Should add_count() be generic?

[mgirlich]

add_count() is not generic in contrast to count(). As add_count() uses mostly exported dplyr verbs it almost works correctly in dbplyr and dtplyr. The issue is that the resulting table is grouped.

There are basically two solutions to this:

1. make add_count() generic or
2. Remove the check for is.data.frame(x) in add_count():

# last lines of `add_count()`
if (is.data.frame(x)) {
    out <- dplyr_reconstruct(out, x)
}
out

Then it would only be necessary to implement dplyr_reconstruct() for tbl_lazy in dbplyr and dtplyr_step in dtplyr