[FR] rowwise semantics for flattening data.frame

[yutannihilation]

At first, I thought the behaviour of flatten_dfr() reported on #374 was a bug but it's not. As a data.frame is basically a list, flatten() transforms a list of data.frames to a list:

reprex::reprex_info()
#> Created by the reprex package v0.1.1.9000 on 2017-10-31

library(purrr)
x <- data.frame(a = 1:2, b= 10:11)
flatten(list(x, x))
#> $a
#> [1] 1 2
#> 
#> $b
#> [1] 10 11
#> 
#> $a
#> [1] 1 2
#> 
#> $b
#> [1] 10 11

and then converts this to a data.frame by bind_rows(). So, theoretically, there is nothing weird that the columns will be overwritten by the ones of the same name.

flatten_dfr(list(x, x))
#> # A tibble: 2 x 2
#>       a     b
#>   <int> <int>
#> 1     1    10
#> 2     2    11

That said, I feel this is counter-intuitive; considering that a list of multiple vectors is flattened into one concatenated vector, we naturally expect a list of data.frames will be a data.frame whose content is a concatenation of the rows in the data.frames:

#> # A tibble: 4 x 2
#>       a     b
#>   <int> <int>
#> 1     1    10
#> 2     2    11
#> 3     1    10
#> 4     2    11

Is it possible to use rowwise semantics for data.frame? (If not, I think flatten_dfr() and flatten_dfc() should be removed to avoid confusion. In most cases, bind_rows() and bind_cols() are the ones that is needed actually.)

[hadley]

Duplicate of #374