Behavior of *_at() functions with unnamed input

[echasnovski]

After using *_at() functions, I feel like their behavior is a little inconsistent. In documentation it says "modify_at() only modifies elements given by names or positions." and "Only those elements corresponding to .at will be modified." I read this as: try finding elements specified in .at and modify them, while not touching anything else.

When .at is a string, it works exactly like that when at least one list element is named. However, it throws error when there are no names, in which case I'd expect it to not find any .at elements and modify nothing while returning the input.

A reprex:

library(purrr)

map_at(list(1, a = 2), .at = "a", ~. + 1)
#> [[1]]
#> [1] 1
#> 
#> $a
#> [1] 3
map_at(list(1, b = 2), .at = "a", ~. + 1)
#> [[1]]
#> [1] 1
#> 
#> $b
#> [1] 2

map_at(list(1, 2), .at = "a", ~. + 1)
#> Error: character indexing requires a named object
modify_at(list(1, 2), .at = "a", ~. + 1)
#> Error: character indexing requires a named object

^{Created on 2019-05-25 by the reprex package (v0.3.0)}

[hadley]

I think we will deprecate/superseded the _at functions (#874) so we no longer to plan this to fix this.