Dplyr filter top and bottom rows by value simultaneously on grouped data

Question

# Sample dataframe
set.seed(123)
d = data.frame(x = runif(120), grp = gl(3, 40))

# Select top_n
d %>%
  group_by(grp) %>%
  top_n(n=3, wt=x)

How do I select both top and bottom observations within the same pipe? Have tried the following but does not work

# helper function
my_top_bott = function(x, n, wt) {
      x1 = x %>% top_n(n=n, wt=wt)
      x2 = x %>% top_n(n=n, wt=-wt)
      x = bind_rows(x1, x2)
      return(x)
    }

# Pipe
d %>%
      group_by(grp) %>%
      my_top_bott(., n=3, wt=x)

Answer 1

One possibility could be:

d %>%
 group_by(grp) %>%
 filter(dense_rank(x) <= 3 | dense_rank(desc(x)) <= 3)

         x grp  
      <dbl> <fct>
 1 0.0456   1    
 2 0.957    1    
 3 0.0421   1    
 4 0.994    1    
 5 0.963    1    
 6 0.0246   1    
 7 0.858    2    
 8 0.0458   2    
 9 0.895    2    
10 0.0948   2    
11 0.815    2    
12 0.000625 2    
13 0.103    3    
14 0.985    3    
15 0.0936   3    
16 0.954    3    
17 0.0607   3    
18 0.954    3 

Or a possibility proposed by @IceCreamToucan:

d %>%
 group_by(grp) %>%
 filter(!between(dense_rank(x), 3 + 1, n() - 3))

Or a possibility involving match():

d %>%
    group_by(grp) %>%
    filter(!is.na(x[match(x, sort(x)[c(1:3, (n()-2):n())])]))

Answer 2

You could also use the row_number().

d %>%
  group_by(grp) %>%
  arrange(desc(x)) %>%
  filter(row_number() > max(row_number()) - 3 | row_number() <= 3)

         x grp  
     <dbl> <fct>
 1 0.995   2    
 2 0.975   2    
 3 0.975   1    
 4 0.974   3    
 5 0.974   3    
 6 0.960   1    
 7 0.960   3    
 8 0.951   2    
 9 0.874   1    
10 0.127   2    
11 0.104   2    
12 0.0693  1    
13 0.0520  1    
14 0.0279  2    
15 0.0146  3    
16 0.0114  3    
17 0.00864 1    
18 0.00333 3