anki/pylib/anki/collection.py

# Copyright: Ankitects Pty Ltd and contributors
# License: GNU AGPL, version 3 or later; http://www.gnu.org/licenses/agpl.html

from __future__ import annotations

from typing import Any, Generator, List, Literal, Optional, Sequence, Tuple, Union

import anki._backend.backend_pb2 as _pb

# protobuf we publicly export - listed first to avoid circular imports
SearchNode = _pb.SearchNode
Progress = _pb.Progress
EmptyCardsReport = _pb.EmptyCardsReport
GraphPreferences = _pb.GraphPreferences
BuiltinSort = _pb.SortOrder.Builtin
Preferences = _pb.Preferences
UndoStatus = _pb.UndoStatus
OpChanges = _pb.OpChanges
OpChangesWithCount = _pb.OpChangesWithCount
OpChangesWithID = _pb.OpChangesWithID
DefaultsForAdding = _pb.DeckAndNotetype
BrowserRow = _pb.BrowserRow

import copy
import os
import pprint
import re
import sys
import time
import traceback
import weakref
from dataclasses import dataclass, field

import anki.latex
from anki import hooks
from anki._backend import RustBackend
from anki.cards import Card, CardID
from anki.config import Config, ConfigManager
from anki.consts import *
from anki.dbproxy import DBProxy
from anki.decks import DeckManager
from anki.errors import AnkiError, DBError
from anki.lang import TR, FormatTimeSpan
from anki.media import MediaManager, media_paths_from_col_path
from anki.models import ModelManager, NoteType
from anki.notes import Note, NoteID
from anki.scheduler.v1 import Scheduler as V1Scheduler
from anki.scheduler.v2 import Scheduler as V2Scheduler
from anki.scheduler.v3 import Scheduler as V3TestScheduler
from anki.sync import SyncAuth, SyncOutput, SyncStatus
from anki.tags import TagManager
from anki.types import assert_exhaustive
from anki.utils import (
    devMode,
    from_json_bytes,
    ids2str,
    intTime,
    splitFields,
    stripHTMLMedia,
)

anki.latex.setup_hook()


SearchJoiner = Literal["AND", "OR"]


@dataclass
class ReviewUndo:
    card: Card
    was_leech: bool


@dataclass
class Checkpoint:
    name: str


@dataclass
class BackendUndo:
    name: str


UndoResult = Union[None, BackendUndo, Checkpoint, ReviewUndo]


class Collection:
    sched: Union[V1Scheduler, V2Scheduler]

    def __init__(
        self,
        path: str,
        backend: Optional[RustBackend] = None,
        server: bool = False,
        log: bool = False,
    ) -> None:
        self._backend = backend or RustBackend(server=server)
        self.db: Optional[DBProxy] = None
        self._should_log = log
        self.server = server
        self.path = os.path.abspath(path)
        self.reopen()

        self.media = MediaManager(self, server)
        self.models = ModelManager(self)
        self.decks = DeckManager(self)
        self.tags = TagManager(self)
        self.conf = ConfigManager(self)
        self._loadScheduler()

    def name(self) -> Any:
        return os.path.splitext(os.path.basename(self.path))[0]

    def weakref(self) -> Collection:
        "Shortcut to create a weak reference that doesn't break code completion."
        return weakref.proxy(self)

    @property
    def backend(self) -> RustBackend:
        traceback.print_stack(file=sys.stdout)
        print()
        print(
            "Accessing the backend directly will break in the future. Please use the public methods on Collection instead."
        )
        return self._backend

    # I18n/messages
    ##########################################################################

    def tr(self, key: TR.V, **kwargs: Union[str, int, float]) -> str:
        return self._backend.translate(key, **kwargs)

    def format_timespan(
        self,
        seconds: float,
        context: FormatTimeSpan.Context.V = FormatTimeSpan.INTERVALS,
    ) -> str:
        return self._backend.format_timespan(seconds=seconds, context=context)

    # Progress
    ##########################################################################

    def latest_progress(self) -> Progress:
        return self._backend.latest_progress()

