Increase granularity of activity tracking to per turn.

[kevingranade]

Summary

SUMMARY: None

Purpose of change

Increase activity tracking granularity to per turn.
Simplify character class slightly by extracting activity and weariness logic to a helper class.

Describe the solution

Extract the weariness and activity tracking code into a separate class.
Also activity level was not being saved or loaded, do that.
Increase granularity of tracking to per-turn.
Maintains a rolling average of activity level for bmr calculation, and also an instantaneous activity level for other uses.

Describe alternatives you've considered

I'm not at all happy with the names, activity_tracker is okish, but activity_history is just bad.

Testing

Add a number of unit tests, insure the existing weariness and calorie expenditure tests continue to work.
Run around and do active stuff manually and validate the results.

[kevingranade]

Updated with WIP adjustments to make activity ticks more granular by tracking them on a turn-by-turn basis and then applying them once every 5 minutes.

I'm kind of stuck what this does that breaks weariness so badly, it absolutely DESTROYS the weariness tests because weariness passes an inflection point and then just keeps climbing rapidly, I'm not sure what I did to make this happen.
Posted this and then immediately realized what it was and fixed it.

This works, but needs some polish and picking the commits apart, maybe a little more testing

[anothersimulacrum]

I think this should look like this to account for the turns thing

Suggested change

	accumulated_activity += current_activity;
	// skipped turns?
	current_activity = new_level;
	num_events++;
	int turns_moved = calendar::turn - current_turn;
	accumulated_activity += current_activity * turns_moved;
	current_activity = new_level;
	num_events += turns_moved;

[kevingranade]

Sort of, we can do number of turns, but the activity level if anything should be NO_EXERCISE, representative of the fact that no activity was logged in the interim.

[anothersimulacrum]

I'd have to look through to see when skipped turns trigger, but this is the correct behavior for what I was considering, when an activity takes more than a characters moves to complete. E.g. if I drag furniture and it takes me 6 turns (but it is not an activity), I should be counted as whatever activity level for 6 turns, instead of just 1, and nothing for 5 turns.

[kevingranade]

The activity should be logging its exertion level every turn.

[anothersimulacrum]

I am talking about non-activity actions like that. Dragging is now an activity, but if it were not this would be the correct behavior.
I am not sure how many of these we have, however. Melee combat comes to mind as somewhere this could pop up.

[kevingranade]

Still, anything the player does should log an activity, and anything that takes more than one turn should log each turn.
If that isn't happening, we can fix it.

[kevingranade]

Rebased and fixed the failing starve test. It was being thrown off by the rolling average of activity level, so if another test happened right before it, it was picking up an unusually high activity level.

[kevingranade]

Rebased on latest master and should be ready to go, I incorporated @anothersimulacrum 's activity level reporting feature into the activity tracker.

For future consideration, I want to unify move point, stamina, and weariness tracking behind a method called Character::expend_effort( int moves, float exertion_level ) That way we never expend moves without logging effort, and we never log effort without expending moves, and the product of moves spent and exertion level should be sufficient to determine stamina and fatigue for the action as well as being the sole place to spend moves.

[actual-nh]

For future consideration, I want to unify move point, stamina, and weariness tracking behind a method called Character::expend_effort( int moves, float exertion_level ) That way we never expend moves without logging effort, and we never log effort without expending moves, and the product of moves spent and exertion level should be sufficient to determine stamina and fatigue for the action as well as being the sole place to spend moves.

Spending of calories?

[kevingranade]

Spending of calories?

Yes, exertion level would feed into calorie expenditure.

[actual-nh]

This PR has some of the same thinking as #47653, namely trying to smooth out some of the current fluctuations due to things happening every X minutes. If it goes in prior to #47653, it will unfortunately require a bit of a rewrite on my part of #47653 - or vice-versa, if #47653 goes in first - but that would just be an annoyance.

For the future, the finer activity tracking would make it possible to have less-than-full intervals for figuring out weariness recovery - one could benefit from 3 minutes of rest, for instance.

[actual-nh]

BTW, the two failing builds are both at the activity_counter_from_1_to_300 test; tracker.average_activity() vs expected_activity 1.00863f vs 150.5f. (The Travis one is for mingw-w64.)

[anothersimulacrum]

I've assigned myself to this so I (hopefully) remember to review it.

[kevingranade]

I THINK what's happening with that failing test is that calendar::turn is relatively advanced, and the loop in activity_tracker::new_turn() is incorporating the difference between calendar::turn_zero and calendar::turn into the tracker, resulting in the activity score getting greatly depressed. The thing I don't get is why it isn't always happening.

[actual-nh]

I can see why putting the do_turn after advancing turns makes sense. But why update body prior to do_turn?

[kevingranade]

Because that's the order in which it occurs in game::do_turn().
Technically moves are incremented at the end of the turn instead of the start, but update_body() shouldn't be interacting with moves, so it shouldn't matter.

[AlexMooney]

Seems like with the change to Character::weariness() this could be left alone.