Should pluck<- actually be chuck<-?

[DavisVaughan]

I am not sure that pluck() and pluck<-() are consistent. pluck() returns NULL if the element does not exist, but pluck<-() throws an error. In that way, pluck<-() feels more like the counterpart to chuck().

I am not sure that pluck<-() should even exist, since assign_in() is very strict and requires that the plucked-location must exist. I am just arguing that is should be named chuck<-() instead.

library(purrr)

pluck(mtcars, "y")
#> NULL

chuck(mtcars, "y")
#> Error: Can't find name `y` in vector

# should this be `chuck<-`()?
pluck(mtcars, "y") <- 1 
#> Error: Can't find name `y` in vector

# this should be `pluck<-`() I think
mtcars$y <- 1

# what I think should be `chuck<-`() uses chuck() already
`pluck<-`
#> function (.x, ..., value) 
#> {
#>     assign_in(.x, list2(...), value)
#> }
#> <bytecode: 0x7f9558ae34e8>
#> <environment: namespace:purrr>
purrr:::assign_in
#> function (x, where, value) 
#> {
#>     chuck(x, !!!where)
#>     call <- reduce_subset_call(quote(x), as.list(where))
#>     call <- call("<-", call, value)
#>     eval_bare(call)
#>     x
#> }
#> <bytecode: 0x7f9558ae2790>
#> <environment: namespace:purrr>

^{Created on 2019-02-04 by the reprex package (v0.2.1.9000)}

[lionel-]

We considered this and thought the interface would be simpler overall this way because casual users wouldn't need to learn the difference between the two and remember that they have to use chuck<- rather than pluck<-. Note that a lenient pluck<- isn't possible because there is no type information for non-existing locations.

[lionel-]

Just got some feedback from @jennybc about modify_in() and assign_in().

• modify_in() might need an option to ignore non-existing locations instead of failing.
• assign_in() might need an option to force assignment in non-existing locations (using lists for deep assignments where there is no type information? Or error in that case?).

The latter suggests we should have both pluck<- and chuck<-.

Also would be useful to have a can_pluck() predicate.

cc @hadley

[lionel-]

From @hadley:can_pluck() could be has_element()