Safe Haskell	Safe-Inferred
Language	Haskell2010

FSRS.Scheduler

Contents

Cards
Scheduler
Reviewing
Fuzz

Description

A card scheduler built on top of the FSRS-7 memory model.

The memory model in FSRS.Algorithm is fully specified by upstream; the scheduling policy around it is not, so this module follows the widely used reference scheduler from py-fsrs: a card walks through a list of learning steps, graduates into the review queue, and drops into relearning steps when it lapses.

Two things differ from py-fsrs, both because FSRS-7 is built for continuous intervals where FSRS-6 was built for whole days:

Elapsed time is measured in fractional days rather than truncated to whole days, and it is fed to the model directly. There is no same-day special case, because transitionCoefficient already is one.
Scheduled intervals are not rounded to whole days. Set schedulerMinimumInterval to 1 (the default) to keep review-queue intervals at a day or more anyway, or lower it to let FSRS-7 schedule sub-day reviews.

Fuzzing is explicit rather than ambient: reviewCard is deterministic, and reviewCardFuzzed takes the random sample as an argument, so scheduling stays a pure function of its inputs.

Synopsis

data CardState
- = Learning
- | Review
- | Relearning
data Card = Card {
- cardState :: !CardState
- cardStep :: !(Maybe Int)
- cardMemory :: !(Maybe MemoryState)
- cardDue :: !UTCTime
- cardLastReview :: !(Maybe UTCTime)
}
newCard :: UTCTime -> Card
cardRetrievability :: Scheduler -> Card -> UTCTime -> Maybe Retrievability
data Scheduler = Scheduler {
- schedulerParameters :: !Parameters
- schedulerDesiredRetention :: !Retrievability
- schedulerLearningSteps :: ![NominalDiffTime]
- schedulerRelearningSteps :: ![NominalDiffTime]
- schedulerMinimumInterval :: !Days
- schedulerMaximumInterval :: !Days
}
defaultScheduler :: Scheduler
nextReviewInterval :: Scheduler -> Stability -> Days
data ReviewLog = ReviewLog {
- logRating :: !Rating
- logReviewTime :: !UTCTime
- logElapsedDays :: !Days
- logStateBefore :: !CardState
- logMemoryBefore :: !(Maybe MemoryState)
- logMemoryAfter :: !MemoryState
- logInterval :: !NominalDiffTime
}
reviewCard :: Scheduler -> Card -> Rating -> UTCTime -> (Card, ReviewLog)
reviewCardFuzzed :: Double -> Scheduler -> Card -> Rating -> UTCTime -> (Card, ReviewLog)
previewIntervals :: Scheduler -> Card -> UTCTime -> [(Rating, NominalDiffTime)]
fuzzRanges :: [(Days, Days, Double)]
fuzzBounds :: Scheduler -> Days -> Maybe (Days, Days)
fuzzInterval :: Scheduler -> Double -> Days -> Days

Cards

data CardState Source #

Where a card sits in the learn / review / relearn cycle.

Constructors

Learning	Working through `schedulerLearningSteps`.
Review	Graduated; intervals come from the memory model.
Relearning	Lapsed, working through `schedulerRelearningSteps`.

Instances

Instances details

Bounded CardState Source #
Instance details Defined in FSRS.Scheduler Methods minBound :: CardState # maxBound :: CardState #
Enum CardState Source #
Instance details Defined in FSRS.Scheduler Methods succ :: CardState -> CardState # pred :: CardState -> CardState # toEnum :: Int -> CardState # fromEnum :: CardState -> Int # enumFrom :: CardState -> [CardState] # enumFromThen :: CardState -> CardState -> [CardState] # enumFromTo :: CardState -> CardState -> [CardState] # enumFromThenTo :: CardState -> CardState -> CardState -> [CardState] #
Read CardState Source #
Instance details Defined in FSRS.Scheduler Methods readsPrec :: Int -> ReadS CardState # readList :: ReadS [CardState] # readPrec :: ReadPrec CardState # readListPrec :: ReadPrec [CardState] #
Show CardState Source #
Instance details Defined in FSRS.Scheduler Methods showsPrec :: Int -> CardState -> ShowS # show :: CardState -> String # showList :: [CardState] -> ShowS #
Eq CardState Source #
Instance details Defined in FSRS.Scheduler Methods (==) :: CardState -> CardState -> Bool # (/=) :: CardState -> CardState -> Bool #
Ord CardState Source #
Instance details Defined in FSRS.Scheduler Methods compare :: CardState -> CardState -> Ordering # (<) :: CardState -> CardState -> Bool # (<=) :: CardState -> CardState -> Bool # (>) :: CardState -> CardState -> Bool # (>=) :: CardState -> CardState -> Bool # max :: CardState -> CardState -> CardState # min :: CardState -> CardState -> CardState #

data Card Source #

A card's scheduling state.