    # Scheduler
    ##########################################################################

    supportedSchedulerVersions = (1, 2)

    def schedVer(self) -> Any:
        ver = self.conf.get("schedVer", 1)
        if ver in self.supportedSchedulerVersions:
            return ver
        else:
            raise Exception("Unsupported scheduler version")

    def _loadScheduler(self) -> None:
        ver = self.schedVer()
        if ver == 1:
            self.sched = V1Scheduler(self)
        elif ver == 2:
            if self.is_2021_test_scheduler_enabled():
                self.sched = V3TestScheduler(self)  # type: ignore
            else:
                self.sched = V2Scheduler(self)

    def upgrade_to_v2_scheduler(self) -> None:
        self._backend.upgrade_scheduler()
        self.clear_python_undo()
        self._loadScheduler()

    def is_2021_test_scheduler_enabled(self) -> bool:
        return self.get_config_bool(Config.Bool.SCHED_2021)

    def set_2021_test_scheduler_enabled(self, enabled: bool) -> None:
        if self.is_2021_test_scheduler_enabled() != enabled:
            self.set_config_bool(Config.Bool.SCHED_2021, enabled)
            self._loadScheduler()

    # DB-related
    ##########################################################################

    # legacy properties; these will likely go away in the future

    @property
    def crt(self) -> int:
        return self.db.scalar("select crt from col")

    @crt.setter
    def crt(self, crt: int) -> None:
        self.db.execute("update col set crt = ?", crt)

    @property
    def mod(self) -> int:
        return self.db.scalar("select mod from col")

    # legacy
    def setMod(self) -> None:
        # this is now a no-op, as modifications to things like the config
        # will mark the collection modified automatically
        pass

    flush = setMod

    def modified_by_backend(self) -> bool:
        # Until we can move away from long-running transactions, the Python
        # code needs to know if the transaction should be committed, so we need
        # to check if the backend updated the modification time.
        return self.db.last_begin_at != self.mod

    def save(self, name: Optional[str] = None, trx: bool = True) -> None:
        "Flush, commit DB, and take out another write lock if trx=True."
        # commit needed?
        if self.db.modified_in_python or self.modified_by_backend():
            self.db.modified_in_python = False
            self.db.commit()
            if trx:
                self.db.begin()
        elif not trx:
            # if no changes were pending but calling code expects to be
            # outside of a transaction, we need to roll back
            self.db.rollback()

        self._save_checkpoint(name)

    def autosave(self) -> None:
        """Save any pending changes.
        If a checkpoint was taken in the last 5 minutes, don't save."""
        if not self._have_outstanding_checkpoint():
            # if there's no active checkpoint, we can save immediately
            self.save()
        elif time.time() - self._last_checkpoint_at > 300:
            self.save()

    def close(self, save: bool = True, downgrade: bool = False) -> None:
        "Disconnect from DB."
        if self.db:
            if save:
                self.save(trx=False)
            else:
                self.db.rollback()
            self._clear_caches()
            self._backend.close_collection(downgrade_to_schema11=downgrade)
            self.db = None
            self.media.close()
            self._closeLog()

    def close_for_full_sync(self) -> None:
        # save and cleanup, but backend will take care of collection close
        if self.db:
            self.save(trx=False)
            self._clear_caches()
            self.db = None
            self.media.close()
            self._closeLog()

    def rollback(self) -> None:
        self._clear_caches()
        self.db.rollback()
        self.db.begin()

    def _clear_caches(self) -> None:
        self.models._clear_cache()

    def reopen(self, after_full_sync: bool = False) -> None:
        assert not self.db
        assert self.path.endswith(".anki2")

        self._last_checkpoint_at = time.time()
        self._undo: _UndoInfo = None

        (media_dir, media_db) = media_paths_from_col_path(self.path)

        log_path = ""
        should_log = not self.server and self._should_log
        if should_log:
            log_path = self.path.replace(".anki2", "2.log")

