GitHub/Anki
1
0
Fork 0
mirror of https://github.com/ankitects/anki.git synced 2025-09-18 22:12:21 -04:00
Code Issues Projects Releases Packages 1 Wiki Activity Actions
Anki/pylib/anki/sched.py
Damien Elmes 5f9792392a don't cap child counts to parents when reviewing in v2 
https://forums.ankiweb.net/t/anki-2-1-41-beta/7305/59

When originally implemented in 21023ed3e5,
a given deck's limit was bound by its parents. This lead to a deck list
that seemed more logical in the parent limit < child limit case, as
child counts couldn't exceed a parent's, but it obscured the fact that
child decks could still be clicked on to show cards. And in the parent
limit > child limit case, the count shown for the child on the deck list
did not reflect how many cards were actually available and would be
delivered.

This change updates the reviewer to ignore parent limits when getting
review counts for the deck, which makes the behaviour consistent with
the deck list, which was recently changed to ignore parent limits.

Neither solution is ideal - this was a tradeoff v2 made in order to keep
fetching of review cards from multiple decks reasonably performant. The
experimental scheduling work moves back to respecting limits on
individual children, so this should hopefully improve in the future.

Also removed _revForDeck(), which was unused.
2021-03-02 10:23:06 +10:00

723 lines
25 KiB
Python
Raw Blame History

# Copyright: Ankitects Pty Ltd and contributors
# License: GNU AGPL, version 3 or later; http://www.gnu.org/licenses/agpl.html
from __future__ import annotations
import random
import time
from heapq import *
from typing import Any, List, Optional, Tuple, Union
import anki
from anki import hooks
from anki.cards import Card
from anki.consts import *
from anki.decks import QueueConfig
from anki.schedv2 import Scheduler as V2
from anki.utils import ids2str, intTime
# queue types: 0=new/cram, 1=lrn, 2=rev, 3=day lrn, -1=suspended, -2=buried
# revlog types: 0=lrn, 1=rev, 2=relrn, 3=cram
# positive revlog intervals are in days (rev), negative in seconds (lrn)
class Scheduler(V2):
name = "std"
haveCustomStudy = True
_spreadRev = True
_burySiblingsOnAnswer = True
def __init__( # pylint: disable=super-init-not-called
self, col: anki.collection.Collection
) -> None:
self.col = col.weakref()
self.queueLimit = 50
self.reportLimit = 1000
self.dynReportLimit = 99999
self.reps = 0
self.lrnCount = 0
self.revCount = 0
self.newCount = 0
self.today: Optional[int] = None
self._haveQueues = False
self._updateCutoff()
def answerCard(self, card: Card, ease: int) -> None:
self.col.log()
assert 1 <= ease <= 4
self.col.markReview(card)
if self._burySiblingsOnAnswer:
self._burySiblings(card)
card.reps += 1
self.reps += 1
# former is for logging new cards, latter also covers filt. decks
card.wasNew = card.type == CARD_TYPE_NEW # type: ignore
wasNewQ = card.queue == QUEUE_TYPE_NEW
new_delta = 0
review_delta = 0
if wasNewQ:
# came from the new queue, move to learning
card.queue = QUEUE_TYPE_LRN
# if it was a new card, it's now a learning card
if card.type == CARD_TYPE_NEW:
card.type = CARD_TYPE_LRN
# init reps to graduation
card.left = self._startingLeft(card)