Constructors

Card
Fields cardState :: !CardState cardStep :: !(Maybe Int) Index into the current step list; `Nothing` in the `Review` state. cardMemory :: !(Maybe MemoryState) `Nothing` until the card has been reviewed once. cardDue :: !UTCTime cardLastReview :: !(Maybe UTCTime)

Instances

Instances details

Show Card Source #
Instance details Defined in FSRS.Scheduler Methods showsPrec :: Int -> Card -> ShowS # show :: Card -> String # showList :: [Card] -> ShowS #
Eq Card Source #
Instance details Defined in FSRS.Scheduler Methods (==) :: Card -> Card -> Bool # (/=) :: Card -> Card -> Bool #

newCard :: UTCTime -> Card Source #

A card that has never been reviewed, due at the given time.

cardRetrievability :: Scheduler -> Card -> UTCTime -> Maybe Retrievability Source #

A card's retrievability at the given time, or Nothing if it has never been reviewed.

Scheduler

data Scheduler Source #

Scheduling policy.

schedulerMinimumInterval must not exceed schedulerMaximumInterval.

Constructors

Scheduler
Fields schedulerParameters :: !Parameters schedulerDesiredRetention :: !Retrievability The recall probability a scheduled review aims for. schedulerLearningSteps :: ![NominalDiffTime] schedulerRelearningSteps :: ![NominalDiffTime] schedulerMinimumInterval :: !Days Floor for review-queue intervals. schedulerMaximumInterval :: !Days Ceiling for review-queue intervals.

Instances

Instances details

Show Scheduler Source #
Instance details Defined in FSRS.Scheduler Methods showsPrec :: Int -> Scheduler -> ShowS # show :: Scheduler -> String # showList :: [Scheduler] -> ShowS #
Eq Scheduler Source #
Instance details Defined in FSRS.Scheduler Methods (==) :: Scheduler -> Scheduler -> Bool # (/=) :: Scheduler -> Scheduler -> Bool #

defaultScheduler :: Scheduler Source #

The FSRS-7 defaults: 90% desired retention, one-minute and ten-minute learning steps, a ten-minute relearning step, and intervals between one day and a century.

nextReviewInterval :: Scheduler -> Stability -> Days Source #

The interval a graduated card of the given stability earns, clamped to the scheduler's interval bounds.

Reviewing

data ReviewLog Source #

What happened in a single review.

Constructors

ReviewLog
Fields logRating :: !Rating logReviewTime :: !UTCTime logElapsedDays :: !Days Days since the previous review; `0` for a card's first review. logStateBefore :: !CardState logMemoryBefore :: !(Maybe MemoryState) logMemoryAfter :: !MemoryState logInterval :: !NominalDiffTime The interval that was scheduled, after any fuzz.

Instances

Instances details

Show ReviewLog Source #
Instance details Defined in FSRS.Scheduler Methods showsPrec :: Int -> ReviewLog -> ShowS # show :: ReviewLog -> String # showList :: [ReviewLog] -> ShowS #
Eq ReviewLog Source #
Instance details Defined in FSRS.Scheduler Methods (==) :: ReviewLog -> ReviewLog -> Bool # (/=) :: ReviewLog -> ReviewLog -> Bool #

reviewCard :: Scheduler -> Card -> Rating -> UTCTime -> (Card, ReviewLog) Source #

Review a card. Deterministic: no fuzz is applied.

reviewCardFuzzed :: Double -> Scheduler -> Card -> Rating -> UTCTime -> (Card, ReviewLog) Source #

Review a card, fuzzing the scheduled interval.

The first argument is a uniform sample from [0, 1], which the caller draws however it likes; values outside that range are clamped. Fuzz only ever applies to a card that ends up in the Review state with an interval of at least 2.5 days — see fuzzInterval.

previewIntervals :: Scheduler -> Card -> UTCTime -> [(Rating, NominalDiffTime)] Source #

The interval each of the four ratings would earn, without reviewing. Handy for showing the four buttons' answers in a UI.

Fuzz

fuzzRanges :: [(Days, Days, Double)] Source #

(start, end, factor) triples describing how much an interval may be nudged: every day of the interval that falls inside a range contributes factor days of slack. Taken from py-fsrs.

fuzzBounds :: Scheduler -> Days -> Maybe (Days, Days) Source #

The window an interval may be fuzzed into, or Nothing for intervals shorter than 2.5 days, which are left alone.

fuzzInterval :: Scheduler -> Double -> Days -> Days Source #

Nudge an interval by a random amount within fuzzBounds.

The first argument is a uniform sample from [0, 1]; it is clamped, so any finite value is safe. Intervals below 2.5 days are returned unchanged.