        # connect
        if not after_full_sync:
            self._backend.open_collection(
                collection_path=self.path,
                media_folder_path=media_dir,
                media_db_path=media_db,
                log_path=log_path,
            )
        else:
            self.media.connect()
        self.db = DBProxy(weakref.proxy(self._backend))
        self.db.begin()

        self._openLog()

    def modSchema(self, check: bool) -> None:
        "Mark schema modified. Call this first so user can abort if necessary."
        if not self.schemaChanged():
            if check and not hooks.schema_will_change(proceed=True):
                raise AnkiError("abortSchemaMod")
        self.db.execute("update col set scm=?", intTime(1000))
        self.save()

    def schemaChanged(self) -> bool:
        "True if schema changed since last sync."
        return self.db.scalar("select scm > ls from col")

    def usn(self) -> int:
        if self.server:
            return self.db.scalar("select usn from col")
        else:
            return -1

    def beforeUpload(self) -> None:
        "Called before a full upload."
        self.save(trx=False)
        self._backend.before_upload()
        self.close(save=False, downgrade=True)

    # Object creation helpers
    ##########################################################################

    def get_card(self, id: CardID) -> Card:
        return Card(self, id)

    def update_card(self, card: Card) -> None:
        """Save card changes to database, and add an undo entry.
        Unlike card.flush(), this will invalidate any current checkpoint."""
        self._backend.update_card(card=card._to_backend_card(), skip_undo_entry=False)

    def get_note(self, id: NoteID) -> Note:
        return Note(self, id=id)

    def update_note(self, note: Note) -> OpChanges:
        """Save note changes to database, and add an undo entry.
        Unlike note.flush(), this will invalidate any current checkpoint."""
        return self._backend.update_note(
            note=note._to_backend_note(), skip_undo_entry=False
        )

    getCard = get_card
    getNote = get_note

    # Utils
    ##########################################################################

    def nextID(self, type: str, inc: bool = True) -> Any:
        type = f"next{type.capitalize()}"
        id = self.conf.get(type, 1)
        if inc:
            self.conf[type] = id + 1
        return id

    def reset(self) -> None:
        "Rebuild the queue and reload data after DB modified."
        self.autosave()
        self.sched.reset()

    # Deletion logging
    ##########################################################################

    def _logRem(self, ids: List[Union[int, NoteID]], type: int) -> None:
        self.db.executemany(
            "insert into graves values (%d, ?, %d)" % (self.usn(), type),
            ([x] for x in ids),
        )

    # Notes
    ##########################################################################

    def new_note(self, notetype: NoteType) -> Note:
        return Note(self, notetype)

    def add_note(self, note: Note, deck_id: int) -> OpChanges:
        out = self._backend.add_note(note=note._to_backend_note(), deck_id=deck_id)
        note.id = NoteID(out.note_id)
        return out.changes

    def remove_notes(self, note_ids: Sequence[NoteID]) -> OpChanges:
        hooks.notes_will_be_deleted(self, note_ids)
        return self._backend.remove_notes(note_ids=note_ids, card_ids=[])

    def remove_notes_by_card(self, card_ids: List[CardID]) -> None:
        if hooks.notes_will_be_deleted.count():
            nids = self.db.list(
                f"select nid from cards where id in {ids2str(card_ids)}"
            )
            hooks.notes_will_be_deleted(self, nids)
        self._backend.remove_notes(note_ids=[], card_ids=card_ids)

    def card_ids_of_note(self, note_id: NoteID) -> Sequence[CardID]:
        return [CardID(id) for id in self._backend.cards_of_note(note_id)]

    def defaults_for_adding(
        self, *, current_review_card: Optional[Card]
    ) -> DefaultsForAdding:
        """Get starting deck and notetype for add screen.
        An option in the preferences controls whether this will be based on the current deck
        or current notetype.
        """
        if card := current_review_card:
            home_deck = card.currentDeckID()
        else:
            home_deck = 0

        return self._backend.defaults_for_adding(
            home_deck_of_current_review_card=home_deck,
        )

    def default_deck_for_notetype(self, notetype_id: NoteID) -> Optional[int]:
        """If 'change deck depending on notetype' is enabled in the preferences,
        return the last deck used with the provided notetype, if any.."""
        if self.get_config_bool(Config.Bool.ADDING_DEFAULTS_TO_CURRENT_DECK):
            return None

        return (
            self._backend.default_deck_for_notetype(
                ntid=notetype_id,
            )
            or None
        )

