Passing arbitrary arguments to multiple nested functions in R

Question

Can't find a question similar to what I am trying to do. I have a workhorse function that can take any of a class of custom-made sub-functions, with their respective arguments, and do certain processing on it, like the following toy examples:

func1 <- function(a, b, c) {a + b * c}

func2 <- function(x, y, z, bob, bleb) {(x / y)^bleb * z/bob}

workhorse <- function(m, n, o, FUN, ...) {
  result <- FUN(x, y, ...)
  [process result with m, n, and o]
}

By itself, this works perfectly well as intended, as in

foo <- workhorse(m, n, o, func1, a, b, c)
foo2 <- workhorse(m, n, o, func2, x, y, z, bob, bleb)

I am trying to create a wrapper for workhorse (meta_workhorse) that will accept a list of functions (by quoted name) to run through workhorse and combine the results, something like below.

FUN_list <- c('func1', 'func2')
ARGS_list <- list(c(a, b, c), c(x, y, z, bob, bleb))

meta_workhorse <- function(mm, nn, oo, FUN_list, ARGS_list) {
  meta_result <- sapply(1:length(FUN_list), function(x) {
    workhorse(mm, nn, oo, get(FUN_list[x]), ARGS_list[x], ...)
  }
}

The problem is that each of the sub-functions take different arguments, so I need to link the function name going into the get() function with the corresponding list of arguments in ARGS_list. I've tried everything I could think of, including do.calls, setting formals(workhorse), to no avail. How do I pass the arguments in this case?

EDIT: Nicola's answer solved most of the issue, but still have one subfuction (using a packaged function) keeps giving the error "unused argument ... = list(x, y)". Looking for some suggestions while I dig through previous ellipsis questions.

nicola · Accepted Answer · 2014-10-12 22:17:27Z

3

I guess you should declare your ARGS_list as a list of lists, instead of a normal list. Then, you can proceed to use do.call. An example (hoping that is what you are looking for):

   func1 <- function(a, b, c) {a + b * c}
   func2 <- function(x, y, z, bob, bleb) {(x / y)^bleb * z/bob}
   workhorse <- function(m, n, o, FUN, ...) {
     result <- FUN(...)
     result+m+n*o
   }
   FUN_list <- c('func1', 'func2')    
   ARGS_list <- list(list(a=1, b=2, c=3), list(x=4, y=5, z=6, bob=7, bleb=8))
   meta_workhorse <- function(mm, nn, oo, FUN_list, ARGS_list) {
      sapply(1:length(FUN_list), function(x) {
        do.call(workhorse,c(list(m=mm,n=nn,o=oo,FUN=get(FUN_list[x])),ARGS_list[[x]]))
      })
   }

   meta_workhorse(1,2,3,FUN_list,ARGS_list)
   # [1] 14.000000  7.143805

answered Oct 12, 2014 at 22:17

nicola

24.6k3 gold badges38 silver badges60 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

skwalas Over a year ago

Thank you, that provided part of the solution. Still did not work on the real code, then I figured out that it had to be passed as ... = ARGS_list[[x]], since it was being passed through the sub functions as the ellipses. Works just fine now.

skwalas Over a year ago

Spoke too soon. This works for most functions. I have one that repeatedly gives the following error: unused argument (... = list(n.trees = 20). This seems to imply that it cannot handle the arguments being passed as a list, in contrast to the other functions I'm working with. However, it still works if the arguments are passed directly (i.e. one function at a time). Is there a way to parse a list of arguments so that it looks like arguments being called directly? I'm stuck, again.

nicola Over a year ago

Provide as similar as possible prototypes of your sub-functions and try to isolate the error to make it reproducible. Then edit your question or (maybe better) ask another question.

skwalas Over a year ago

traced the issue to the specific packaged function that I used in my wrapper function (gbm::gbm.fit). I ended up extracting the dot arguments with list(...), then recreating a do.call specifically for the gbm.fit function listing all of the standard arguments plus the dot arguments. However,even this did not work until I modified it to unlist(list(...)) first. I am at a loss why gbm.fit would behave this way, but the original answer seems to work fine for other packaged functions such as e1071:svm and glmnet::glm. In any case, I seem to be back on track for now. thanks.

Collectives™ on Stack Overflow

Passing arbitrary arguments to multiple nested functions in R

1 Answer 1

4 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related