Improved Axis API

[simpuid]

Added functions to get current value, previous value and delta value of Axis<T>
previous value means the value before calling Axis<T>::update and current value means the value after calling Axis<T>::update. delta value is the difference between current and previous values and it can be 0.0f32. delta would be None when axis' value is not set after calling Axis<T>::update
This will help to detect actual changes to axis value by the caller system easily, like in the updated gamepad_input example.
The other way is to approx compare delta with 0.0f32 but that's tiring (see the first commit)

[cart]

I'm a little hesitant about this one. If there are multiple "axis input events" in the same update, that would result in some of the delta being missed.

I also think most axis use cases use absolute values / don't require deltas (ex: "how fast should I move this thing" and "is this axis past a certain threshold"), so I think I'm fine letting the user track this themselves. For use cases that need to track "did something happen since the last time we ran", I think exposing the axis events in an Events queue is a more generic approach.

[simpuid]

I'm a little hesitant about this one. If there are multiple "axis input events" in the same update, that would result in some of the delta being missed.

Actually the delta would be the sum of all the deltas

I also think most axis use cases use absolute values / don't require deltas (ex: "how fast should I move this thing" and "is this axis past a certain threshold"), so I think I'm fine letting the user track this themselves. For use cases that need to track "did something happen since the last time we ran", I think exposing the axis events in an Events queue is a more generic approach.

Events seems like a good approach to solve this problem. I shall open a new PR with AxisChange(GamepadAxisType, f32) and ButtonChange(GamepadButtonType, f32) variants of GamepadEventType. This would allow users to apply custom deadzones for axis and buttons which was discussed in #681

Closing because it's redundant

[cart]

Actually the delta would be the sum of all the deltas

I don't thats true. If there are two EventType::AxisChanged events processed in the same frame (lets say with original value 0, new1 == 0.5, new2 == 0.75), this is what I'm pretty sure happens:

1. new1 is processed, set is called. previous = 0, current = 0.5

• if we called delta() now, we would get a value of 0.5

2. new2 is procesed, set is called. previous = 0.5, current = 0.75

• if we called delta() now, we would get a value of 0.25

3. some user system runs in update and calls .delta(), we get 0.25 despite the fact that it actually should be 0.75

[simpuid]

new2 is procesed, set is called. previous = 0.5, current = 0.75

it would be previous() = 0, current() = 0.75 and delta() = 0.75
If you look closely in the Axis<T>::set function, the previous values are inserted when they are absent (vacant entry), they are never updated .

Axis<T>::update clears the previous values so the previous value is inserted only at the first call of Axis<T>::set after Axis<T>::update

Whatever the case, the event approach looks better

[cart]

Oops yeah you're right. Thanks for bearing with me 😄