In tally_ and count_, support programmatic specification of wt

[dholstius]

Passing the name(s) of variables (as character) "just works" in most underscored ("programmatic") dplyr functions. For example:

tally(group_by_(my_data, "foo"), wt = bar)

But, it looks like this has yet to be supported in tally_ (and hence, also in count_).

Proposed: these should also work:

tally_(my_data, wt = "bar")
count_(my_data, vars = "foo", wt = "bar")

Happy to construct test cases, pull request, etc. but not sure if this might be intended behavior — or if/how it intersects with the intended behavior of lazyeval::interp (relied upon by tally_).

[krlmlr]

@hadley: I agree that this is inconsistent, and wt should accept both character and formula in tally_(). The inconsistencies are even more severe in tidyr.

[hadley]

@krlmlr yes, I think we'll definitely want to take another pass at this for the next release (using everything we've learned in lazyeval etc)

[hadley]

@lionel- another one for you

[lionel-]

Should this be tally(my_data, wt = !! symbol("bar"))?

@hadley what's the rule for which verbs should support strings? group_by() supports strings but select() does not.

[hadley]

Yes, I think so.

I think generally verbs shouldn't support strings without some syntactic signal. select() has one_of(), and with tidyeval we can now use !!.

[lionel-]

sounds good, closing this then.