# dynamic?
if card.odid and card.type == CARD_TYPE_REV:
if self._resched(card):
# reviews get their ivl boosted on first sight
card.ivl = self._dynIvlBoost(card)
card.odue = self.today + card.ivl
new_delta = +1
if card.queue in (QUEUE_TYPE_LRN, QUEUE_TYPE_DAY_LEARN_RELEARN):
self._answerLrnCard(card, ease)
elif card.queue == QUEUE_TYPE_REV:
self._answerRevCard(card, ease)
review_delta = +1
else:
raise Exception(f"Invalid queue '{card}'")
self.update_stats(
card.did,
new_delta=new_delta,
review_delta=review_delta,
milliseconds_delta=+card.timeTaken(),
)
card.mod = intTime()
card.usn = self.col.usn()
card.flush()
def counts(self, card: Optional[Card] = None) -> Tuple[int, int, int]:
counts = [self.newCount, self.lrnCount, self.revCount]
if card:
idx = self.countIdx(card)
if idx == QUEUE_TYPE_LRN:
counts[QUEUE_TYPE_LRN] += card.left // 1000
else:
counts[idx] += 1
new, lrn, rev = counts
return (new, lrn, rev)
def countIdx(self, card: Card) -> int:
if card.queue == QUEUE_TYPE_DAY_LEARN_RELEARN:
return QUEUE_TYPE_LRN
return card.queue
def answerButtons(self, card: Card) -> int:
if card.odue:
# normal review in dyn deck?
if card.odid and card.queue == QUEUE_TYPE_REV:
return 4
conf = self._lrnConf(card)
if card.type in (CARD_TYPE_NEW, CARD_TYPE_LRN) or len(conf["delays"]) > 1:
return 3
return 2
elif card.queue == QUEUE_TYPE_REV:
return 4
else:
return 3
# Getting the next card
##########################################################################
def _getCard(self) -> Optional[Card]:
"Return the next due card id, or None."
# learning card due?
c = self._getLrnCard()
if c:
return c
# new first, or time for one?
if self._timeForNewCard():
c = self._getNewCard()
if c:
return c
# card due for review?
c = self._getRevCard()
if c:
return c
# day learning card due?
c = self._getLrnDayCard()
if c:
return c
# new cards left?
c = self._getNewCard()
if c:
return c
# collapse or finish
return self._getLrnCard(collapse=True)
# Learning queues
##########################################################################
def _resetLrnCount(self) -> None:
# sub-day
self.lrnCount = (
self.col.db.scalar(
f"""
select sum(left/1000) from (select left from cards where
did in %s and queue = {QUEUE_TYPE_LRN} and due < ? limit %d)"""
% (self._deckLimit(), self.reportLimit),
self.dayCutoff,
)
or 0
)
# day
self.lrnCount += self.col.db.scalar(
f"""
select count() from cards where did in %s and queue = {QUEUE_TYPE_DAY_LEARN_RELEARN}
and due <= ? limit %d"""
% (self._deckLimit(), self.reportLimit),
self.today,
)
def _resetLrn(self) -> None:
self._resetLrnCount()
self._lrnQueue: List[Any] = []
self._lrnDayQueue: List[Any] = []
self._lrnDids = self.col.decks.active()[:]
# sub-day learning
def _fillLrn(self) -> Union[bool, List[Any]]:
if not self.lrnCount:
return False
if self._lrnQueue:
return True
self._lrnQueue = self.col.db.all(
f"""
select due, id from cards where
did in %s and queue = {QUEUE_TYPE_LRN} and due < ?
limit %d"""
% (self._deckLimit(), self.reportLimit),
self.dayCutoff,
)
for i in range(len(self._lrnQueue)):
self._lrnQueue[i] = (self._lrnQueue[i][0], self._lrnQueue[i][1])
# as it arrives sorted by did first, we need to sort it
self._lrnQueue.sort()
return self._lrnQueue
def _getLrnCard(self, collapse: bool = False) -> Optional[Card]:
if self._fillLrn():
cutoff = time.time()
if collapse:
cutoff += self.col.conf["collapseTime"]
if self._lrnQueue[0][0] < cutoff:
id = heappop(self._lrnQueue)[1]
card = self.col.getCard(id)
self.lrnCount -= card.left // 1000
return card
return None
def _answerLrnCard(self, card: Card, ease: int) -> None:
# ease 1=no, 2=yes, 3=remove
conf = self._lrnConf(card)
if card.odid and not card.wasNew: # type: ignore
type = REVLOG_CRAM
elif card.type == CARD_TYPE_REV:
type = REVLOG_RELRN
else:
