geom_text_repel

[KasperSkytte]

Would be super awesome if geom_text_repel() would work with plotly since I mainly use plotly for plots totally crowded with text

[cpsievert]

Duplicate of #566

[klin333]

if anyone is desperate for ggrepel in plotly, you can always use the non-exposed internal functions of ggrepel to compute the repelled label x y coordinates, then plot the already repelled coordinates with plotly via the standard geom_text/ggplotly or any native plotly functions. Of course this isn't supported by ggrepel and will be very brittle, you will have to mess around with function arguments depending on ggrepel version.

library(tidyverse)
library(stringr)

# my packages: plotly_4.9.2;  ggplot2_3.2.1; ggrepel_0.8.0 
repel_coords <- function(x, y, label, char_width, char_height, xlim, ylim, trans_x = 'identity', trans_y = 'identity', var_names = c('repel_x', 'repel_y'), 
                         point_padding_x = 0, point_padding_y = 0, hjust = 0, vjust = 0, maxiter = 8000, force = 1e-6, direction = 'both') {
  N <- length(x)
  hjust <- rep(hjust, N)
  vjust <- rep(vjust, N)
  
  exp10 <- function(x) 10^x
  trans_map <- c(
    'log' = 'exp',
    'log10' = 'exp10',
    'identity' = 'identity'
  )
  stopifnot(trans_x %in% names(trans_map))
  stopifnot(trans_y %in% names(trans_map))
  inv_x <- get(trans_map[trans_x])
  inv_y <- get(trans_map[trans_y])
  trans_x <- get(trans_x)
  trans_y <- get(trans_y)
  
  x <- trans_x(x)
  xlim <- trans_x(xlim)
  y <- trans_y(y)
  ylim <- trans_y(ylim)
  
  data_points <- tibble(x, y) %>% as.matrix()
  l <- str_split(label, '\n')
  heights <- map_dbl(l, length) * char_height
  widths <- map_dbl(l, compose(max, str_length)) * char_width
  boxes <- tibble(x-widths/2, y-heights/2, x+widths/2, y+heights/2) %>% as.matrix()
  df <- ggrepel:::repel_boxes(data_points, point_padding_x, point_padding_y, boxes, xlim, ylim, hjust, vjust, maxiter = maxiter, force = force, direction = direction)
  df <- as_tibble(df)
  
  df$x <- inv_x(df$x)
  df$y <- inv_y(df$y)
  
  df <- df %>% set_names(var_names)
  df
}

# example usage
data_df <- tibble(x = rnorm(30, 0, 5), y = rnorm(30, 0, 5))

label_df <- tribble(
  ~x, ~y, ~label,
  1,  2,  'item1',
  1,  2.1, 'item2',
  3,  4, 'item3',
  3.1,  4, 'item4'
) %>%
  nest() %>%
  mutate(repel = map(data, function(d) repel_coords(d$x, d$y, d$label, 1, 1, c(-10, 10), c(-10, 10), force = 1e-4))) %>%
  unnest()

pl <- data_df %>%
  ggplot(aes(x = x, y = y)) + 
  geom_point() + 
  geom_point(data = label_df, color = 'red') + 
  # geom_text(data = label_df, aes(label = label)) 
  geom_text(data = label_df, aes(x = repel_x, y = repel_y, label = label))

ggplotly(pl)

[cpsievert]

Nice @klin333! If you were to post this example code on https://community.rstudio.com/ I'm sure it would get a lot more visibility :)