Default for top_n

[romainfrancois]

Currently the default behavior of top_n is to use the last variable when none is specified. Just wondering if it would perhaps make more sense to use the indices.

For example:

iris %.% group_by(Species) %.% top_n(2)

will use the last variable Species. Of course, within each group, all values for Species are equal, so this will give us the entire data.

So I'm wondering if iris %.% group_by(Species) %.% top_n(2) should express "the two first rows in each group"

[hadley]

I think this probably should be the job for another function, since it's not quite the regular defintion of top_n(). I was thinking of building slice() out of integer_filter().

[romainfrancois]

About slice, i'd like to avoid materialising the indices, which should not be too hard if we define the perimeter allowed by the arguments.