data.table::transpose applied to lists changes element types

[MLopez-Ibanez]

Applying data.table::transpose to a data.table that contains numerical and string values will convert everything to strings. By comparison purrr::transpose preserves the original types by creating a list of lists. Note that it is very easy to convert a list of lists to a list of vectors (that is, get the output of data.table::transpose from the output of purrr::transpose) but very hard the other way.

# Minimal reproducible example; please be sure to set verbose=TRUE where possible!

library(data.table)
dt <- data.table(num=runif(3), char = letters[1:3])
str(data.table::transpose(dt)) # I want a data.table, so fine.
# Classes ‘data.table’ and 'data.frame':	2 obs. of  3 variables:
#  $ V1: chr  "0.21412930265069" "a"
#  $ V2: chr  "0.0449177662376314" "b"
#  $ V3: chr  "0.299894759198651" "c"
#  - attr(*, ".internal.selfref")=<externalptr> 
str(data.table::transpose(as.list(dt))) # I want a list, so it should be a list of lists.
# List of 3
#  $ : chr [1:2] "0.21412930265069" "a"
#  $ : chr [1:2] "0.0449177662376314" "b"
#  $ : chr [1:2] "0.299894759198651" "c"
library(purrr)
str(purrr::transpose(dt)) # Exactly right!
# List of 3
#  $ :List of 2
#   ..$ num : num 0.214
#   ..$ char: chr "a"
#  $ :List of 2
#   ..$ num : num 0.0449
#   ..$ char: chr "b"
#  $ :List of 2
#   ..$ num : num 0.3
#   ..$ char: chr "c"

This is a very frequent operation, which is currently very hard to do with data.table. See

https://stackoverflow.com/questions/3492379/data-frame-rows-to-a-list
https://stackoverflow.com/questions/27784076/r-data-table-to-list-of-rows-better-ways
https://stackoverflow.com/questions/41005943/turn-all-data-table-rows-into-a-list-r
https://stackoverflow.com/questions/56901903/convert-a-data-table-to-list-of-rows-that-are-data-tables-and-apply-a-function-t

(and more)

[MLopez-Ibanez]

The solution recommended often in stackoverflow (and used by data.table:::split.data.table):

lapply(seq_len(nrow(dt)), function(ind) as.list(dt[ind]))

is even slower than the same operation with data.frame (around 5 times slower in my computer).