type = REVLOG_LRN
leaving = False
# lrnCount was decremented once when card was fetched
lastLeft = card.left
# immediate graduate?
if ease == BUTTON_THREE:
self._rescheduleAsRev(card, conf, True)
leaving = True
# graduation time?
elif ease == BUTTON_TWO and (card.left % 1000) - 1 <= 0:
self._rescheduleAsRev(card, conf, False)
leaving = True
else:
# one step towards graduation
if ease == BUTTON_TWO:
# decrement real left count and recalculate left today
left = (card.left % 1000) - 1
card.left = self._leftToday(conf["delays"], left) * 1000 + left
# failed
else:
card.left = self._startingLeft(card)
resched = self._resched(card)
if "mult" in conf and resched:
# review that's lapsed
card.ivl = max(1, conf["minInt"], int(card.ivl * conf["mult"]))
else:
# new card; no ivl adjustment
pass
if resched and card.odid:
card.odue = self.today + 1
delay = self._delayForGrade(conf, card.left)
if card.due < time.time():
# not collapsed; add some randomness
delay *= int(random.uniform(1, 1.25))
card.due = int(time.time() + delay)
# due today?
if card.due < self.dayCutoff:
self.lrnCount += card.left // 1000
# if the queue is not empty and there's nothing else to do, make
# sure we don't put it at the head of the queue and end up showing
# it twice in a row
card.queue = QUEUE_TYPE_LRN
if self._lrnQueue and not self.revCount and not self.newCount:
smallestDue = self._lrnQueue[0][0]
card.due = max(card.due, smallestDue + 1)
heappush(self._lrnQueue, (card.due, card.id))
else:
# the card is due in one or more days, so we need to use the
# day learn queue
ahead = ((card.due - self.dayCutoff) // 86400) + 1
card.due = self.today + ahead
card.queue = QUEUE_TYPE_DAY_LEARN_RELEARN
self._logLrn(card, ease, conf, leaving, type, lastLeft)
def _lrnConf(self, card: Card) -> QueueConfig:
if card.type == CARD_TYPE_REV:
return self._lapseConf(card)
else:
return self._newConf(card)
def _rescheduleAsRev(self, card: Card, conf: QueueConfig, early: bool) -> None:
lapse = card.type == CARD_TYPE_REV
if lapse:
if self._resched(card):
card.due = max(self.today + 1, card.odue)
else:
card.due = card.odue
card.odue = 0
else:
self._rescheduleNew(card, conf, early)
card.queue = QUEUE_TYPE_REV
card.type = CARD_TYPE_REV
# if we were dynamic, graduating means moving back to the old deck
resched = self._resched(card)
if card.odid:
card.did = card.odid
card.odue = 0
card.odid = 0
# if rescheduling is off, it needs to be set back to a new card
if not resched and not lapse:
card.queue = card.type = CARD_TYPE_NEW
card.due = self.col.nextID("pos")
def _startingLeft(self, card: Card) -> int:
if card.type == CARD_TYPE_REV:
conf = self._lapseConf(card)
else:
conf = self._lrnConf(card)
tot = len(conf["delays"])
tod = self._leftToday(conf["delays"], tot)
return tot + tod * 1000
def _graduatingIvl(
self, card: Card, conf: QueueConfig, early: bool, adj: bool = True
) -> int:
if card.type == CARD_TYPE_REV:
# lapsed card being relearnt
if card.odid:
if conf["resched"]:
return self._dynIvlBoost(card)
return card.ivl
if not early:
# graduate
ideal = conf["ints"][0]
else:
# early remove
ideal = conf["ints"][1]
if adj:
return self._adjRevIvl(card, ideal)
else:
return ideal
def _rescheduleNew(self, card: Card, conf: QueueConfig, early: bool) -> None:
"Reschedule a new card that's graduated for the first time."
