Problem with small values in Random.int on init

[PezzA]

Raised this question on stack overflow and it was suggested to create a bug report here.

http://stackoverflow.com/questions/37522369/elm-random-number-on-init-strange-behaviour

I've been working through the examples here:

http://guide.elm-lang.org/architecture/effects/random.html

And what I'm trying to do with the dice is to get to generate a D6 when then component is created.

module Components.DiceRoller exposing (Model, Msg, init, update, view)

import Html exposing (..)
import Html.App as Html
import Html.Attributes exposing (..)
import Html.Events exposing (..)
import Random
import String exposing (..)

main =
    Html.program
        { init = init
        , view = view
        , update = update
        , subscriptions = subscriptions
        }



-- MODEL


type alias Model =
    { dieFace : Int
    }


init : ( Model, Cmd Msg )
init =
    ( Model 0, (Random.generate NewFace (Random.int 1 6)) )



-- UPDATE


type Msg
    = NewFace Int


update : Msg -> Model -> ( Model, Cmd Msg )
update msg model =
    case msg of
        NewFace newFace ->
            ( Model newFace, Cmd.none )



-- SUBSCRIPTIONS


subscriptions : Model -> Sub Msg
subscriptions model =
    Sub.none



-- VIEW


dieFaceImage : Int -> String
dieFaceImage dieFace =
    concat [ "/src/img/40px-Dice-", (toString dieFace), ".svg.png" ]


view : Model -> Html Msg
view model =
    let
        imagePath =
            dieFaceImage model.dieFace
    in
        div []
            [ img [ src imagePath ] []
            , span [] [ text imagePath ]
            ]

However if I use 1 6 as the random min max, I only get 1 different value per minute. This only happens on init, it's fine in the orig example where it a UI action.

[mariosangiorgio]

As I mentioned on the StackOverflow answer, this behaviour is a bit odd but I am not sure if that's expected or not. What is happening is that, for this combination of parameters, there are quite big intervals of seeds that generate the same first value of the pseudo-random sequence.

It's worth noting that, if we exclude the first value, the sequence of the numbers produced look random enough.

I had another look at the implementation of the random number generator, comparing it both to the Haskell implementation and to the original paper. I found that one of the magic numbers might have a wrong value.

We have magicNum7 = 2137383399 but I think it should be magicNum7 = 2147483399. Updating it is not affecting this particular issue, but it's still worth amending the constant to the correct value.

[mariosangiorgio]

As I mentioned on the StackOverflow answer, this behaviour is a bit odd but I am not sure if that's expected or not. What is happening is that, for this combination of parameters, there are quite big intervals of seeds that generate the same first value of the pseudo-random sequence.

It's worth noting that, if we exclude the first value, the sequence of the numbers produced look random enough.

I had another look at the implementation of the random number generator, comparing it both to the Haskell implementation and to the original paper. I found that one of the magic numbers might have a wrong value.

We have magicNum7 = 2137383399 but I think it should be magicNum7 = 2147483399. Updating it is not affecting this particular issue, but it's still worth amending the constant to the correct value.

[jvoigtlaender]

@PezzA, @mariosangiorgio, in case you are still wondering what the issue here is: It's not the "magic numbers" thing, it's the fact that the randome generator is seeded from the current time at program start. See https://github.com/elm-lang/core/issues/673 for more explanation.

[jvoigtlaender]

@PezzA, @mariosangiorgio, in case you are still wondering what the issue here is: It's not the "magic numbers" thing, it's the fact that the randome generator is seeded from the current time at program start. See https://github.com/elm-lang/core/issues/673 for more explanation.

[evancz]

Not sure exactly what's wrong here, but maybe @mgold does. In any case, added this to #724 for tracking anything related to Random.

[evancz]

Not sure exactly what's wrong here, but maybe @mgold does. In any case, added this to #724 for tracking anything related to Random.

[mgold]

This will almost certainly be resolved by the new RNG.

[mgold]

This will almost certainly be resolved by the new RNG.

[connec]

I also encountered this working through the Random section in the guide.

I think it's interesting (though probably not valuable) to observe that if "Initialize the dice to a random value" was added as an exercise, it would probably have to be removed or else it would generate a lot of confusion.

[connec]

I also encountered this working through the Random section in the guide.

I think it's interesting (though probably not valuable) to observe that if "Initialize the dice to a random value" was added as an exercise, it would probably have to be removed or else it would generate a lot of confusion.

[seethroughtrees]

I've encountered this same issue. It seems if I use Random.Int with 3 or less. I will get repeated numbers in the same fashion as seen above. Anything over 3, and I get random as expected.

Like the above, this only occurs when triggered on init. Is there a workaround here?

I can create and provide an example to reproduce or versions if necessary.

[seethroughtrees]

I've encountered this same issue. It seems if I use Random.Int with 3 or less. I will get repeated numbers in the same fashion as seen above. Anything over 3, and I get random as expected.

Like the above, this only occurs when triggered on init. Is there a workaround here?

I can create and provide an example to reproduce or versions if necessary.

[lydell]

@seethroughtrees http://package.elm-lang.org/packages/mgold/elm-random-pcg/5.0.0/Random-Pcg

[lydell]

@seethroughtrees http://package.elm-lang.org/packages/mgold/elm-random-pcg/5.0.0/Random-Pcg

[mgold]

BTW that new RNG is #778. Until it's merged and released, use my library linked just above.

[mgold]

BTW that new RNG is #778. Until it's merged and released, use my library linked just above.

[seethroughtrees]

Perfect. Thanks for the update @lydell and @mgold . Good to have that link in this thread too.

[seethroughtrees]

Perfect. Thanks for the update @lydell and @mgold . Good to have that link in this thread too.