`[` mistakenly returns rows, not columns, when subsetting with i and specifying drop

[DavisVaughan]

I don't think this is intended behavior, and it is actually causing an issue for me in tibbletime. Here's the simple version.

---

In base R, specifying drop with a data frame is ignored and the column is returned.

# Iris column subsetting
head(iris[1])
#>   Sepal.Length
#> 1          5.1
#> 2          4.9
#> 3          4.7
#> 4          4.6
#> 5          5.0
#> 6          5.4

# Specifying drop still returns the column
head(iris[1, drop = FALSE])
#> Warning in `[.data.frame`(iris, 1, drop = FALSE): 'drop' argument will be
#> ignored
#>   Sepal.Length
#> 1          5.1
#> 2          4.9
#> 3          4.7
#> 4          4.6
#> 5          5.0
#> 6          5.4

---

In tibble, if drop is specified it is ignored but it treats i like you are wanting to subset by rows, not columns.

# iris tbl
library(tibble)
iris_tbl <- as_tibble(iris)

# This works as intended
iris_tbl[1]
#> # A tibble: 150 x 1
#>    Sepal.Length
#>           <dbl>
#>  1         5.10
#>  2         4.90
#>  3         4.70
#>  4         4.60
#>  5         5.00
#>  6         5.40
#>  7         4.60
#>  8         5.00
#>  9         4.40
#> 10         4.90
#> # ... with 140 more rows

# This returns rows not columns
iris_tbl[1, drop = FALSE]
#> # A tibble: 1 x 5
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#>          <dbl>       <dbl>        <dbl>       <dbl> <fctr> 
#> 1         5.10        3.50         1.40       0.200 setosa

# Same thing but with two columns and drop=TRUE
iris_tbl[c(1,3)]
#> # A tibble: 150 x 2
#>    Sepal.Length Petal.Length
#>           <dbl>        <dbl>
#>  1         5.10         1.40
#>  2         4.90         1.40
#>  3         4.70         1.30
#>  4         4.60         1.50
#>  5         5.00         1.40
#>  6         5.40         1.70
#>  7         4.60         1.40
#>  8         5.00         1.50
#>  9         4.40         1.40
#> 10         4.90         1.50
#> # ... with 140 more rows

# Correctly displays warning, but returns rows
iris_tbl[c(1,3), drop = TRUE]
#> Warning: drop ignored
#> # A tibble: 2 x 5
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#>          <dbl>       <dbl>        <dbl>       <dbl> <fctr> 
#> 1         5.10        3.50         1.40       0.200 setosa 
#> 2         4.70        3.20         1.30       0.200 setosa

---

The issue seems to be coming from the use of if (nargs() <= 2L) in tbl-df.R. When drop is specified, then the number of args becomes 3. Is there another way to check for what you want here? It seems like you just want something like if( !missing(i) & missing(j) ) but I don't think that is a drop in replacement by itself.

[DavisVaughan]

Here is an easy solution. The problem really stems from that fact that when drop is specified, nargs() counts an extra argument. You really don't care if drop is specified or not, so just ignore drop as an argument.

  # Ignore drop as an argument
  .nargs <- if(!missing(drop)) {
    nargs() - 1
  } else {
    nargs()
  }

# Escape early if nargs() == 2L; ie, column subsetting
if (.nargs <= 2L) {
   ...
}

---

With this implemented, all tests still pass and the following works:

library(tibble)
iris_tbl <- as_tibble(iris)

# This works as intended
iris_tbl[1]
#> # A tibble: 150 x 1
#>    Sepal.Length
#>           <dbl>
#>  1         5.10
#>  2         4.90
#>  3         4.70
#>  4         4.60
#>  5         5.00
#>  6         5.40
#>  7         4.60
#>  8         5.00
#>  9         4.40
#> 10         4.90
#> # ... with 140 more rows

# drop is ignored, so .nargs = 2
iris_tbl[1, drop = FALSE]
#> # A tibble: 150 x 1
#>    Sepal.Length
#>           <dbl>
#>  1         5.10
#>  2         4.90
#>  3         4.70
#>  4         4.60
#>  5         5.00
#>  6         5.40
#>  7         4.60
#>  8         5.00
#>  9         4.40
#> 10         4.90
#> # ... with 140 more rows

# drop is ignored, so .nargs = 3
# same behavior as base so this is good
iris_tbl[1, , drop = FALSE]
#> # A tibble: 1 x 5
#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#>          <dbl>       <dbl>        <dbl>       <dbl> <fctr> 
#> 1         5.10        3.50         1.40       0.200 setosa

# Same thing but with two columns and drop=TRUE
iris_tbl[c(1,3)]
#> # A tibble: 150 x 2
#>    Sepal.Length Petal.Length
#>           <dbl>        <dbl>
#>  1         5.10         1.40
#>  2         4.90         1.40
#>  3         4.70         1.30
#>  4         4.60         1.50
#>  5         5.00         1.40
#>  6         5.40         1.70
#>  7         4.60         1.40
#>  8         5.00         1.50
#>  9         4.40         1.40
#> 10         4.90         1.50
#> # ... with 140 more rows

# Correctly displays warning, and now returns column
iris_tbl[c(1,3), drop = TRUE]
#> Warning: drop ignored
#> # A tibble: 150 x 2
#>    Sepal.Length Petal.Length
#>           <dbl>        <dbl>
#>  1         5.10         1.40
#>  2         4.90         1.40
#>  3         4.70         1.30
#>  4         4.60         1.50
#>  5         5.00         1.40
#>  6         5.40         1.70
#>  7         4.60         1.40
#>  8         5.00         1.50
#>  9         4.40         1.40
#> 10         4.90         1.50
#> # ... with 140 more rows

---

Or, more compactly,

# Ignore drop as an argument
  .nargs <- nargs() - !missing(drop)

[krlmlr]

Thanks for raising this, and for suggesting a solution.