Comment by Alex Johnson (JIRA)

I see this as well using Swift 2.2 (Jan 6th snapshot). It looks like something has changed in the way that nil is stored. It seems to be converted to an instance of Any, which is confusing to me because that shouldn't be possible.

Compare the result of the dictionary creation in Swift 2.2:

  1> let d: [String:Any] = ["foo": (nil as Int?)]
d: [String : Any] = 1 key/value pair {
  [0] = {
    key = "foo"
    value = {
      payload_data_0 = 0x0000000000000000
      payload_data_1 = 0x0000000100602701
      payload_data_2 = 0x0005000000000000
      instance_type = 0x00000001004f5028
    }
  }
}

... to the result in Swift 2.1:

  1> let d: [String:Any] = ["foo": (nil as Int?)]
d: [String : Any] = 1 key/value pair {
  [0] = {
    key = "foo"
    value = nil
  }
}

And compare the result of accessing the value in Swift 2.2:

  3> let v = d["foo"]!
v: Any = {
  payload_data_0 = 0x0000000000000000
  payload_data_1 = 0x0000000000000001
  payload_data_2 = 0x0000000000000000
  instance_type = 0x00000001004f5028
}

... to the result in Swift 2.1:

  3> let v = d["foo"]!
v: Int? = nil

Is this resolved, or no longer valid? If either, this should be closed.

The coercion in the original example is diagnosed correctly—you cannot coerce from Any to a concrete type. Even though the output differs, nothing seems to suggest that there was an issue with downcasting the value instead.

The output is back to normal now, but I wonder: was it a mirroring issue or something else, and how do we test for the expected output? @lorentey Would be most appreciative of any tips.

Use as? or as! for runtime dynamic type checking.

@tbkka Sorry, I should have expressed myself more clearly. I am willing to add a regression test before closing this and was asking where to look at (which 'print' or 'dump' function does the REPL call?).

Unfortunately, I'm not that familiar with how REPL testing works. @gottesmm , can you give any pointers?

The output is back to normal now, but I wonder: was it a mirroring issue or something else, and how do we test for the expected output? @lorentey Would be most appreciative of any tips.

It looks like that would have been an LLDB data formatting bug.

REPL tests are here: https://github.com/apple/llvm-project/tree/next/lldb/test/Shell/SwiftREPL