    # legacy

    def noteCount(self) -> int:
        return self.db.scalar("select count() from notes")

    def newNote(self, forDeck: bool = True) -> Note:
        "Return a new note with the current model."
        return Note(self, self.models.current(forDeck))

    def addNote(self, note: Note) -> int:
        self.add_note(note, note.model()["did"])
        return len(note.cards())

    def remNotes(self, ids: Sequence[NoteID]) -> None:
        self.remove_notes(ids)

    def _remNotes(self, ids: List[NoteID]) -> None:
        pass

    # Cards
    ##########################################################################

    def isEmpty(self) -> bool:
        return not self.db.scalar("select 1 from cards limit 1")

    def cardCount(self) -> Any:
        return self.db.scalar("select count() from cards")

    def remove_cards_and_orphaned_notes(self, card_ids: Sequence[CardID]) -> None:
        "You probably want .remove_notes_by_card() instead."
        self._backend.remove_cards(card_ids=card_ids)

    def set_deck(self, card_ids: Sequence[CardID], deck_id: int) -> OpChanges:
        return self._backend.set_deck(card_ids=card_ids, deck_id=deck_id)

    def get_empty_cards(self) -> EmptyCardsReport:
        return self._backend.get_empty_cards()

    # legacy

    def remCards(self, ids: List[CardID], notes: bool = True) -> None:
        self.remove_cards_and_orphaned_notes(ids)

    def emptyCids(self) -> List[CardID]:
        print("emptyCids() will go away")
        return []

    # Card generation & field checksums/sort fields
    ##########################################################################

    def after_note_updates(
        self, nids: List[NoteID], mark_modified: bool, generate_cards: bool = True
    ) -> None:
        self._backend.after_note_updates(
            nids=nids, generate_cards=generate_cards, mark_notes_modified=mark_modified
        )

    # legacy

    def updateFieldCache(self, nids: List[NoteID]) -> None:
        self.after_note_updates(nids, mark_modified=False, generate_cards=False)

    # this also updates field cache
    def genCards(self, nids: List[NoteID]) -> List[int]:
        self.after_note_updates(nids, mark_modified=False, generate_cards=True)
        # previously returned empty cards, no longer does
        return []

    # Finding cards
    ##########################################################################

    def find_cards(
        self,
        query: str,
        order: Union[bool, str, BuiltinSort.Kind.V] = False,
        reverse: bool = False,
    ) -> Sequence[CardID]:
        """Return card ids matching the provided search.

        To programmatically construct a search string, see .build_search_string().

        If order=True, use the sort order stored in the collection config
        If order=False, do no ordering

        If order is a string, that text is added after 'order by' in the sql statement.
        You must add ' asc' or ' desc' to the order, as Anki will replace asc with
        desc and vice versa when reverse is set in the collection config, eg
        order="c.ivl asc, c.due desc".

        If order is a BuiltinSort.Kind value, sort using that builtin sort, eg
        col.find_cards("", order=BuiltinSort.Kind.CARD_DUE)

        The reverse argument only applies when a BuiltinSort.Kind is provided;
        otherwise the collection config defines whether reverse is set or not.
        """
        if isinstance(order, str):
            mode = _pb.SortOrder(custom=order)
        elif isinstance(order, bool):
            if order is True:
                mode = _pb.SortOrder(from_config=_pb.Empty())
            else:
                mode = _pb.SortOrder(none=_pb.Empty())
        else:
            mode = _pb.SortOrder(
                builtin=_pb.SortOrder.Builtin(kind=order, reverse=reverse)
            )
        return [
            CardID(id) for id in self._backend.search_cards(search=query, order=mode)
        ]

    def find_notes(self, *terms: Union[str, SearchNode]) -> Sequence[NoteID]:
        """Return note ids matching the provided search or searches.

        If more than one search is provided, they will be ANDed together.

        Eg: col.find_notes("test", "another") will search for "test AND another"
        and return matching note ids.

