Implement snapshot in FRP

Question

I am implementing the FRP framework in Scala and I seem to have run into a problem. Motivated by some thinking, I decided to limit this question to the public interface of my structure, so Behaviors could be evaluated only in the "present", namely:

behaviour.at(now) 

This is also consistent with Conal’s assumption in Franz's work that behavior is always evaluated / selected with increasing time. It restricts conversions to Behaviors, but otherwise we find ourselves in huge problems with Behaviors, which represent some input:

 val slider = Stepper(0, sliderChangeEvent) 

Using this behavior, an estimate of future values ​​will be incorrect, and an unlimited amount of memory will be required to evaluate past values ​​(all occurrences used in the "slider" event must be saved).

I am having problems specifying the “snapshot” operation in Behaviors, given this limitation. My problem is best explained by an example (using the slider mentioned above):

 val event = mouseB // an event that occurs when the mouse is pressed val sampler = slider.snapshot(event) val stepper = Stepper(0, sampler) 

My problem is that if the "mouseB" event occurred when this code is executed, the current "stepper" value will be the last "slider" pattern (the value at the time it last appeared). If the time of the last occurrence is in the past, we therefore complete the “slider” estimate using the elapsed time, which violates the above rule (and your original assumption). I see a couple of ways to solve this problem:

• We record the past (save all past events in the event), allowing us to evaluate behavior with past times (using an unlimited amount of memory).
• We modify the “snapshot” to accept a temporary argument (“sample after this time”) and apply this time> = now
• In a more wacky move, we could somehow limit the creation of FRP objects to the initial setup of the program and only start processing events / input after this installation is complete.

I also just could not implement the “pattern” or remove the “stepper” / “switcher” (but I really don't want to do any of these things). Anyone have any thoughts on this? I didn’t understand something?

Answer 1

Oh, I understand what you mean now.

Your "you can only try on the" now ", I think the restriction is not tight enough. It should be a little stronger so as not to look into the past. Since you are using the ecological concept of now , I would define the functions of building behavior in terms of this (for now cannot go forward simply by executing definitions which in my last answer will be messy) For example:

Stepper(i,e) is the behavior with the value of i in the interval [now,e1] (where e1 is the time of the first appearance of e after now ) and the value of the last occurrence of e after.

With this semantics, your prediction about the stepper value that fell into this riddle is dismantled, and the pedometer will now have the value 0. I don’t know if you need this semantics, but it seems to me quite natural.

Answer 2

From what I can say, you are worried about the state of the race: what happens if the event occurs during code execution.

Purely functional code does not like to know that it is executing. Functional methods are in the best condition in a clean environment, so it doesn't matter in which order the code is executed. The way out of this dilemma is to pretend that each change occurred in one sensitive (internal, probably) piece of the imperative code; pretend that any functional declarations in the FRP structure occur 0 times, therefore, it is impossible to change anything during their declaration.

No one should ever sleep or really do something time sensitive in the code section that talks about behavior and things. In fact, the code that works with FRP objects should be clean, then you have no problem.

This does not necessarily preclude running it on multiple threads, but to support the fact that you may need to reorganize your internal views. Welcome to the world of implementing the FRP library. I suspect that your inner view will change many times during this process. :-)

Answer 3

I am confused in your confusion. I see that Stepper will “set” the behavior for the new value whenever an event occurs. So what happens is this:

When the mouseB event mouseB , the value of the slider ( snapshot ) behavior will be read. This value will be "set" in the behavior of Stepper .

So, it is true that Stepper will “remember” meanings from the past; The fact is that he only remembers the last meaning from the past, and not all.

Semantically, it is best to model Stepper as a function that luqui offers.