card.ivl = self._graduatingIvl(card, conf, early)
card.due = self.today + card.ivl
card.factor = conf["initialFactor"]
def _logLrn(
self,
card: Card,
ease: int,
conf: QueueConfig,
leaving: bool,
type: int,
lastLeft: int,
) -> None:
lastIvl = -(self._delayForGrade(conf, lastLeft))
ivl = card.ivl if leaving else -(self._delayForGrade(conf, card.left))
def log() -> None:
self.col.db.execute(
"insert into revlog values (?,?,?,?,?,?,?,?,?)",
int(time.time() * 1000),
card.id,
self.col.usn(),
ease,
ivl,
lastIvl,
card.factor,
card.timeTaken(),
type,
)
try:
log()
except:
# duplicate pk; retry in 10ms
time.sleep(0.01)
log()
def removeLrn(self, ids: Optional[List[int]] = None) -> None:
"Remove cards from the learning queues."
if ids:
extra = f" and id in {ids2str(ids)}"
else:
# benchmarks indicate it's about 10x faster to search all decks
# with the index than scan the table
extra = " and did in " + ids2str(
d.id for d in self.col.decks.all_names_and_ids()
)
# review cards in relearning
self.col.db.execute(
f"""
update cards set
due = odue, queue = {QUEUE_TYPE_REV}, mod = %d, usn = %d, odue = 0
where queue in ({QUEUE_TYPE_LRN},{QUEUE_TYPE_DAY_LEARN_RELEARN}) and type = {CARD_TYPE_REV}
%s
"""
% (intTime(), self.col.usn(), extra)
)
# new cards in learning
self.forgetCards(
self.col.db.list(
f"select id from cards where queue in ({QUEUE_TYPE_LRN},{QUEUE_TYPE_DAY_LEARN_RELEARN}) %s"
% extra
)
)
def _lrnForDeck(self, did: int) -> int:
cnt = (
self.col.db.scalar(
f"""
select sum(left/1000) from
(select left from cards where did = ? and queue = {QUEUE_TYPE_LRN} and due < ? limit ?)""",
did,
intTime() + self.col.conf["collapseTime"],
self.reportLimit,
)
or 0
)
return cnt + self.col.db.scalar(
f"""
select count() from
(select 1 from cards where did = ? and queue = {QUEUE_TYPE_DAY_LEARN_RELEARN}
and due <= ? limit ?)""",
did,
self.today,
self.reportLimit,
)
# Reviews
##########################################################################
def _deckRevLimit(self, did: int) -> int:
return self._deckNewLimit(did, self._deckRevLimitSingle)
def _resetRev(self) -> None:
self._revQueue: List[Any] = []
self._revDids = self.col.decks.active()[:]
def _fillRev(self, recursing: bool = False) -> bool:
"True if a review card can be fetched."
if self._revQueue:
return True
if not self.revCount:
return False
while self._revDids:
did = self._revDids[0]
lim = min(self.queueLimit, self._deckRevLimit(did))
if lim:
# fill the queue with the current did
self._revQueue = self.col.db.list(
f"""
select id from cards where
did = ? and queue = {QUEUE_TYPE_REV} and due <= ? limit ?""",
did,
self.today,
lim,
)
if self._revQueue:
# ordering
if self.col.decks.get(did)["dyn"]:
# dynamic decks need due order preserved
self._revQueue.reverse()
else:
# random order for regular reviews
r = random.Random()
r.seed(self.today)
r.shuffle(self._revQueue)
