Runtime error on certain recursive random-number generators

[alex-lew]

This issue can be replicated in Chrome and Safari on Mac OS X in elm-lang.org/try.

import Html exposing (Html, div, text, button)
import Html.Events exposing (onClick)
import Random


-- PROBLEM

rollIf : Bool -> Random.Generator Int
rollIf b =
  if b then
    Random.int 1 6
  else
    tryAgain

rollAgain : Random.Generator Int
rollAgain =
  roll

roll : Random.Generator Int
roll =
  Random.bool
    |> Random.andThen rollIf


-- SUPPORT CODE

main =
  Html.program
    { init = (0, Cmd.none)
    , update = update
    , view = view
    , subscriptions = always Sub.none
    }

view model =
  div []
    [ text (toString model)
    , button [onClick Generate] [text "Generate!"]
    ]

type Msg = Generate | NewValue Int

update msg model =
  case msg of
    Generate ->
      (model, Random.generate NewValue roll)

    NewValue v ->
      (v, Cmd.none)

Clicking "Generate" a few times will cause the following runtime error:

TypeError: undefined is not an object (evaluating '_p59._0')

If the definition of rollAgain is removed in favor of using roll directly, the runtime error disappears and the code works as expected.

[process-bot]

Thanks for the issue! Make sure it satisfies this checklist. My human colleagues will appreciate it!

Here is what to expect next, and if anyone wants to comment, keep these things in mind.

[alex-lew]

(This might be the same as #1527?)

[jvoigtlaender]

That or, given the specifics of your instance, actually a duplicate of #1516.

[alex-lew]

Thanks -- hadn't seen that! In this case, reordering the definitions in Elm doesn't help, so it might be different?

[jvoigtlaender]

Maybe. But the simplified example you added to #1527 is very similar to what my SSCCE in #1516 came from. In any case, the underlying problem is the same in all three issues, and they would almost certainly be fixed by the same compiler change. Which also means they are unlikely to be fixed anytime soon, given Evan's statement in #1516.

[jvoigtlaender]

To be sure, you could take a look at the JS produced for your example and see whether it has the "order problem".

[evancz]

Added this to #1377 so it can be addressed alongside a bunch of similar issues.

Thanks for the report!

[evancz]

Alright, I got this working in my local builds! Thanks for the report. This one was particularly helpful in figuring out a general solution!