yodaiken / dolphinsr Goto Github PK

I'm adapting this for a tiny project of mine, and as far as I see, a logic error can consistently be caused by generating a few cards, and then only picking "hard" as an answer.

This seems to happen in applyToLearningCardState, with an end state of rating = "hard" and prev.consecutiveCorrect = 1.

Hi there,
I'm playing with your implementation of spaced repetition and confused slightly on when to use 'easy', 'good', 'hard', or 'again' for the rating of a review.

'again' seems obvious for when the review is incorrect.
But if it's correct... how do I decide between 'easy', 'good', or 'hard'?

Forgive me if this is obvious to those who have poked more at Anki or SM2 or some such versions of this stuff :D

Thanks!

I am having trouble getting this up and running with my collection of cards. In the instructions, it sounds like a review only needs to store:

{
  master: Id,
  combination: Combination,
  ts: Date,
  rating: Rating,
}

however, we need to procure consecutiveCorrector, factor, mode and lapses from previous reviews somehow. Don't we need to store these fields in the review (in our datastore) each time a review is created (or at minimum in a card's initial review) so that the scheduler can correctly process and space cards for future reviews?

In the dist code, I traced card state variables back to the root and it appears they all originate from a loaded review. Taking a look at the ComputeCardSchedule.test.js however, these fields are static and only defined in some card states:

    const learning: LearningCardState = {
      master,
      combination: { front: [1, 0], back: [0] },
      mode: 'learning',
      consecutiveCorrect: 0,
      lastReviewed: dates.today,
    };
    const lapsed: LapsedCardState = {
      master,
      combination: { front: [1, 0], back: [0, 1] },
      mode: 'lapsed',
      consecutiveCorrect: 0,
      lastReviewed: dates.today,
      factor: 1000,
      interval: 0,
      lapses: 1,
    };