Need to distinguish rows vs. cols

[hadley]

Maybe map_dfc() and map_dfr()? (Nice because then all types have a 3 letter abbreviation).

[hadley]

Or have map_df(x, .direction = 1).

[lionel-]

Maybe .collate = c("rows", "cols") to be consistent with the rows-based functionals.

[hadley]

I think I'd rather keep these as separate functions: these are special cases because data frames are so important in R.

[jimhester]

I guess this would work as a simple implementation. The naming code is a little weird though...

map_dfc <- function(.x, .f, ..., .sep = "") {
  .f <- as_function(.f, ...)
  res <- map(.x, .f)
  lens <- lengths(res)
  ret <- vector("list", sum(lens))
  i <- 1
  for (j in seq_along(res)) {
    for (k in seq_along(res[[j]])) {
      ret[[i]] <- res[[j]][[k]]
      i <- i + 1
    }
  }
  nms <- map(res, names)
  nms <- map(nms, ~ if (is.null(.)) "" else .)
  names(ret) <- unlist(map2(names(nms), nms, paste, sep = .sep))
  ret
}

map_dfc(mtcars, mean)
#> $mpg
#> [1] 20.09062
#> 
#> $cyl
#> [1] 6.1875
#> 
#> $disp
#> [1] 230.7219
#> 
#> $hp
#> [1] 146.6875
#> 
#> $drat
#> [1] 3.596563
#> 
#> ...
map_dfc(mtcars, ~ tibble::data_frame(mean = mean(.), median = median(.)), .sep = "_")
#> $mpg_mean
#> [1] 20.09062
#> 
#> $mpg_median
#> [1] 19.2
#> 
#> $cyl_mean
#> [1] 6.1875
#> 
#> $cyl_median
#> [1] 6
#> 
#> $disp_mean
#> [1] 230.7219
#> 
#> $disp_median
#> [1] 196.3
#> ...

[hadley]

@jimhester why not use bind_cols() for symmetry?