unexpected output from aggregate

Question

While experimenting with aggregate for another question here, I encountered a rather strange result. I\'m unable to figure out why and am wondering if what I\'m doi

Answer 1

The issue here is how aggregate.data.frame() determines the groups.

In aggregate.data.frame() there is a loop which forms the grouping variable grp. In that loop, grp is altered/updated via:

grp <- grp * nlevels(ind) + (as.integer(ind) - 1L)

The problem with your example if that once by is converted to factors, and the loop has gone over all of these factors, in your example grp ends up being:

Browse[2]> grp
[1] Inf Inf Inf Inf

Essentially the looping update pushed the values of grp to a number indistinguishable from Inf.

Having done that, aggregate.data.frame() later does this

y <- y[match(sort(unique(grp)), grp, 0L), , drop = FALSE]

and this is where the earlier problem now manifests itself as

dim(y[match(sort(unique(grp)), grp, 0L), , drop = FALSE])

because

match(sort(unique(grp)), grp, 0L)

clearly returns just 1:

> match(sort(unique(grp)), grp, 0L)
[1] 1

as there is only one unique value of grp.