CurrentTarget / Target ops

[mariusmuja]

This PR is a result of the discussion we had in #119 (comment). It still uses the non-typesafe cast for currentTarget, but allows for a nicer syntax and uses the TagWith... typeclasses for restricting the type the target can be cast to.

The syntax is the following:

• currentTarget uses html.Input by default as the underlying event target:

input(
  onInput.value --> string,
  onClick.checked --> bool,
  onKeyDown.valueAsNumber --> number
)

The PR removes onInputString, onInputNumber and onInputChecked and updates the deprecation message on inputString, inputNumber and inputChecked to use onInput.value, onInput.valueAsNumber and onChange.checked instead.

• it's possible to force a different event target for currentTarget, but only if the TagWith... typeclasses are implemented for it (html.TextArea, html.Select ).

textArea(
 onChange.value[html.TextArea] --> string,
)

• for target, the type from TypedTargetEvent is used by default

input(
  onSearch.target.value --> stringStream,
  onSearch.target.valueAsNumber --> doubleStream,
  onSearch.target.checked --> boolStream,
)

• in case the type of TypedTargetEvent is not specific enough it's possible to force a more specific type:

input(
  //onClick.target.value --> stringStream,   // does not compile because onClick is TypedTargetEvent[dom.Element], not specific enough
  onClick.target[html.Input].value --> stringStream,
  onClick.target[html.Input].valueAsNumber --> doubleStream,
  onChange.target[html.Input].checked --> boolStream,
)

[cornerman]

Thanks for working on this. I did not have time to make a full review of the changes just yet.

But I do not get why we need these generic methods for currentTarget onChange.value[html.TextArea]. Just using onChange.value will behave the same way, as casting is unchecked in scalajs anyhow. What is the benefit of having it?

Also, why do we decide that currentTarget can assume html.Input, but target cannot? Here, we do not have typesafety anyhow.

[mariusmuja]

But I do not get why we need these generic methods for currentTarget onChange.value[html.TextArea]. Just using onChange.value will behave the same way, as casting is unchecked in scalajs anyhow. What is the benefit of having it?

Also, why do we decide that currentTarget can assume html.Input, but target cannot? Here, we do not have typesafety anyhow.

You are correct, there's no benefit from having the generic methods for casting to different event targets, they will behave the same way. I've removed them and simplified the PR.

[cornerman]

rename TagWith to TargetOps in test string here

[cornerman]

I would argue against mapping to EventTarget here, as _.map(_.target) seems good enough to me and we do not have the same helper functions for other event properties. Instead we can only define a target object with the above methods in an implicit class for the EmitterBuilder.

Like this:

  implicit class TargetAsInput[E <: Event, O <: Event](builder: EmitterBuilder[E, O]) {
      object target {
          def value = ???
          def valueAsNumber = ???
          def checked = ???
      }
   }

[mariusmuja]

Good point, I'll update the PR.

[cornerman]

I think this PR improves the handling for using currentTarget and target. For now, this is nice to use. But I am still unhappy that we do not have any kind of typesafety here - which we should definitely tackle in an upcoming PR.