R: Using pipe %>% and pkg::fo leads to error "Error in .::base : unused argument"

Question

I am using magrittr's pipe %>%, followed immediately by a function called by: package::function, and get the error: Error in .::base : unused argument (mean)

What is the problem?

library(magrittr)
c(1,2) %>%
  base::mean
#> Error in .::base: unused argument (mean)

Answer 1

What's happening is that magrittr is getting confused as to exactly what function you want to insert the previous value into. When you do just

c(1,2) %>%
  mean

magrittr is able to easily see that mean is a symbol that points to the mean function. But when you do base::mean, things get a bit trickier because :: is also a function in R. Let's compare the difference between base::mean and base::mean() in R in terms of how they are translated into function calls.

as.list(quote(base::mean))
# [[1]]
# `::`    
# [[2]]
# base    
# [[3]]
# mean

as.list(quote(base::mean()))
#  [[1]]
# base::mean

You can see these parse differently. When you just type base::mean R will see the :: function first and will try to pass in the numbers there. Basically to's trying to call

`::`(., base, mean)

which doesn't make sense can that's what gives you that specific error message

But if you explicitly add the (), R can see that you are trying to call the function returned from base::mean so it will add the parameter to the right place. So you can do

c(1,2) %>%
  base::mean()

or

c(1,2) %>%
    (base::mean)

Answer 2

Note that there is a version of magrittr that does not have this issue but it hasn't been pushed to CRAN for a very long time now.

As documented in this issue, github version of magrittr pipes can successfully deal with

c(1,2) %>%
    base::mean

[1] 1.5

This was fixed 4 years ago but never made it to CRAN. Since most people would be using the CRAN version, I would not suggest writing code that depends on this fix.