Repel text labels away from each other in a ggplot2 figure.

plot_repel.R

library(ggplot2)
library(FField)

# You'll have to play with repulsion, cex.x, and cex.y to get satisfactory results.
plot_text <- function(x, y, label, repulsion = 1, cex.x = 110, cex.y = 40) {
  dat <- data.frame(xpos = x, ypos = y, label = label)
  dat$label <- as.character(dat$label)
  
  # Use the FField package to repel the text labels away from each other.
  dat <- cbind(
    dat,
    FFieldPtRep(coords = dat[,c("xpos", "ypos")], rep.fact = repulsion)
  )
 
  # Get the bounding box sizes of the text labels. 
  dat$w <- local({
    plot.new()
    retval <- strwidth(dat$label, cex = cex.x)
    dev.off()
    retval
  })
  dat$h <- local({
    plot.new()
    retval <- strheight(dat$label, cex = cex.y)
    dev.off()
    retval
  })
  
  # Feel free to customize this as you wish.
  ggplot(dat) +
    # Points for the data.
    geom_point(
      aes(x = xpos, y = ypos),
      size = 4
    ) +
    # Line segments connecting the text labels to points.
    geom_segment(
      aes(x = xpos, xend = x, y = ypos, yend = y),
      color = "#666666"
    ) +
    # White rectangles behind text labels to enhance legibility.
    geom_rect(
      aes(xmin = x - w / 2, xmax = x + w / 2,
          ymin = y - h / 2, ymax = y + h / 2),
      fill = "white", alpha = 0.5
    ) +
    # Text labels.
    geom_text(aes(x = x, y = y, label = label), size = 8) +
    theme_bw(base_size = 24)
}

# Seed for reproducibility.
set.seed(43)
d <- data.frame(
  x = runif(10),
  y = runif(10),
  label = letters[1:10]
)

plot_text(d$x, d$y, d$label, repulsion = 0.001, cex.x = 10, cex.y = 20)

plot_wordcloud.R

library(ggplot2)
library(wordcloud)

# Seed for reproducibility.
set.seed(43)
d <- data.frame(
  xpos = runif(10),
  ypos = runif(10),
  label = replicate(paste(sample(letters, 6), collapse=""), n = 10)
)

font.size <- 12

# d <- cbind(d, wordlayout(d$x, d$y, d$label, units = "inches"))
d <- cbind(d, wordlayout(d$x, d$y, d$label, units = "figure", cex = font.size / 4))
d$x <- d$x + d$width / 2
d$y <- d$y + d$ht / 2

ggplot(d) +
  # Points for the data.
  geom_point(
    aes(x = xpos, y = ypos),
    size = 4
  ) +
  # Line segments connecting the text labels to points.
  geom_segment(
    aes(x = xpos, xend = x, y = ypos, yend = y),
    color = "#666666"
  ) +
  # White rectangles behind text labels to enhance legibility.
  geom_rect(
    aes(xmin = x - width / 2, xmax = x + width / 2,
        ymin = y - ht / 2, ymax = y + ht / 2),
    fill = "blue", alpha = 0.5
  ) +
  # Text labels.
  geom_text(aes(x = x, y = y, label = label), size = font.size) +
  theme_bw(base_size = 24)

repel.png

wordcloud.png

Thanks to Grigori Kapoustin for the FField package. I learned about strwidth and strheight from this stackoverflow post.

If this does not suit you, please consider some alternatives:

• directlabels, and see a few examples here.
• wordcloud, and have a look at the wordlayout function.

Thinking about how to make a new geom or grob...

• Read these:
  • https://github.com/hadley/ggplot2/wiki/Creating-a-new-geom
  • https://github.com/hadley/ggplot2/blob/master/vignettes/extending-ggplot2.Rmd
• Copy code from this: https://github.com/hadley/ggplot2/blob/master/R/geom-text.r

After exploring the limitations of these packages I decided to make my own package: https://github.com/slowkow/ggrepel