>% grid.force()
  print(gg)
  xx <- gg %>% getGrob("textrepeltree", grep = TRUE)
  xx
}
res <- extract_labels(g)
n_distinct(res$data$label)   # all 500 are here
length(res$children)         # this is also 500

Here is an image of the output produced by the code:

[slowkow]

Sorry, but I don't know how to help.

My understanding is that ggrepel should only be creating a grob for labels that are being plotted, but it should not be creating grobs for unplotted labels. If that's not happening, then maybe the plot is not being rendered appropriately? I don't know.

If anyone eventually finds something that works, I'd appreciate if you share it here so everyone can learn.

[jttoivon]

Thanks for the reply!

I got it working! It seems that the main problem was incorrect units in the viewport() call. I assumed the units would be in inches, but that isn't the case. The second problem was that I was doing the plotting and extracting separately. But now in my fixed example the extracted data should correspond to what is plotted.

library(tidyverse)
library(ggrepel)
library(grid)
#dev.size()
create_data <- function(n=10) {
  x <- runif(n)
  y <- runif(n)
  label <- paste0("***",1:n,"***")
  df <- tibble(x=x, y=y, label=label)
  df
}
set.seed(123)
tmp <- create_data(n=500)
g <- tmp %>% ggplot(aes(x, y, label=label)) + geom_text_repel()
extract_labels <- function(g, width=9.8, height=7) {
  force(g)
  # Trying to imitate solution to this https://github.com/slowkow/ggrepel/issues/24
  grid.newpage()
  #vp <- viewport(width = width, height = height)  # Wrong units!!!
  vp <- viewport(width = unit(width, "in"), height = unit(height, "in"))
  pushViewport(vp)
  gg <- g %>% ggplotGrob() %>% grid.force()
  print(g, vp=vp)  # Make the plot appear
  repeltree <- gg %>% getGrob("textrepeltree", grep = TRUE)
  popViewport()
  repeltree
}
res <- extract_labels(g, 5, 5)
n_distinct(res$data$label)   # all 500 are here
length(res$children)         # now this prints 20!!!!

[slowkow]

This is great! Thank you so much for sharing 🙏

I think this might help others who are trying to use the outputs from ggrepel.