        Eg: col.find_notes(SearchNode(deck="test"), "foo") will return notes
        that have a card in deck called "test", and have the text "foo".
        """
        return [
            NoteID(did)
            for did in self._backend.search_notes(self.build_search_string(*terms))
        ]

    def find_and_replace(
        self,
        *,
        note_ids: Sequence[NoteID],
        search: str,
        replacement: str,
        regex: bool = False,
        field_name: Optional[str] = None,
        match_case: bool = False,
    ) -> OpChangesWithCount:
        "Find and replace fields in a note. Returns changed note count."
        return self._backend.find_and_replace(
            nids=note_ids,
            search=search,
            replacement=replacement,
            regex=regex,
            match_case=match_case,
            field_name=field_name or "",
        )

    def field_names_for_note_ids(self, nids: Sequence[int]) -> Sequence[str]:
        return self._backend.field_names_for_notes(nids)

    # returns array of ("dupestr", [nids])
    def findDupes(self, fieldName: str, search: str = "") -> List[Tuple[Any, list]]:
        nids = self.findNotes(search, SearchNode(field_name=fieldName))
        # go through notes
        vals: Dict[str, List[int]] = {}
        dupes = []
        fields: Dict[int, int] = {}

        def ordForMid(mid: int) -> int:
            if mid not in fields:
                model = self.models.get(mid)
                for c, f in enumerate(model["flds"]):
                    if f["name"].lower() == fieldName.lower():
                        fields[mid] = c
                        break
            return fields[mid]

        for nid, mid, flds in self.db.all(
            f"select id, mid, flds from notes where id in {ids2str(nids)}"
        ):
            flds = splitFields(flds)
            ord = ordForMid(mid)
            if ord is None:
                continue
            val = flds[ord]
            val = stripHTMLMedia(val)
            # empty does not count as duplicate
            if not val:
                continue
            vals.setdefault(val, []).append(nid)
            if len(vals[val]) == 2:
                dupes.append((val, vals[val]))
        return dupes

    findCards = find_cards
    findNotes = find_notes
    findReplace = find_and_replace

    # Search Strings
    ##########################################################################

    def build_search_string(
        self,
        *nodes: Union[str, SearchNode],
        joiner: SearchJoiner = "AND",
    ) -> str:
        """Join one or more searches, and return a normalized search string.

        To negate, wrap in a negated search term:

            term = SearchNode(negated=col.group_searches(...))

        Invalid searches will throw an exception.
        """
        term = self.group_searches(*nodes, joiner=joiner)
        return self._backend.build_search_string(term)

    def group_searches(
        self,
        *nodes: Union[str, SearchNode],
        joiner: SearchJoiner = "AND",
    ) -> SearchNode:
        """Join provided search nodes and strings into a single SearchNode.
        If a single SearchNode is provided, it is returned as-is.
        At least one node must be provided.
        """
        assert nodes

        # convert raw text to SearchNodes
        search_nodes = [
            node if isinstance(node, SearchNode) else SearchNode(parsable_text=node)
            for node in nodes
        ]

        # if there's more than one, wrap them in a group
        if len(search_nodes) > 1:
            return SearchNode(
                group=SearchNode.Group(
                    nodes=search_nodes, joiner=self._pb_search_separator(joiner)
                )
            )
        else:
            return search_nodes[0]

    def join_searches(
        self,
        existing_node: SearchNode,
        additional_node: SearchNode,
        operator: Literal["AND", "OR"],
    ) -> str:
        """
        AND or OR `additional_term` to `existing_term`, without wrapping `existing_term` in brackets.
        Used by the Browse screen to avoid adding extra brackets when joining.
        If you're building a search query yourself, you probably don't need this.
        """
        search_string = self._backend.join_search_nodes(
            joiner=self._pb_search_separator(operator),
            existing_node=existing_node,
            additional_node=additional_node,
        )

        return search_string

    def replace_in_search_node(
        self, existing_node: SearchNode, replacement_node: SearchNode
    ) -> str:
        """If nodes of the same type as `replacement_node` are found in existing_node, replace them.

