Debugger and anonymous functions, purrr::map()

[MattP]

I'm wondering if there is a way to jump into anonymous functions when debugging them in, say, a purrr::map() call. Essentially, I get some errors when developing a function that, inside, uses some anonymous function arguments to map. I would like to do a call to debugonce() or similar to step into the anonymous function, inspect its values, and figure out what is wrong. I cannot figure out how to step into here -- I get into the guts of the purrr::map() implementation, not the anonymous function.

One solution is to pepper the anonymous code with browser() calls, and that works, but I am looking for something that can invoke the debugger without altering the code...
Advice appreciated.

library(dplyr)
library(purrr)

#;; I can run the interactive debugger if I name all of the functions
#;; Along the way
v <- 1:3
innerFunc <- function(i) {
    if (i == 2) stop()
    i
}

outerFunc <- function(vvec) {
    map(vvec, innerFunc)
}

#;; This allows me to step into the inner function at every invocation, and
#;; I can inspect the value of `i` before it stops.
debug(innerFunc)
outerFunc(v)



#;;;;;;;
#;; However, I don't know how to step into the inner function if I define it
#;; anonymously.  Is there a way?
outerFunc <- function(vvec) {
    map(vvec, ~{
        if (.x == 2) stop()
        .x
    })
}

#;; How do I step into the inner anonymous function via the debugger?
#;; This leads me down a rabbit-hole of map implementation guts.
debugonce(outerFunc)
outerFunc(v)

[michaelbgarcia]

How about setting options(error = recover). After running your function, select the third option and it brings you into the inner function at the point of failure.

library(dplyr)
library(purrr)

options(error = recover)
v <- 1:3
outerFunc <- function(vvec) {
  map(vvec, ~{
    if (.x == 2) stop()
    .x
  })
}

outerFunc(v)

[nirgrahamuk]

For my 2cents, anonymous functions are useful only when taking the time to think of a name for a function is more trouble than its worth, a simple function which is unlikely to fail and does something relatively trivial is one thing, but I think that from a programming best practice point of view, when you find that you need to debug a function that you had previously in-lined anonymously, this should potentially signal a warning to you, that the function might be worthwhile naming after all.