[plomtask] / plomtask / db.py

"""Database management."""
from __future__ import annotations
from os import listdir
from os.path import isfile
from difflib import Differ
from sqlite3 import connect as sql_connect, Cursor, Row
from typing import Any, Self, TypeVar, Generic
from plomtask.exceptions import HandledException, NotFoundException
from plomtask.dating import valid_date

EXPECTED_DB_VERSION = 5
MIGRATIONS_DIR = 'migrations'
FILENAME_DB_SCHEMA = f'init_{EXPECTED_DB_VERSION}.sql'
PATH_DB_SCHEMA = f'{MIGRATIONS_DIR}/{FILENAME_DB_SCHEMA}'


class UnmigratedDbException(HandledException):
    """To identify case of unmigrated DB file."""


class DatabaseFile:
    """Represents the sqlite3 database's file."""
    # pylint: disable=too-few-public-methods

    def __init__(self, path: str) -> None:
        self.path = path
        self._check()

    @classmethod
    def create_at(cls, path: str) -> DatabaseFile:
        """Make new DB file at path."""
        with sql_connect(path) as conn:
            with open(PATH_DB_SCHEMA, 'r', encoding='utf-8') as f:
                conn.executescript(f.read())
            conn.execute(f'PRAGMA user_version = {EXPECTED_DB_VERSION}')
        return cls(path)

    @classmethod
    def migrate(cls, path: str) -> DatabaseFile:
        """Apply migrations from_version to EXPECTED_DB_VERSION."""
        migrations = cls._available_migrations()
        from_version = cls._get_version_of_db(path)
        migrations_todo = migrations[from_version+1:]
        for j, filename in enumerate(migrations_todo):
            with sql_connect(path) as conn:
                with open(f'{MIGRATIONS_DIR}/{filename}', 'r',
                          encoding='utf-8') as f:
                    conn.executescript(f.read())
            user_version = from_version + j + 1
            with sql_connect(path) as conn:
                conn.execute(f'PRAGMA user_version = {user_version}')
        return cls(path)

    def _check(self) -> None:
        """Check file exists, and is of proper DB version and schema."""
        if not isfile(self.path):
            raise NotFoundException
        if self._user_version != EXPECTED_DB_VERSION:
            raise UnmigratedDbException()
        self._validate_schema()

    @staticmethod
    def _available_migrations() -> list[str]:
        """Validate migrations directory and return sorted entries."""
        msg_too_big = 'Migration directory points beyond expected DB version.'
        msg_bad_entry = 'Migration directory contains unexpected entry: '
        msg_missing = 'Migration directory misses migration of number: '
        migrations = {}
        for entry in listdir(MIGRATIONS_DIR):
            if entry == FILENAME_DB_SCHEMA:
                continue
            toks = entry.split('_', 1)
            if len(toks) < 2:
                raise HandledException(msg_bad_entry + entry)
            try:
                i = int(toks[0])
            except ValueError as e:
                raise HandledException(msg_bad_entry + entry) from e
            if i > EXPECTED_DB_VERSION:
                raise HandledException(msg_too_big)
            migrations[i] = toks[1]
        migrations_list = []
        for i in range(EXPECTED_DB_VERSION + 1):
            if i not in migrations:
                raise HandledException(msg_missing + str(i))
            migrations_list += [f'{i}_{migrations[i]}']
        return migrations_list

    @staticmethod
    def _get_version_of_db(path: str) -> int:
        """Get DB user_version, fail if outside expected range."""
        sql_for_db_version = 'PRAGMA user_version'
        with sql_connect(path) as conn:
            db_version = list(conn.execute(sql_for_db_version))[0][0]
        if db_version > EXPECTED_DB_VERSION:
            msg = f'Wrong DB version, expected '\
                    f'{EXPECTED_DB_VERSION}, got unknown {db_version}.'
            raise HandledException(msg)
        assert isinstance(db_version, int)
        return db_version

    @property
    def _user_version(self) -> int:
        """Get DB user_version."""
        return self._get_version_of_db(self.path)

    def _validate_schema(self) -> None:
        """Compare found schema with what's stored at PATH_DB_SCHEMA."""

