[FIXED] Why does pivot_longer convert column type from 'double' to 'list'

Issue

After my data frame is converted to long, manipulated, then converted back to wide, the column types seem to have changed.

x = data.frame(A = rnorm(100), 
               b = rnorm(100)) %>%
  mutate(id = row_number())

typeof(x[,'A'])
# produces "double"

x3 = x %>% pivot_longer(-id) %>%
  pivot_wider(names_from = name, values_from = value)

typeof(x3[,'A'])
# produces "list"

typeof(x3[,'A'] %>% unlist())
# produces "double"

This is a problem because I want to loop through an array and assign parts of the df to parts of the array. For example:

arr = array(dim = c(2,100))
# arr has type 'logical'

arr[,1] = x3[,'A']
# arr now has type 'list'

arr[,2] = x3[,'A']

#last line gives me: 'Error in arr[, 2] = x3[, "A"] : incorrect number of subscripts on matrix'

Assigning a slice of the array to a list seems to convert the whole thing to a list. I believe I can get round this by replacing the last line with arr[,2] = x3[,'A'] %>% unlist(), but it's suck strange behaviour I want to know what's going on.

Solution

The reason is that x3 is a tibble and using x3[, 'A'] is still a tibble which you could check via class(x3[,'A']). And as a tibble or a data.frame is basically a list typeof will give you a type of list (check e.g. typeof(mtcars)). This is one of the differences between a data.frame and a tibble. In case of a data.frame x3[,'A'] would be simplified to a vector by default which is not the case for a tibble.

If your want a vector then you have to be more explicit when slicing from a tibble by using x3[,'A', drop = TRUE] or x3[["A"]] or x3$A:

library(dplyr, warn=FALSE)
library(tidyr)
set.seed(123)

x = data.frame(A = rnorm(100), 
               b = rnorm(100)) %>%
  mutate(id = row_number())

x3 <- x %>% 
  pivot_longer(-id) %>%
  pivot_wider(names_from = name, values_from = value)

class(x3[,'A'])
#> [1] "tbl_df"     "tbl"        "data.frame"

typeof(x3$A)
#> [1] "double"

typeof(x3[,'A', drop = TRUE])
#> [1] "double"

typeof(x3[['A']])
#> [1] "double"

Answered By - stefan

Answer Checked By - Willingham (PHPFixing Volunteer)