The tyrian from purplekingdomgames

The existing Http example is prone to being rate limited.

Consider: Supporting `Try` / `MonadError` like functionality on `Task`

Two specific places the re-evaluate:

The TaskRunner where it runs delay. If we allowed handleErrorWith type functionality, we could avoid doing the try catch.
The HTTP Cmd has a try catch in it - could it be factored out?

Launch mechanism

When Indigo starts, instead of using main, you embed the game and invoke a launch method. The advantage of this is that you can supply (crude) flags (like command line args) during start up, as well as the ID of the element to hook onto.

Should Tyrian do a similar thing?

`Random` `Cmd` should return a more specific type

For example, calling Random.double should return Cmd[RandomValue.NextDouble] not Cmd[RandomValue].

Tyrian g8 templates: sbt

Add basic support for SSR

Should just be a case of providing a method you can call that performs the presentation work based on a given model. Then you could use the same codebase for SPA's or SSR.

Parcel loses track of source maps

I suspect all that's happening is that parcel keeps moving things around during the build and the browser can't keep up.

There's some suggestion here that we might be able to tell parcel where to look - but my initial stab at this wasn't very fruitful.

https://parceljs.org/features/targets/#sourcemap

Add WebSocket support

Http decoders: Add a default String decoder

Remove Cats

Cats core is currently a dependency, but unless I'm mistaken it is only there to provide some simple and wholly optional monoid syntax that is:

Easily mimicked
Literally not as nice as NOT using monoid operators in most cases.

So I'd prefer to ditch it and make Tyrian a zero dependancy lib for now. That doesn't rule out bringing it back if some future need arises. The home-grown Task implementation springs to mind.

Notice of intention to change licence type from BSD 3 to MIT (Scalm)

Dear former contributors to Scalm!

(According to github, that's: @julienrf, @salc2, and @battermann.)

As mentioned on the Scalm gitter account, I've forked the excellent (but seemingly abandoned) Scalm as I hope to use it in conjunction with Indigo. I've brought it up to date, converted it to Scala 3, and I've started changing the apis and internals in to suit my own purposes. I do plan to release it (why not?).

The current licence agreement specifically mentions Julien Richard-Foy and EPFL, and I'd like to switch across to an MIT licence and change the named people to be consistent with Indigo's licence agreement and representative of who is looking after this version of the project (the original will remain as-is).

I would appreciate it if you could respond with your approval, or indeed your objections.

I realise most of you haven't looked at Scalm for 3 years or so (which is part of the reason for the fork!) so if I don't hear from you within 30 days, I'll go ahead and assume I have your consent.

Thank you for building the project in the first place!

All the best,

Dave

Setup GHA CI

Remove "simple" Tyrian constructor?

I think it just lures you into a false sense of security, and you instantly need to full one to actually do anything useful.

Consider Seq instead of List

HTML input value is not reflected on the view

It seems the value attribute of an input field is not correctly reflected on the view when the model changes.

case class Model(input: String)

input(`type` := "text", attribute("value", model.input))

The input field is updated either on Msg.SymbolChanged or reset (input = "") on Msg.Subscribe.

// view
input(`type` := "text", attribute("value", model.input), onInput(s => Msg.SymbolChanged(s)))

// update
case Msg.SymbolChanged(in) =>
  model.copy(input = in) -> Cmd.Empty

case Msg.Subscribe =>
  val nm  = model.copy(input = "")
  val cmd = model.ws.map(ws => Cmd.Batch(ws.publish(WsIn.Subscribe.asJson.noSpaces), refocusInput))
  nm -> cmd.getOrElse(Cmd.Empty)

Everytime the model is updated in the first message, I can see the HTML input value is updated (seeing this by inspecting the element in the browser). However, when the Subscribe message is handled and the input in the model is reset, the value attribute is properly reset as well, but the view remains with the previous value.

Here's an example from the trading application, right after clicking subscribe, where I first noticed the issue.

The value attribute is reset but the view still shows the previously entered value "CHFEUR". I'd expect the input field to be cleared up as well.

Increase code strictness

Enabled strict equality and look into explicit nulls.

Investigate: Organise Browser Tests

What are the options?

Create a `TyrianApp` trait

Like Indigo, you can choose to just inherit a trait and it tells you when to do while implementing the initialisation for you in the background.

