R + dplyr: Partial Deduplication of Rows in a Tibble

A very common question is how to remove all the duplicated lines in a data frame in R, something which can be done with a variety of tools (I like dplyr+distinct).

However, what if your dataset contains several duplicated lines, but you do not want to remove all of them, but only those for some combination of the variables?

I do not know how to achieve that, so any suggestion is welcome.

Please have a look at the reprex at the end of the post.

Thanks!

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union


df <- tibble(x=rep(seq(5), 3), y=rep(LETTERS[1:5],3),
             z=c(rep(c("h","j","k","t","u"), 2), LETTERS[1:5])
             )
df
#> # A tibble: 15 × 3
#>        x y     z    
#>    <int> <chr> <chr>
#>  1     1 A     h    
#>  2     2 B     j    
#>  3     3 C     k    
#>  4     4 D     t    
#>  5     5 E     u    
#>  6     1 A     h    
#>  7     2 B     j    
#>  8     3 C     k    
#>  9     4 D     t    
#> 10     5 E     u    
#> 11     1 A     A    
#> 12     2 B     B    
#> 13     3 C     C    
#> 14     4 D     D    
#> 15     5 E     E

df_ded <- df |>
    distinct()

df_ded
#> # A tibble: 10 × 3
#>        x y     z    
#>    <int> <chr> <chr>
#>  1     1 A     h    
#>  2     2 B     j    
#>  3     3 C     k    
#>  4     4 D     t    
#>  5     5 E     u    
#>  6     1 A     A    
#>  7     2 B     B    
#>  8     3 C     C    
#>  9     4 D     D    
#> 10     5 E     E

## I want to deduplicate only the rows with x==3 and z=="k"

df_ded_partial <- df |>
    distinct(x==3, z=="k") ## but this is not what I mean.

## How to achieve it?

df_ded_partial
#> # A tibble: 3 × 2
#>   `x == 3` `z == "k"`
#>   <lgl>    <lgl>     
#> 1 FALSE    FALSE     
#> 2 TRUE     TRUE      
#> 3 TRUE     FALSE

^{Created on 2023-02-14 with reprex v2.0.2}

We can use group_modify() and check for the condition using the .y argument which is a tibble of the current group. So we can say: if the condition is met return the distinct(.x) group otherwise return the whole group .x.

library(dplyr)

df |>
  group_by(x, z) |>
  group_modify(~ if(.y$x == 3 && .y$z == "k") distinct(.x) else .x)

#> # A tibble: 14 x 3
#> # Groups:   x, z [10]
#>        x z     y    
#>    <int> <chr> <chr>
#>  1     1 A     A    
#>  2     1 h     A    
#>  3     1 h     A    
#>  4     2 B     B    
#>  5     2 j     B    
#>  6     2 j     B    
#>  7     3 C     C    
#>  8     3 k     C    
#>  9     4 D     D    
#> 10     4 t     D    
#> 11     4 t     D    
#> 12     5 E     E    
#> 13     5 u     E    
#> 14     5 u     E

Data from OP

df <- tibble(x=rep(seq(5), 3), y=rep(LETTERS[1:5],3),
             z=c(rep(c("h","j","k","t","u"), 2), LETTERS[1:5])
)

^{Created on 2023-02-14 by the reprex package (v2.0.1)}