On retaining class data.table (or not)

[Henrik-P]

After answering How to retain data.table class when piped through dplyr::filter? on SO, I thought some observations could be worth mentioning here.

Create a data.table:

library(data.table)
d <- data.table(mtcars)
class(d)
# "data.table" "data.frame"

(1) Apply dplyr verbs directly on the data.table

• class data.table is retained when using select

  library(magrittr)
  d %>% dplyr::select(hp, mpg) %>% class
  # "data.table" "data.frame"
  

• class data.table is not retained when using filter

  d %>% dplyr::filter(hp > 100) %>% class
  # "data.frame"
  

First, I realize that this most likely not be 'the correct' way of using dplyr verbs on a data.table (hence the dtplyr package), but this was the premise of the SO question. Still, the .data argument and Value are described in the same way in ?filter and ?select, so from this information only it's hard to tell why .data of class data.table is treated differently in the two functions.

(2) I suppose the correct dtplyr way would be to explicitly convert the data.table to a "data table tbl" using tbl_dt to retain class data.table:

library(dtplyr)
d %>% tbl_dt() %>% dplyr::select(hp, mpg) %>% class
# [1] "tbl_dt"     "tbl"        "data.table" "data.frame"

d %>% tbl_dt() %>% dplyr::filter(hp > 100) %>% class
# [1] "tbl_dt"     "tbl"        "data.table" "data.frame"

(3) However, merely loading dtplyr is enough to retain class data.table. Which was a bit unexpected.

d %>% dplyr::select(hp, mpg) %>% class
# [1] "data.table" "data.frame"

d %>% dplyr::filter(hp > 100) %>% class
# [1] "data.table" "data.frame"

[hadley]

I don't think 3) should be too unexpected - where better for filter.data.table method to live?

[Henrik-P]

I assume you are right :) For some reason I thought an explicitly conversion to a "data table tbl" was necessary to retain the data.table class.