Improve tag generation to allow Strings as tag contents

I keep trying to do this:

title("My Website")

when it should be

title(text("My Website"))

Would be nice to have an additional constructor that just allows that?

Dev Mode: Store model state on update

If we allowed a dev mode that stored the model state on update, than a hot reload would not reset the page to the beginning.

Add Indigo bridge example

Borrow naming convention from Laminar

Something I've come across a lot is what do you call things, e.g. class="" can't be class(..) because it's a reserved word.

Here's how Laminar does it, even if we don't just borrow this library, I think there's benefits in consistency - and they're already thought this through.

https://github.com/raquo/scala-dom-types#naming-differences-compared-to-native-html--dom

Upgrade to Scala 3.1.1 and sbt 1.6.2

Support Anchor URL changes

window.onhashchange seems to be the mechanism, so I guess (and you could do this right now) you would subscribe the event (Sub) and send a Msg on change with the details, update your model, and render different contents.

Add support for IndexedDB

BUG: `value` attribute is unusable

If you try to set value := "hello" you can an ambiguous overload error.

The workaround is to use: attribute("value", "hello")

Deprecate Hooks?

They're very Snabbdom specific, it's exposing a lot about the engines internals to allow it.

Upgrade to Mill 0.10.0

Allow 'no value' attributes to accept a predicate

In Elm, a field like hidden can be expressed as hidden(true|false) so that you can decide whether to use the hidden attribute in-line.

My also require the production of the notion of an empty attribute.

Indigo integration

Should direct support for Indigo be added so that Tyrian and easily embed and interact with Indigo games?

It's entirely possible to integrate Indigo games now but the experience is not slick, and there is no specialist machinery on either side to offer smooth communications from one to the other.

Since they are both Scala.js, could Tyrian invoke Indigo directly?

Docs: Add a description of how to use scalajs-bundler

I was working with someone the other day and the detail of this isn't immediately obvious unless you already know what you're doing...

Create a "focus" `Cmd`

Consider replicating some of the behaviours in this Elm package, particularly the focus command:

https://package.elm-lang.org/packages/elm/browser/latest/Browser.Dom

Add a built-in Cmd to handle simple local storage

WebSockets / TyrianIndigoBride terminology change

Currently they have a send and a subscribe method. Should either be send and receive or publish and subscribe. I'm going with the latter because I can't spell receive without thinking about it excessively hard.

No way to declare doctype

I don't think we have a way to write:

<!DOCTYPE HTML>

On the off chance we want to generate whole pages via SSR.

Investigate replacing internal custom `Task` implementation with something more standard.

There is a hand rolled Task implementation that is used to power Tyrian, that looks like it ought to be replaceable with something more standard/modern.

However, the job is not to Finally Tagless all the things, at least not yet, as that would have a serious impact on the APIs.

WebSockets: Can the heartbeat be a form of `Sub`?

The heartbeat is currently using a setTimeout right in the WebSockets code.

Generate common styles

It would be nice to generate things like:

Style("font-family" -> "Arial, Helvetica, sans-serif")

..to help avoid the typos.

Tyrian g8 templates: Mill

Add websockets example

Review: Component/Child Code

There is the beginnings of a component pattern in Tyrian (inherited) that is not connected to anything, nor is it featured in the examples - not even the sub component example.

Suspect it's heading towards halogen style component nesting. It isn't needed for anything but we should decide what to do with it. Keep it and make it work, or remove it.

Upgrade to Scala.js 1.9.0

Replace Snabbdom

Snabbdom is the JS Virtual DOM implementation that powers Tyrian's rendering at the moment.

I'd like to investigate replacing it with a pure Scala.js Incremental DOM (or something).

The point of this is to make the SPA and SSR rendering happen via the same process.

Step one is to do the naive thing and just re-render on-change, and replace the innerHtml of the top level element completely. That will work (probably), but will be slow (probably, but is cheap memory-wise), everything after that is an optimisation.

We'll need a way to run tests in the browser so we can talk to a real DOM - there is a Scala.js selenium thing for that.
We could also do with figuring out a way to benchmark this compared a) to Snabbdom and b) to other solutions, like Elm and React. Don't need to be the fastest on the market but can't be appallingly slow either.