Unify argument flexibility of both single value and named list of values

[DavisVaughan]

• Also remove NEWS bullet about moving towards ptype = list()

unchop()

• ptype = list()
  • Change to ptype = NULL (original behavior)
  • Use is_ptype() and wrap into a named list with names of cols if TRUE
  • Useful when you only specify 1 column to unchop, which is most of the time

unnest()

• ptype = list()
  • Inherits unchop behavior
  • Change to ptype = NULL (original behavior)

pivot_longer()

• values_ptypes = list() and names_ptypes = list()

  • Change to = NULL
  • Use is_ptype() and wrap into a named list with names of names_from or values_from

• values_transform = list() and names_transform = list()

  • Change to = NULL
  • Wrap if not a list, then map over list with as_function()

unnest_longer() / unnest_wider() / hoist()

• ptype = list()

• transform = list()

---

Already done, or not touching

pivot_wider()

• values_fill = NULL

  • Allows scalar value
  • Allows named list of scalar values

• values_fn = NULL / unused_fn = NULL

  • Allows single function
  • Allows named list of functions

replace_na() / complete()

• fill = list()
  • Not updating this one. This controls both the overall selection AND the replacement value. Allowing fill = 0 would fill all columns in the data frame with 0 where there were missing values, which means all columns had to be the same type, which is rare.

[DavisVaughan]

I think this will break dbplyr because it has a method for pivot_longer() that sets names_ptypes = list(), and that has a new meaning now. It looks like these arguments aren't supported in dtplyr, so it should be fine there. I don't think any of the other functions should have any issues.

We may have to temporarily make names_ptypes = list() and values_ptypes = list() in pivot_longer() be treated like NULL with a warning until dbplyr can update