Consider allowing f in epi[x]_slide to be an unnamed data masking expression yielding a tibble

[brookslogan]

This would improve compatibility with dplyr. However, it may introduce some ambiguities with the function&formula f case, and require some modifications to exactly how these cases are evaluated: if there is an f but no ..., we can't immediately distinguish between functions, formulas and data masking expressions; e.g., just using class won't work as it would force data masking expressions. Instead, we would need to (defuse and) evaluate f itself using data masking, then check the result's class: functions and formulas would need to be (validated and) evaluated in the normal fashion, while tibbles would be unpacked into multiple columns as in mutate and summarize. For consistency, f may also need to be evaluated in this way even if args are provided in ..., but we should only allow args in ... if it's actually a function (or formula?).

Generalizing f in this way may make it harder to validate f, and may make it impossible to immediately validate f without access to any data. Thus, whether or not to proceed may require some thought.

[dshemetov]

This is addressed in #477 right? @brookslogan

[brookslogan]

Yes. We side-stepped the tricky implementation here by just requiring the user add an extra comma... perhaps not ideal, but these seem like power-user features anyway and they can deal with it.