        You can use this to replace any "deck" clauses in a search with a different deck for example.
        """
        return self._backend.replace_search_node(
            existing_node=existing_node, replacement_node=replacement_node
        )

    def _pb_search_separator(self, operator: SearchJoiner) -> SearchNode.Group.Joiner.V:
        # pylint: disable=no-member
        if operator == "AND":
            return SearchNode.Group.Joiner.AND
        else:
            return SearchNode.Group.Joiner.OR

    # Browser rows
    ##########################################################################

    def browser_row_for_card(
        self, cid: int
    ) -> Tuple[Generator[Tuple[str, bool], None, None], BrowserRow.Color.V, str, int]:
        row = self._backend.browser_row_for_card(cid)
        return (
            ((cell.text, cell.is_rtl) for cell in row.cells),
            row.color,
            row.font_name,
            row.font_size,
        )

    # Config
    ##########################################################################

    def get_config(self, key: str, default: Any = None) -> Any:
        try:
            return self.conf.get_immutable(key)
        except KeyError:
            return default

    def set_config(self, key: str, val: Any) -> None:
        self.conf.set(key, val)

    def remove_config(self, key: str) -> None:
        self.conf.remove(key)

    def all_config(self) -> Dict[str, Any]:
        "This is a debugging aid. Prefer .get_config() when you know the key you need."
        return from_json_bytes(self._backend.get_all_config())

    def get_config_bool(self, key: Config.Bool.Key.V) -> bool:
        return self._backend.get_config_bool(key)

    def set_config_bool(self, key: Config.Bool.Key.V, value: bool) -> None:
        self._backend.set_config_bool(key=key, value=value)

    def get_config_string(self, key: Config.String.Key.V) -> str:
        return self._backend.get_config_string(key)

    def set_config_string(self, key: Config.String.Key.V, value: str) -> None:
        self._backend.set_config_string(key=key, value=value)

    # Stats
    ##########################################################################

    def stats(self) -> "anki.stats.CollectionStats":
        from anki.stats import CollectionStats

        return CollectionStats(self)

    def card_stats(self, card_id: CardID, include_revlog: bool) -> str:
        import anki.stats as st

        if include_revlog:
            revlog_style = "margin-top: 2em;"
        else:
            revlog_style = "display: none;"

        style = f"""<style>
.revlog-learn {{ color: {st.colLearn} }}
.revlog-review {{ color: {st.colMature} }}
.revlog-relearn {{ color: {st.colRelearn} }}
.revlog-ease1 {{ color: {st.colRelearn} }}
table.review-log {{ {revlog_style} }}
</style>"""

        return style + self._backend.card_stats(card_id)

    def studied_today(self) -> str:
        return self._backend.studied_today()

    def graph_data(self, search: str, days: int) -> bytes:
        return self._backend.graphs(search=search, days=days)

    def get_graph_preferences(self) -> bytes:
        return self._backend.get_graph_preferences()

    def set_graph_preferences(self, prefs: GraphPreferences) -> None:
        self._backend.set_graph_preferences(input=prefs)

    def congrats_info(self) -> bytes:
        "Don't use this, it will likely go away in the future."
        return self._backend.congrats_info().SerializeToString()

    # legacy

    def cardStats(self, card: Card) -> str:
        return self.card_stats(card.id, include_revlog=False)

    # Timeboxing
    ##########################################################################
    # fixme: there doesn't seem to be a good reason why this code is in main.py
    # instead of covered in reviewer, and the reps tracking is covered by both
    # the scheduler and reviewer.py. in the future, we should probably move
    # reps tracking to reviewer.py, and remove the startTimebox() calls from
    # other locations like overview.py. We just need to make sure not to reset
    # the count on things like edits, which we probably could do by checking
    # the previous state in moveToState.

    def startTimebox(self) -> None:
        self._startTime = time.time()
        self._startReps = self.sched.reps

    def timeboxReached(self) -> Union[Literal[False], Tuple[Any, int]]:
        "Return (elapsedTime, reps) if timebox reached, or False."
        if not self.conf["timeLim"]:
            # timeboxing disabled
            return False
        elapsed = time.time() - self._startTime
        if elapsed > self.conf["timeLim"]:
            return (self.conf["timeLim"], self.sched.reps - self._startReps)
        return False

