compute retrievability for all states

[joshdavham]

Currently, you can only get a card’s retrievability when it’s in the Review state. I thought it would be good enhancement if you could also compute the retrievability for cards in the other states as well.

Changes:

• Compute retrievability for Learning and Relearning cards the same way it's computed for Review cards
• New cards have 0 retrievability
  • Rationale: given that new cards have S=0, then for t>0, R(t,S) -> 0 as S -> 0(+)
• Made the datetime argument optional for the get_retrievability function.
  • If the function is invoked without a given datetime, the datetime defaults to datetime.now(timezone.utc) similar to the Card class's __init__ method or the FSRS class's review_card method.
• Bump minor version: 3.0.0 -> 3.1.0
• Add test: test_retrievability to demonstrate this change works

Let me know if this makes sense or if you have questions 👍

[L-M-Sherlock]

LGTM

[ishiko732]

Setting self.stability = 0 will trigger an error.

[L-M-Sherlock]

self.stability = 0 only when the state is new, right?

[ishiko732]

Continuously selecting "Again" will cause the stability to approach zero indefinitely.

[L-M-Sherlock]

Good point. I haven't known that. @joshdavham, does it happen in py-fsrs?

[joshdavham]

@ishiko732 Yeah, explicitly setting a card's stability to 0 would cause an error if the card is not in the New state. This would probably have to be done by the user though, which I'm not sure they would do. Stability is supposed to be computed internally by the code, not set by user.

[joshdavham]

@L-M-Sherlock You're asking if stability asymptotically approaches 0 when a card's rated again a bunch of times? Yeah it does

I'm not sure if it would every actually reach 0 and cause an error however