warning message "argument is not an atomic vector; coercing" when mutating in a list column

I do some reshaping within my data. I’m basically doing two things:

• All x cases with the same value should be nested/summarized into a list column.
• For the remaining x values I will have a character value in y separated by one or more commas. I want to split up these character values into separate values and then put that into the existing y list column as character vector. So e.g. for x == "second", I do have the char vector "C, D" (so just one value) and I want to create a char vector c("C", "D"), i.e. with length 2.

The code below seems to do what I want, but I’m getting a warning message. Although it is only a warning message, I want to make sure I’m doing the right thing.

library(tidyverse)
df <- data.frame(x = c("first", "first", "second", "third"),
                 y = c("A", "B", "C, D", "E, F, G"))

df

       x       y
1  first       A
2  first       B
3 second    C, D
4  third E, F, G

.

df |>
  group_by(x) |>
  summarise(y = list(y)) |> 
  rowwise() |>
  mutate(y = list(as.vector(y, mode = "character"))) |>
  ungroup() |>
  mutate(across(y, ~if_else(!str_detect(x, "first"), str_split(., ", "), y)))

which (correctly) gives:

# A tibble: 3 x 2
  x      y        
  <chr>  <list>   
1 first  <chr [2]>
2 second <chr [2]>
3 third  <chr [2]>

But with a warning:

Warning message:
Problem with `mutate()` input `..1`.
i `..1 = across(...)`.
i argument is not an atomic vector; coercing 

Waht can/should I do?

>Solution :

You can get this result more easily with tidyr::separate_rows:

df |>
  separate_rows(y) |>
  group_by(x) |>
  summarise(y = list(y))

#     x      y        
#     <chr>  <list>   
#   1 first  <chr [2]>
#   2 second <chr [2]>
#   3 third  <chr [3]>