        def reformat_rows(rows: list[str]) -> list[str]:
            new_rows = []
            for row in rows:
                new_row = []
                for subrow in row.split('\n'):
                    subrow = subrow.rstrip()
                    in_parentheses = 0
                    split_at = []
                    for i, c in enumerate(subrow):
                        if '(' == c:
                            in_parentheses += 1
                        elif ')' == c:
                            in_parentheses -= 1
                        elif ',' == c and 0 == in_parentheses:
                            split_at += [i + 1]
                    prev_split = 0
                    for i in split_at:
                        segment = subrow[prev_split:i].strip()
                        if len(segment) > 0:
                            new_row += [f'    {segment}']
                        prev_split = i
                    segment = subrow[prev_split:].strip()
                    if len(segment) > 0:
                        new_row += [f'    {segment}']
                new_row[0] = new_row[0].lstrip()
                new_row[-1] = new_row[-1].lstrip()
                if new_row[-1] != ')' and new_row[-3][-1] != ',':
                    new_row[-3] = new_row[-3] + ','
                    new_row[-2:] = ['    ' + new_row[-1][:-1]] + [')']
                new_rows += ['\n'.join(new_row)]
            return new_rows

        sql_for_schema = 'SELECT sql FROM sqlite_master ORDER BY sql'
        msg_err = 'Database has wrong tables schema. Diff:\n'
        with sql_connect(self.path) as conn:
            schema_rows = [r[0] for r in conn.execute(sql_for_schema) if r[0]]
        schema_rows = reformat_rows(schema_rows)
        retrieved_schema = ';\n'.join(schema_rows) + ';'
        with open(PATH_DB_SCHEMA, 'r', encoding='utf-8') as f:
            stored_schema = f.read().rstrip()
        if stored_schema != retrieved_schema:
            diff_msg = Differ().compare(retrieved_schema.splitlines(),
                                        stored_schema.splitlines())
            raise HandledException(msg_err + '\n'.join(diff_msg))


class DatabaseConnection:
    """A single connection to the database."""

    def __init__(self, db_file: DatabaseFile) -> None:
        self.conn = sql_connect(db_file.path)

    def commit(self) -> None:
        """Commit SQL transaction."""
        self.conn.commit()

    def exec(self, code: str, inputs: tuple[Any, ...] = tuple()) -> Cursor:
        """Add commands to SQL transaction."""
        return self.conn.execute(code, inputs)

    def exec_on_vals(self, code: str, inputs: tuple[Any, ...]) -> Cursor:
        """Wrapper around .exec appending adequate " (?, …)" to code."""
        q_marks_from_values = '(' + ','.join(['?'] * len(inputs)) + ')'
        return self.exec(f'{code} {q_marks_from_values}', inputs)

    def close(self) -> None:
        """Close DB connection."""
        self.conn.close()

    def rewrite_relations(self, table_name: str, key: str, target: int | str,
                          rows: list[list[Any]], key_index: int = 0) -> None:
        # pylint: disable=too-many-arguments
        """Rewrite relations in table_name to target, with rows values.

        Note that single rows are expected without the column and value
        identified by key and target, which are inserted inside the function
        at key_index.
        """
        self.delete_where(table_name, key, target)
        for row in rows:
            values = tuple(row[:key_index] + [target] + row[key_index:])
            self.exec_on_vals(f'INSERT INTO {table_name} VALUES', values)

    def row_where(self, table_name: str, key: str,
                  target: int | str) -> list[Row]:
        """Return list of Rows at table where key == target."""
        return list(self.exec(f'SELECT * FROM {table_name} WHERE {key} = ?',
                              (target,)))

    # def column_where_pattern(self,
    #                          table_name: str,
    #                          column: str,
    #                          pattern: str,
    #                          keys: list[str]) -> list[Any]:
    #     """Return column of rows where one of keys matches pattern."""
    #     targets = tuple([f'%{pattern}%'] * len(keys))
    #     haystack = ' OR '.join([f'{k} LIKE ?' for k in keys])
    #     sql = f'SELECT {column} FROM {table_name} WHERE {haystack}'
    #     return [row[0] for row in self.exec(sql, targets)]

    def column_where(self, table_name: str, column: str, key: str,
                     target: int | str) -> list[Any]:
        """Return column of table where key == target."""
        return [row[0] for row in
                self.exec(f'SELECT {column} FROM {table_name} '
                          f'WHERE {key} = ?', (target,))]

    def column_all(self, table_name: str, column: str) -> list[Any]:
        """Return complete column of table."""
        return [row[0] for row in
                self.exec(f'SELECT {column} FROM {table_name}')]

    def delete_where(self, table_name: str, key: str,
                     target: int | str) -> None:
        """Delete from table where key == target."""
        self.exec(f'DELETE FROM {table_name} WHERE {key} = ?', (target,))


