unique and duplicated's default value for 'by='

[arunsrinivasan]

I am not particularly happy with the current option of unique/duplicated having by = key(x) as the default, especially because there's no message or warning as to which columns are being used to compute unique values on.

Now that on= argument is also in place, and we are entertaining the idea of having to set keys less, maybe it's right time to lift this default value so that it's similar to base R? i.e., with cols = seq_along(x) - see #1283.

[franknarf1]

I agree.

(I'm guessing a "question" issue is soliciting opinions..? I've no opinion on the by->cols issue; I like by, but could live with cols. I have never written the cols arg of the other functions out, rather passing by position every time. In contrast, I always write out by=.)

[arunsrinivasan]

Meant it to be internals, sorry about that.

[MichaelChirico]

I agree--I rarely use the default value for by in my current usage--even if x is keyed, more often I do something equivalent to by=key(x)[1:2], i.e., only using a subset of the assigned keys.

[jangorecki]

that would also refer to uniqueN

[cderv]

Help documentation for these functions seems not to have changed. It still says that by = key(x) is used by default. I think we should update the documentation as it is not verry clear right now.

I discovered this reading through the NEWS for all changes in 1.9.8.