How to use namespaced function with dplyr::mutate_each?

Question

I am trying to use dplyr::mutate_each with some external functions without attaching actual libraries

dplyr::tbl_df(iris) %>% 
    dplyr::mutate_each(dplyr::funs(stringi::stri_trim_both))

but it fails with following error:

Error: unsupported type for column 'Sepal.Length' (CLOSXP, classes = function)

When I use data.table instead of data.frame:

Error in `[.data.table`(`_dt`, , `:=`(Sepal.Length, stringi::stri_trim_both), : RHS of assignment is not NULL, not an an atomic vector (see ?is.atomic) and not a list column.

If I use local variable as below everything works as expected.

trim_both <-  stringi::stri_trim_both
dplyr::tbl_df(iris) %>% dplyr::mutate_each(dplyr::funs(trim_both))

It is not an optimal solution but I can live with that. Nevertheless I would be grateful for an explanation what is the source of the problem.

Session info:

R version 3.1.1 (2014-07-10)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] dplyr_0.4.1

loaded via a namespace (and not attached):
[1] assertthat_0.1       DBI_0.3.1            lazyeval_0.1.10.9000
[4] magrittr_1.5         parallel_3.1.1       Rcpp_0.11.4         
[7] stringi_0.4-1        tools_3.1.1         

Note: This problem no longer occurs in dplyr 0.7.2.

Answer 1

The underlying reason is that dplyr::funs_ calls dplyr:::make_call. And dplyr:::make_call differentiates between cases using the class of the object generated by lazyeval::lazy_dots.

class(lazyeval::lazy_dots(trim_both)[[1]]$expr)
## "name"
class(lazyeval::lazy_dots(stringi::stri_trim_both)[[1]]$expr)
## "call"

See the function my_funs below for a solution to this. I have not tested this in any detail and I am sure that there is a reason that this was different in dplyr, so do not use this as a default. It's mostly meant to clarify the problem

# calling my_funs_ (instead of funs_)
my_funs <- function (...) 
  my_funs_(lazyeval::lazy_dots(...))

my_funs_ <- function(dots){
  dots <- lazyeval::as.lazy_dots(dots)
  env <- lazyeval::common_env(dots)
  names(dots) <- dplyr:::names2(dots)
  # difference here
  dots[] <- lapply(dots, function(x) {
    if (is.character(x$expr)) {
      x$expr <- substitute(f(.), list(f = as.name(x$expr)))
    }
    else if (is.name(x$expr)) {
      x$expr <- substitute(f(.), list(f = x$expr))
    }
    else if (is.call(x$expr)) {
      x$expr <- substitute(f(.), list(f = x$expr)) #### this line was different
      # originally x$expr <- x$expr
    }
    else {
      stop("Unknown inputs")
    }
    x
  })
  missing_names <- names(dots) == ""
  ### this is also different 
  default_names <- vapply(dots[missing_names], function(x) as.character(x)[1], 
                          character(1))
  ## originally dplyr:::make_name(x) instead of as.character(x)[1]
  names(dots)[missing_names] <- default_names
  class(dots) <- c("fun_list", "lazy_dots")
  dots
}

dplyr::tbl_df(iris) %>% 
  dplyr::mutate_each(my_funs(stringi::stri_trim_both))