dplyr::distinct appears to consider empty rows different

[JohnMount]

dplyr::distinct() appears to consider empty rows different. Notice in the example below dplyr::distinct() returns a 2 row data frame where both rows are identical. This is a corner-case where there are no columns, but I think in this case dplyr::distinct() should not return more than 1 row in this case. Notice adding a column in the example then decreases the number of rows considered distinct.

suppressPackageStartupMessages(library("dplyr"))
packageVersion("dplyr")
#> [1] '0.7.1.9000'

d <- data.frame(x= c(1, 1))

d0 <- select(d, one_of(character(0)))
dD <- distinct(d0)
print(dD)
#> data frame with 0 columns and 2 rows

d2 <- mutate(dD, newCol = 1)
print(d2)
#>   newCol
#> 1      1
#> 2      1

distinct(d2)
#>   newCol
#> 1      1

[krlmlr]

Thanks. Honestly, I'd rather throw an error here, distinct() on a data frame without columns smells like division by zero to me. @hadley?

[JohnMount]

Error out is not a bad idea. I consider this enough of a dangerous corner case that I already re-coded my application check for this situation and not call distinct() in this situation.

[hadley]

I think if there are no columns, we could just special case distinct() to return the first row.

[romainfrancois]

just returning 1 row feels less surprising than error. I'll do that.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/