BaseModelId = TypeVar('BaseModelId', int, str)
BaseModelInstance = TypeVar('BaseModelInstance', bound='BaseModel[Any]')


class BaseModel(Generic[BaseModelId]):
    """Template for most of the models we use/derive from the DB."""
    table_name = ''
    to_save: list[str] = []
    to_save_versioned: list[str] = []
    to_save_relations: list[tuple[str, str, str, int]] = []
    id_: None | BaseModelId
    cache_: dict[BaseModelId, Self]
    to_search: list[str] = []
    can_create_by_id = False
    _exists = True

    def __init__(self, id_: BaseModelId | None) -> None:
        if isinstance(id_, int) and id_ < 1:
            msg = f'illegal {self.__class__.__name__} ID, must be >=1: {id_}'
            raise HandledException(msg)
        if isinstance(id_, str) and "" == id_:
            msg = f'illegal {self.__class__.__name__} ID, must be non-empty'
            raise HandledException(msg)
        self.id_ = id_

    def __hash__(self) -> int:
        hashable = [self.id_] + [getattr(self, name) for name in self.to_save]
        for definition in self.to_save_relations:
            attr = getattr(self, definition[2])
            hashable += [tuple(rel.id_ for rel in attr)]
        for name in self.to_save_versioned:
            hashable += [hash(getattr(self, name))]
        return hash(tuple(hashable))

    def __eq__(self, other: object) -> bool:
        if not isinstance(other, self.__class__):
            return False
        return hash(self) == hash(other)

    def __lt__(self, other: Any) -> bool:
        if not isinstance(other, self.__class__):
            msg = 'cannot compare to object of different class'
            raise HandledException(msg)
        assert isinstance(self.id_, int)
        assert isinstance(other.id_, int)
        return self.id_ < other.id_

    @property
    def as_dict(self) -> dict[str, object]:
        """Return self as (json.dumps-coompatible) dict."""
        d: dict[str, object] = {'id': self.id_}
        if len(self.to_save_versioned) > 0:
            d['_versioned'] = {}
        for k in self.to_save:
            attr = getattr(self, k)
            if hasattr(attr, 'as_dict'):
                d[k] = attr.as_dict
            d[k] = attr
        for k in self.to_save_versioned:
            attr = getattr(self, k)
            assert isinstance(d['_versioned'], dict)
            d['_versioned'][k] = attr.history
        for r in self.to_save_relations:
            attr_name = r[2]
            d[attr_name] = [x.as_dict for x in getattr(self, attr_name)]
        return d

    # cache management
    # (we primarily use the cache to ensure we work on the same object in
    # memory no matter where and how we retrieve it, e.g. we don't want
    # .by_id() calls to create a new object each time, but rather a pointer
    # to the one already instantiated)

    def __getattribute__(self, name: str) -> Any:
        """Ensure fail if ._disappear() was called, except to check ._exists"""
        if name != '_exists' and not super().__getattribute__('_exists'):
            raise HandledException('Object does not exist.')
        return super().__getattribute__(name)

    def _disappear(self) -> None:
        """Invalidate object, make future use raise exceptions."""
        assert self.id_ is not None
        if self._get_cached(self.id_):
            self._uncache()
        to_kill = list(self.__dict__.keys())
        for attr in to_kill:
            delattr(self, attr)
        self._exists = False

    @classmethod
    def empty_cache(cls) -> None:
        """Empty class's cache, and disappear all former inhabitants."""
        # pylint: disable=protected-access
        # (cause we remain within the class)
        if hasattr(cls, 'cache_'):
            to_disappear = list(cls.cache_.values())
            for item in to_disappear:
                item._disappear()
        cls.cache_ = {}

    @classmethod
    def get_cache(cls: type[BaseModelInstance]) -> dict[Any, BaseModel[Any]]:
        """Get cache dictionary, create it if not yet existing."""
        if not hasattr(cls, 'cache_'):
            d: dict[Any, BaseModel[Any]] = {}
            cls.cache_ = d
        return cls.cache_

    @classmethod
    def _get_cached(cls: type[BaseModelInstance],
                    id_: BaseModelId) -> BaseModelInstance | None:
        """Get object of id_ from class's cache, or None if not found."""
        # pylint: disable=consider-iterating-dictionary
        cache = cls.get_cache()
        if id_ in cache.keys():
            obj = cache[id_]
            assert isinstance(obj, cls)
            return obj
        return None

    def cache(self) -> None:
        """Update object in class's cache.

        Also calls ._disappear if cache holds older reference to object of same
        ID, but different memory address, to avoid doing anything with
        dangling leftovers.
        """
        if self.id_ is None:
            raise HandledException('Cannot cache object without ID.')
        cache = self.get_cache()
        old_cached = self._get_cached(self.id_)
        if old_cached and id(old_cached) != id(self):
            # pylint: disable=protected-access
            # (cause we remain within the class)
            old_cached._disappear()
        cache[self.id_] = self

    def _uncache(self) -> None:
        """Remove self from cache."""
        if self.id_ is None:
            raise HandledException('Cannot un-cache object without ID.')
        cache = self.get_cache()
        del cache[self.id_]