# is the current did empty?
if len(self._revQueue) < lim:
self._revDids.pop(0)
return True
# nothing left in the deck; move to next
self._revDids.pop(0)
# if we didn't get a card but the count is non-zero,
# we need to check again for any cards that were
# removed from the queue but not buried
if recursing:
print("bug: fillRev()")
return False
self._reset_counts()
self._resetRev()
return self._fillRev(recursing=True)
# Answering a review card
##########################################################################
def _answerRevCard(self, card: Card, ease: int) -> None:
delay: int = 0
if ease == BUTTON_ONE:
delay = self._rescheduleLapse(card)
else:
self._rescheduleRev(card, ease)
self._logRev(card, ease, delay, REVLOG_REV)
def _rescheduleLapse(self, card: Card) -> int:
conf = self._lapseConf(card)
card.lastIvl = card.ivl
if self._resched(card):
card.lapses += 1
card.ivl = self._nextLapseIvl(card, conf)
card.factor = max(1300, card.factor - 200)
card.due = self.today + card.ivl
# if it's a filtered deck, update odue as well
if card.odid:
card.odue = card.due
# if suspended as a leech, nothing to do
delay: int = 0
if self._checkLeech(card, conf) and card.queue == QUEUE_TYPE_SUSPENDED:
return delay
# if no relearning steps, nothing to do
if not conf["delays"]:
return delay
# record rev due date for later
if not card.odue:
card.odue = card.due
delay = self._delayForGrade(conf, 0)
card.due = int(delay + time.time())
card.left = self._startingLeft(card)
# queue 1
if card.due < self.dayCutoff:
self.lrnCount += card.left // 1000
card.queue = QUEUE_TYPE_LRN
heappush(self._lrnQueue, (card.due, card.id))
else:
# day learn queue
ahead = ((card.due - self.dayCutoff) // 86400) + 1
card.due = self.today + ahead
card.queue = QUEUE_TYPE_DAY_LEARN_RELEARN
return delay
def _nextLapseIvl(self, card: Card, conf: QueueConfig) -> int:
return max(conf["minInt"], int(card.ivl * conf["mult"]))
def _rescheduleRev(self, card: Card, ease: int) -> None: # type: ignore[override]
# update interval
card.lastIvl = card.ivl
if self._resched(card):
self._updateRevIvl(card, ease)
# then the rest
card.factor = max(1300, card.factor + [-150, 0, 150][ease - 2])
card.due = self.today + card.ivl
else:
card.due = card.odue
if card.odid:
card.did = card.odid
card.odid = 0
card.odue = 0
# Interval management
##########################################################################
def _nextRevIvl(self, card: Card, ease: int) -> int: # type: ignore[override]
"Ideal next interval for CARD, given EASE."
delay = self._daysLate(card)
conf = self._revConf(card)
fct = card.factor / 1000
ivl2 = self._constrainedIvl((card.ivl + delay // 4) * 1.2, conf, card.ivl)
ivl3 = self._constrainedIvl((card.ivl + delay // 2) * fct, conf, ivl2)
ivl4 = self._constrainedIvl(
(card.ivl + delay) * fct * conf["ease4"], conf, ivl3
)
if ease == BUTTON_TWO:
interval = ivl2
elif ease == BUTTON_THREE:
interval = ivl3
elif ease == BUTTON_FOUR:
interval = ivl4
# interval capped?
return min(interval, conf["maxIvl"])
def _constrainedIvl(self, ivl: float, conf: QueueConfig, prev: int) -> int: # type: ignore[override]
"Integer interval after interval factor and prev+1 constraints applied."
new = ivl * conf.get("ivlFct", 1)
return int(max(new, prev + 1))
def _updateRevIvl(self, card: Card, ease: int) -> None:
idealIvl = self._nextRevIvl(card, ease)
card.ivl = min(
max(self._adjRevIvl(card, idealIvl), card.ivl + 1),
self._revConf(card)["maxIvl"],
)
def _adjRevIvl(self, card: Card, idealIvl: int) -> int:
if self._spreadRev:
idealIvl = self._fuzzedIvl(idealIvl)
return idealIvl
# Filtered deck handling
##########################################################################
def _dynIvlBoost(self, card: Card) -> int:
assert card.odid and card.type == CARD_TYPE_REV
assert card.factor
elapsed = card.ivl - (card.odue - self.today)
factor = ((card.factor / 1000) + 1.2) / 2
ivl = int(max(card.ivl, elapsed * factor, 1))
conf = self._revConf(card)
return min(conf["maxIvl"], ivl)
# Leeches
##########################################################################
def _checkLeech(self, card: Card, conf: QueueConfig) -> bool:
"Leech handler. True if card was a leech."