    # Undo
    ##########################################################################

    def undo_status(self) -> UndoStatus:
        "Return the undo status. At the moment, redo is not supported."
        # check backend first
        if status := self._check_backend_undo_status():
            return status

        if not self._undo:
            return UndoStatus()

        if isinstance(self._undo, _ReviewsUndo):
            return UndoStatus(undo=self.tr(TR.SCHEDULING_REVIEW))
        elif isinstance(self._undo, Checkpoint):
            return UndoStatus(undo=self._undo.name)
        else:
            assert_exhaustive(self._undo)
            assert False

    def clear_python_undo(self) -> None:
        """Clear the Python undo state.
        The backend will automatically clear backend undo state when
        any SQL DML is executed, or an operation that doesn't support undo
        is run."""
        self._undo = None

    def undo(self) -> UndoResult:
        """Returns ReviewUndo if undoing a v1/v2 scheduler review.
        Returns None if the undo queue was empty."""
        # backend?
        status = self._backend.get_undo_status()
        if status.undo:
            self._backend.undo()
            self.clear_python_undo()
            return BackendUndo(name=status.undo)

        if isinstance(self._undo, _ReviewsUndo):
            return self._undo_review()
        elif isinstance(self._undo, Checkpoint):
            return self._undo_checkpoint()
        elif self._undo is None:
            return None
        else:
            assert_exhaustive(self._undo)
            assert False

    def op_affects_study_queue(self, changes: OpChanges) -> bool:
        if changes.kind == changes.SET_CARD_FLAG:
            return False
        return changes.card or changes.deck or changes.preference

    def op_made_changes(self, changes: OpChanges) -> bool:
        for field in changes.DESCRIPTOR.fields:
            if field.name != "kind":
                if getattr(changes, field.name, False):
                    return True
        return False

    def _check_backend_undo_status(self) -> Optional[UndoStatus]:
        """Return undo status if undo available on backend.
        If backend has undo available, clear the Python undo state."""
        status = self._backend.get_undo_status()
        if status.undo or status.redo:
            self.clear_python_undo()
            return status
        else:
            return None

    def save_card_review_undo_info(self, card: Card) -> None:
        "Used by V1 and V2 schedulers to record state prior to review."
        if not isinstance(self._undo, _ReviewsUndo):
            self._undo = _ReviewsUndo()

        was_leech = card.note().has_tag("leech")
        entry = ReviewUndo(card=copy.copy(card), was_leech=was_leech)
        self._undo.entries.append(entry)

    def _have_outstanding_checkpoint(self) -> bool:
        self._check_backend_undo_status()
        return isinstance(self._undo, Checkpoint)

    def _undo_checkpoint(self) -> Checkpoint:
        assert isinstance(self._undo, Checkpoint)
        self.rollback()
        undo = self._undo
        self.clear_python_undo()
        return undo

    def _save_checkpoint(self, name: Optional[str]) -> None:
        "Call via .save(). If name not provided, clear any existing checkpoint."
        self._last_checkpoint_at = time.time()
        if name:
            self._undo = Checkpoint(name=name)
        else:
            # saving disables old checkpoint, but not review undo
            if not isinstance(self._undo, _ReviewsUndo):
                self.clear_python_undo()

    def _undo_review(self) -> ReviewUndo:
        "Undo a v1/v2 review."
        assert isinstance(self._undo, _ReviewsUndo)
        entry = self._undo.entries.pop()
        if not self._undo.entries:
            self.clear_python_undo()

        card = entry.card

        # remove leech tag if it didn't have it before
        if not entry.was_leech and card.note().has_tag("leech"):
            card.note().remove_tag("leech")
            card.note().flush()

        # write old data
        card.flush()

        # and delete revlog entry if not previewing
        conf = self.sched._cardConf(card)
        previewing = conf["dyn"] and not conf["resched"]
        if not previewing:
            last = self.db.scalar(
                "select id from revlog where cid = ? " "order by id desc limit 1",
                card.id,
            )
            self.db.execute("delete from revlog where id = ?", last)

        # restore any siblings
        self.db.execute(
            "update cards set queue=type,mod=?,usn=? where queue=-2 and nid=?",
            intTime(),
            self.usn(),
            card.nid,
        )

        # update daily counts
        n = card.queue
        if card.queue in (QUEUE_TYPE_DAY_LEARN_RELEARN, QUEUE_TYPE_PREVIEW):
            n = QUEUE_TYPE_LRN
        type = ("new", "lrn", "rev")[n]
        self.sched._updateStats(card, type, -1)
        self.sched.reps -= 1

        # and refresh the queues
        self.sched.reset()

        return entry

    # legacy

    clearUndo = clear_python_undo
    markReview = save_card_review_undo_info

    def undoName(self) -> Optional[str]:
        "Undo menu item name, or None if undo unavailable."
        status = self.undo_status()
        return status.undo or None

    # DB maintenance
    ##########################################################################

    def fixIntegrity(self) -> Tuple[str, bool]:
        """Fix possible problems and rebuild caches.

