Dropping a column from a data.frame causes unwanted loss of an attribute

Question

I want to drop a column from a data.frame. But when I do, an attribute of the data.frame is lost, which I don't want. First the setup:

d <- data.frame(a = 1, b = 2, c = 3)
attr(d, "test_attribute") <- "something"
d2 <- d
d
#>   a b c
#> 1 1 2 3

The test attribute is present:

attributes(d2) # contains $test_attribute [1] "something"

Now I want to remove the second column - but most ways destroy that attribute:

attributes(d2[, -2]) # it's gone
attributes(dplyr::select(d2, -2)) # it's gone

I found one way to preserve it:

d3 <- d2
d3[2] <- NULL
attributes(d3)

Why does the test_attribute get dropped in the first two cases, but not when using this last method?

Answer 1

?Extract describes this behavior.

Subsetting (except by an empty index) will drop all attributes except names, dim and dimnames.