lf = conf["leechFails"]
if not lf:
return False
# if over threshold or every half threshold reps after that
if card.lapses >= lf and (card.lapses - lf) % (max(lf // 2, 1)) == 0:
# add a leech tag
f = card.note()
f.addTag("leech")
f.flush()
# handle
a = conf["leechAction"]
if a == LEECH_SUSPEND:
# if it has an old due, remove it from cram/relearning
if card.odue:
card.due = card.odue
if card.odid:
card.did = card.odid
card.odue = card.odid = 0
card.queue = QUEUE_TYPE_SUSPENDED
# notify UI
hooks.card_did_leech(card)
return True
else:
return False
# Tools
##########################################################################
def _newConf(self, card: Card) -> QueueConfig:
conf = self._cardConf(card)
# normal deck
if not card.odid:
return conf["new"]
# dynamic deck; override some attributes, use original deck for others
oconf = self.col.decks.confForDid(card.odid)
delays = conf["delays"] or oconf["new"]["delays"]
return dict(
# original deck
ints=oconf["new"]["ints"],
initialFactor=oconf["new"]["initialFactor"],
bury=oconf["new"].get("bury", True),
# overrides
delays=delays,
order=NEW_CARDS_DUE,
perDay=self.reportLimit,
)
def _lapseConf(self, card: Card) -> QueueConfig:
conf = self._cardConf(card)
# normal deck
if not card.odid:
return conf["lapse"]
# dynamic deck; override some attributes, use original deck for others
oconf = self.col.decks.confForDid(card.odid)
delays = conf["delays"] or oconf["lapse"]["delays"]
return dict(
# original deck
minInt=oconf["lapse"]["minInt"],
leechFails=oconf["lapse"]["leechFails"],
leechAction=oconf["lapse"]["leechAction"],
mult=oconf["lapse"]["mult"],
# overrides
delays=delays,
resched=conf["resched"],
)
def _resched(self, card: Card) -> bool:
conf = self._cardConf(card)
if not conf["dyn"]:
return True
return conf["resched"]
# Deck finished state
##########################################################################
def haveBuried(self) -> bool:
sdids = self._deckLimit()
cnt = self.col.db.scalar(
f"select 1 from cards where queue = {QUEUE_TYPE_SIBLING_BURIED} and did in %s limit 1"
% sdids
)
return not not cnt
# Next time reports
##########################################################################
def nextIvl(self, card: Card, ease: int) -> float:
"Return the next interval for CARD, in seconds."
if card.queue in (QUEUE_TYPE_NEW, QUEUE_TYPE_LRN, QUEUE_TYPE_DAY_LEARN_RELEARN):
return self._nextLrnIvl(card, ease)
elif ease == BUTTON_ONE:
# lapsed
conf = self._lapseConf(card)
if conf["delays"]:
return conf["delays"][0] * 60
return self._nextLapseIvl(card, conf) * 86400
else:
# review
return self._nextRevIvl(card, ease) * 86400
# this isn't easily extracted from the learn code
def _nextLrnIvl(self, card: Card, ease: int) -> float:
if card.queue == 0:
card.left = self._startingLeft(card)
conf = self._lrnConf(card)
if ease == BUTTON_ONE:
# fail
return self._delayForGrade(conf, len(conf["delays"]))
elif ease == BUTTON_THREE:
# early removal
if not self._resched(card):
return 0
return self._graduatingIvl(card, conf, True, adj=False) * 86400
else:
left = card.left % 1000 - 1
if left <= 0:
# graduate
if not self._resched(card):
return 0
return self._graduatingIvl(card, conf, False, adj=False) * 86400
else:
return self._delayForGrade(conf, left)
Reference in a new issue View git blame Copy permalink