        Returns tuple of (error: str, ok: bool). 'ok' will be true if no
        problems were found.
        """
        self.save(trx=False)
        try:
            problems = list(self._backend.check_database())
            ok = not problems
            problems.append(self.tr(TR.DATABASE_CHECK_REBUILT))
        except DBError as e:
            problems = [str(e.args[0])]
            ok = False
        finally:
            try:
                self.db.begin()
            except:
                # may fail if the DB is very corrupt
                pass
        return ("\n".join(problems), ok)

    def optimize(self) -> None:
        self.save(trx=False)
        self.db.execute("vacuum")
        self.db.execute("analyze")
        self.db.begin()

    # Logging
    ##########################################################################

    def log(self, *args: Any, **kwargs: Any) -> None:
        if not self._should_log:
            return

        def customRepr(x: Any) -> str:
            if isinstance(x, str):
                return x
            return pprint.pformat(x)

        path, num, fn, y = traceback.extract_stack(limit=2 + kwargs.get("stack", 0))[0]
        buf = "[%s] %s:%s(): %s" % (
            intTime(),
            os.path.basename(path),
            fn,
            ", ".join([customRepr(x) for x in args]),
        )
        self._logHnd.write(f"{buf}\n")
        if devMode:
            print(buf)

    def _openLog(self) -> None:
        if not self._should_log:
            return
        lpath = re.sub(r"\.anki2$", ".log", self.path)
        if os.path.exists(lpath) and os.path.getsize(lpath) > 10 * 1024 * 1024:
            lpath2 = f"{lpath}.old"
            if os.path.exists(lpath2):
                os.unlink(lpath2)
            os.rename(lpath, lpath2)
        self._logHnd = open(lpath, "a", encoding="utf8")

    def _closeLog(self) -> None:
        if not self._should_log:
            return
        self._logHnd.close()
        self._logHnd = None

    ##########################################################################

    def set_user_flag_for_cards(self, flag: int, cids: Sequence[CardID]) -> OpChanges:
        return self._backend.set_flag(card_ids=cids, flag=flag)

    def set_wants_abort(self) -> None:
        self._backend.set_wants_abort()

    def i18n_resources(self) -> bytes:
        return self._backend.i18n_resources()

    def abort_media_sync(self) -> None:
        self._backend.abort_media_sync()

    def abort_sync(self) -> None:
        self._backend.abort_sync()

    def full_upload(self, auth: SyncAuth) -> None:
        self._backend.full_upload(auth)

    def full_download(self, auth: SyncAuth) -> None:
        self._backend.full_download(auth)

    def sync_login(self, username: str, password: str) -> SyncAuth:
        return self._backend.sync_login(username=username, password=password)

    def sync_collection(self, auth: SyncAuth) -> SyncOutput:
        return self._backend.sync_collection(auth)

    def sync_media(self, auth: SyncAuth) -> None:
        self._backend.sync_media(auth)

    def sync_status(self, auth: SyncAuth) -> SyncStatus:
        return self._backend.sync_status(auth)

    def get_preferences(self) -> Preferences:
        return self._backend.get_preferences()

    def set_preferences(self, prefs: Preferences) -> None:
        self._backend.set_preferences(prefs)

    def render_markdown(self, text: str, sanitize: bool = True) -> str:
        "Not intended for public consumption at this time."
        return self._backend.render_markdown(markdown=text, sanitize=sanitize)


# legacy name
_Collection = Collection


@dataclass
class _ReviewsUndo:
    entries: List[ReviewUndo] = field(default_factory=list)


_UndoInfo = Union[_ReviewsUndo, Checkpoint, None]