Fix/optimize card priority calculation and fix learn span behavior

[L-M-Sherlock]

This PR includes several improvements to the FSRS simulator:

1. Enhanced card priority calculation:
   • Modified priority tuple to use i32 instead of usize
   • Adjusted difficulty scaling for accuracy in priority calculation
   • Added filtering for cards beyond learn span
2. Added new test case:
   • Validates that changing learn span doesn't affect review counts per day in the same date

[Luc-Mcgrady]

The learn_span still effects the earlier cards in my deck.
The test fails if you up the numbers high enough

I'll dm you a deck if the test isn't enough.

[L-M-Sherlock]

The inconsistency starts from here:

[L-M-Sherlock]

@Luc-Mcgrady I hope the latest commit solve your problem.

[Luc-Mcgrady]

It's still weird in Anki but I can't get the test to fail. I'm guessing this might be a problem with anki's code?

[L-M-Sherlock]

It's still weird in Anki but I can't get the test to fail. I'm guessing this might be a problem with anki's code?

Could you use dbg to print the input args from your Anki?

[Luc-Mcgrady]

[rslib/src/scheduler/fsrs/simulator.rs:32:9] &req = SimulateFsrsReviewRequest {
    params: [
        0.08572003,
        0.40419477,
        4.1386,
        19.5147,
        5.3538256,
        1.4269736,
        0.9058566,
        8.294402e-5,
        1.3015078,
        0.016733568,
        0.6890899,
        2.0352721,
        0.15819947,
        0.2383912,
        1.4175994,
        0.027590245,
        2.8237,
    ],
    desired_retention: 0.8,
    deck_size: 0,
    days_to_simulate: 10,
    new_limit: 0,
    review_limit: 200,
    max_interval: 36500,
    search: "preset:\"Geography\" -is:suspended",
}
[rslib/src/scheduler/fsrs/simulator.rs:32:9] &req = SimulateFsrsReviewRequest {
    params: [
        0.08572003,
        0.40419477,
        4.1386,
        19.5147,
        5.3538256,
        1.4269736,
        0.9058566,
        8.294402e-5,
        1.3015078,
        0.016733568,
        0.6890899,
        2.0352721,
        0.15819947,
        0.2383912,
        1.4175994,
        0.027590245,
        2.8237,
    ],
    desired_retention: 0.8,
    deck_size: 0,
    days_to_simulate: 20,
    new_limit: 0,
    review_limit: 200,
    max_interval: 36500,
    search: "preset:\"Geography\" -is:suspended",
}

It does set the due date of certain cards to the limit but I cant see how that would effect anything

[L-M-Sherlock]

The due affects the priority. Maybe it causes the inconsistency.

[L-M-Sherlock]

Weird. These cards are new, so they are filtered in the simulator:

fsrs-rs/src/optimal_retention.rs

Lines 189 to 194 in 0625bfa

	if let Some(existing_cards) = existing_cards {
	cards.extend(
	existing_cards
	.into_iter()
	.filter(|card| card.stability > 1e-9),
	);

[Luc-Mcgrady]

Well I just copied the entire output of my deck into the rs file, and besides making my IDE cry I think it proves its not Anki's problem

[L-M-Sherlock]

Nice! Thanks for your assistance in fixing the bug~