do.grouped_dt drops grouping columns

[kismsu]

Hi, I've noticed that do on grouped data.table doesn't send all columns into function.

library(nycflights13)
library(dplyr)

g <- function(x) {
    browser()
}

flights %>%
    data.table() %>%
    select(origin, flight, distance, dep_delay, hour, minute) %>%
    group_by(flight, distance, dep_delay) %>%
    do(
        data.frame(g(.))
    )

Inside g() you'll see

Source: local data table [1 x 3]
Groups: flight, distance, dep_delay

  origin hour minute
1    EWR   14      0

As I believe the goal is to have unified results for all sources, hence function should receive all columns.
The solutions might be (based on http://stackoverflow.com/questions/26636402/how-i-can-get-the-value-of-grouping-column-inside-user-define-function-in-data-t/26641743#26641743) replace in do.grouped_dt

call <- substitute(dt[, cols, by = vars], list(cols = cols)) 

by

call <- substitute(dt[, cols, by = vars, .SDcols = names(dt)], list(cols = cols)) 

or add names to dt_env.

[hadley]

Slightly simpler test case:

mtcars %>% 
  tbl_dt() %>%
  select(mpg, cyl) %>%
  group_by(cyl) %>%
  do(n = names(.)) %>%
  str()