Fix fpsWhen, including consideration for hotswapping

[jvoigtlaender]

fixes #139
yet another alternative to #141 and #142
based on @jwmerrill's version of fpsWhen

All local state now packaged in a signal, to allow Elm Reactor to pick it up during hotswapping.

[jwmerrill]

What's the reason for this? Does, e.g., plain fps (which starts with a true value for isOn) not start without it?

[jwmerrill]

What's the reason for this? Does, e.g., plain fps (which starts with a true value for isOn) not start without it?

[jvoigtlaender]

Exactly. Note that with foldp below, the update-functions isn't called on signal initialization at all. The initial value of a foldp is the explicit initial value argument, not anything obtained from a call of the updating function. So without this here, no timeout will be created at signal initialization (since the only other call to setTimeout is in the update-function, which will not get invoked).

[jwmerrill]

Another alternative would be to unroll the foldp one step, so replace

A3(Signal.foldp, F2(update), initialState, input)

in line 68 with

A3(Signal.foldp, F2(update), update(input.value, initialState), input)

In that case, you'd also want initialState.wasOn to start false in line 37.

If you did that, you could drop the conditional here.

The advantage of unrolling is that it's more obvious that you won't end up with two timeout loops running in parallel.

[jvoigtlaender]

Oh yes, I like that. Also, concerning the initial value handling it brings back some clarity that was a quality of the non-foldp version.

[jwmerrill]

Great. This looks good to me.

[jwmerrill]

Great. This looks good to me.

[evancz]

Great work guys, this looks really good!

[evancz]

Great work guys, this looks really good!

[jvoigtlaender]

Looks like I pushed this commit: jvoigtlaender@f11f7ca too late, it occurs in my branch but not in the merge.

Think it is relevant?

[jvoigtlaender]

Looks like I pushed this commit: jvoigtlaender@f11f7ca too late, it occurs in my branch but not in the merge.

Think it is relevant?

[jvoigtlaender]

I followed up my previous question with a new pull request (#160).

[jvoigtlaender]

I followed up my previous question with a new pull request (#160).