Deep clone does not lead to a complete copy

[jakob-r]

Consider the following example

library(R6)

FooClass = R6Class(
  "Foo",
  public = list(
    initialize = function(value) {
      # force(value) # does not help
      self$foo.fun = function(x) x == value
    },
    foo.fun = NULL
  )
)

foo.obj1 = FooClass$new(3)
foo.obj1$foo.fun(3)
# TRUE

foo.obj2 = foo.obj1$clone()
foo.obj2$foo.fun(3)
# value not found

foo.obj3 = foo.obj1$clone(deep = TRUE)
foo.obj3$foo.fun(3)
# value not found

Or is this expected to fail?

[wch]

That is the expected behavior - the clone() method assumes that any function in the object (like foo.fun in this case) is a method, and when it makes the copy of that function in the clone object, it reassigns the function's environment so that it can find the appropriate self object.

Your version of the foo.fun captures the environment from the initialize method, but that environment gets lost in the cloning process. In this particular case, you'd be better off storing value in self or private instead of capturing it with a closure, but that may or may not be appropriate for your real use case.

This is essentially the same issue as #94.

I think this the documentation should probably do a better job explaining how cloning works.

[jakob-r]

Thanks for your comprehensible answer. That solves the issue for me.