The main change to dplyr is that this reverts most of the changes made in #5334 to support #5326. This previously iterated over data frame input column by column, but we now fully embrace the vctrs definition of a missing value, which is when the entire row is empty for data frames. Since data frame input is probably pretty rare, we are optimizing for the simplicity of the coalesce() implementation over supporting that feature. Users that need this can just map2() over their data frames, calling coalesce() on each column.
Also:
- Now uses
vec_case_when()internally - New
.ptypeand.sizearguments NULLinputs are now dropped (currently they cause an obscure error)
"1b1dd312-a69e-4d4c-b094-289700c3c5f6"