x[1,,drop = TRUE] should return a list

[patperry]

Currently tibble behavior with drop = TRUE is consistent with data.frame for extracting a single column, but not for extracting a single row:

library(tibble)
as_tibble(mtcars)[1, , drop = TRUE]
#> # A tibble: 1 x 11
#>     mpg   cyl  disp    hp  drat    wt  qsec    vs    am  gear  carb
#>   <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1  21.0  6.00   160   110  3.90  2.62  16.5     0  1.00  4.00  4.00
mtcars[1, , drop = TRUE]
#> $mpg
#> [1] 21
#> 
#> $cyl
#> [1] 6
#> 
#> $disp
#> [1] 160
#> 
#> $hp
#> [1] 110
#> 
#> $drat
#> [1] 3.9
#> 
#> $wt
#> [1] 2.62
#> 
#> $qsec
#> [1] 16.46
#> 
#> $vs
#> [1] 0
#> 
#> $am
#> [1] 1
#> 
#> $gear
#> [1] 4
#> 
#> $carb
#> [1] 4

The data.frame behavior makes more sense to me here, because it fits with the general pattern that if x is a matrix-like object, then x[1,,drop = TRUE] returns a vector-like object.

Related: #311

[krlmlr]

Thanks, agreed. We're supporting the drop argument for consistency with packages that expect data frames, and we should also support this case.

[patperry]

I think you have some leeway here for what to do for the following cases:

x <- mtcars[1, ]
x[, , drop = TRUE]

y <- mtcars[, 1, drop = FALSE]
y[, , drop = TRUE]

z <- mtcars[1, 1, drop = FALSE]
z[, , drop = TRUE]

A data.frame returns a list in the first case and a vector in the second and third.

In my own implementation in the frame package, (https://github.com/patperry/r-frame ) I chose to break with the data.frame behavior and only drop dimensions with non-missing indices. This makes the results have the same type in all three cases above (x[, , drop = TRUE] returns x, with similar behavior for y and z).

Probably no one actually uses this construct, so do whatever makes the implementation easiest.

[patperry]

Following up, this is the corner case that actually matters:

x <- data.frame(foo = 31337)
x[1, , drop = TRUE]

A data.frame returns a 31337 here, but I think it's more natural to return list(foo = 31337).

[krlmlr]

I think users should always use drop = FALSE. We're not going to change drop = TRUE behavior for tibble, this is really only for data.frame compatibility

[krlmlr]

This seems to only require splitting the attributes from the resulting list (including "class").

[krlmlr]

Closed in 35dfee5.