Initial command obliterates previous state

[mattrrichard]

On start up, my app randomizes its state as part of its init command:

main =
    let 
        startupCmd = 
            Random.generate NewModel modelGenerator
    in
        Html.program
            { init = ( emptyModel, startupCmd )
            , update = update 
            , view = view
            , subscriptions = subscriptions
            }

The problem here is that its NewModel message ticks after the app's previous state is injected.

Is this something that can be worked around by elm-hot-reload? If not, what about a way to signal to the app that it has been hot reloaded? That way I could simply not issue the startup command in those cases. e.g. :

main = 
    let
        startupCmd = 
            Random.generate NewModel modelGenerator

        init flags =
            if flags.hotSwapped then
                ( emptyModel, Cmd.none )
            else
                ( emptyModel, startupCmd )
    in
        Html.programWithFlags
            { init = init
            , update = update startupCmd discoCmd
            , view = view
            , subscriptions = subscriptions
            }

[fluxxu]

Hi,

I added this and published it to npm:
https://github.com/fluxxu/elm-hot-loader/blob/master/hmr.js#L164

Do you think this can solve your problem?

[mattrrichard]

I think something close to that would work, but I don't believe that __hotSwapped is a valid record field name.

[fluxxu]

I think I came out a better way to solve your problem.
Now you can register callback to patch flags/state:
https://github.com/fluxxu/elm-hot-loader-starter/blob/master/src/index.js#L19
So you can

elm.hot.subscribe(function (event, context) {
  if (event === 'swap') {
    context.flags.hotSwapped = true
  }
})

Then your Main.init will get flags with hotSwapped === true
Please upgrade elm-hot-loader to 0.4.x to use this.

[mattrrichard]

Finally got to test this, and yep it works well for me. Thanks!