    # object retrieval and generation

    @classmethod
    def from_table_row(cls: type[BaseModelInstance],
                       # pylint: disable=unused-argument
                       db_conn: DatabaseConnection,
                       row: Row | list[Any]) -> BaseModelInstance:
        """Make from DB row (sans relations), update DB cache with it."""
        obj = cls(*row)
        assert obj.id_ is not None
        for attr_name in cls.to_save_versioned:
            attr = getattr(obj, attr_name)
            table_name = attr.table_name
            for row_ in db_conn.row_where(table_name, 'parent', obj.id_):
                attr.history_from_row(row_)
        obj.cache()
        return obj

    @classmethod
    def by_id(cls, db_conn: DatabaseConnection,
              id_: BaseModelId | None
              ) -> Self:
        """Retrieve by id_, on failure throw NotFoundException.

        First try to get from cls.cache_, only then check DB; if found,
        put into cache.
        """
        obj = None
        if id_ is not None:
            obj = cls._get_cached(id_)
            if not obj:
                for row in db_conn.row_where(cls.table_name, 'id', id_):
                    obj = cls.from_table_row(db_conn, row)
                    break
        if obj:
            return obj
        raise NotFoundException(f'found no object of ID {id_}')

    @classmethod
    def by_id_or_create(cls, db_conn: DatabaseConnection,
                        id_: BaseModelId | None
                        ) -> Self:
        """Wrapper around .by_id, creating (not caching/saving) if not find."""
        if not cls.can_create_by_id:
            raise HandledException('Class cannot .by_id_or_create.')
        try:
            return cls.by_id(db_conn, id_)
        except NotFoundException:
            obj = cls(id_)
            return obj

    @classmethod
    def all(cls: type[BaseModelInstance],
            db_conn: DatabaseConnection) -> list[BaseModelInstance]:
        """Collect all objects of class into list.

        Note that this primarily returns the contents of the cache, and only
        _expands_ that by additional findings in the DB. This assumes the
        cache is always instantly cleaned of any items that would be removed
        from the DB.
        """
        items: dict[BaseModelId, BaseModelInstance] = {}
        for k, v in cls.get_cache().items():
            assert isinstance(v, cls)
            items[k] = v
        already_recorded = items.keys()
        for id_ in db_conn.column_all(cls.table_name, 'id'):
            if id_ not in already_recorded:
                item = cls.by_id(db_conn, id_)
                assert item.id_ is not None
                items[item.id_] = item
        return list(items.values())

    @classmethod
    def by_date_range_with_limits(cls: type[BaseModelInstance],
                                  db_conn: DatabaseConnection,
                                  date_range: tuple[str, str],
                                  date_col: str = 'day'
                                  ) -> tuple[list[BaseModelInstance], str,
                                             str]:
        """Return list of items in database within (open) date_range interval.

        If no range values provided, defaults them to 'yesterday' and
        'tomorrow'. Knows to properly interpret these and 'today' as value.
        """
        start_str = date_range[0] if date_range[0] else 'yesterday'
        end_str = date_range[1] if date_range[1] else 'tomorrow'
        start_date = valid_date(start_str)
        end_date = valid_date(end_str)
        items = []
        sql = f'SELECT id FROM {cls.table_name} '
        sql += f'WHERE {date_col} >= ? AND {date_col} <= ?'
        for row in db_conn.exec(sql, (start_date, end_date)):
            items += [cls.by_id(db_conn, row[0])]
        return items, start_date, end_date

    @classmethod
    def matching(cls: type[BaseModelInstance], db_conn: DatabaseConnection,
                 pattern: str) -> list[BaseModelInstance]:
        """Return all objects whose .to_search match pattern."""
        items = cls.all(db_conn)
        if pattern:
            filtered = []
            for item in items:
                for attr_name in cls.to_search:
                    toks = attr_name.split('.')
                    parent = item
                    for tok in toks:
                        attr = getattr(parent, tok)
                        parent = attr
                    if pattern in attr:
                        filtered += [item]
                        break
            return filtered
        return items

    # database writing

    def save(self, db_conn: DatabaseConnection) -> None:
        """Write self to DB and cache and ensure .id_.

        Write both to DB, and to cache. To DB, write .id_ and attributes
        listed in cls.to_save[_versioned|_relations].

        Ensure self.id_ by setting it to what the DB command returns as the
        last saved row's ID (cursor.lastrowid), EXCEPT if self.id_ already
        exists as a 'str', which implies we do our own ID creation (so far
        only the case with the Day class, where it's to be a date string.
        """
        values = tuple([self.id_] + [getattr(self, key)
                                     for key in self.to_save])
        table_name = self.table_name
        cursor = db_conn.exec_on_vals(f'REPLACE INTO {table_name} VALUES',
                                      values)
        if not isinstance(self.id_, str):
            self.id_ = cursor.lastrowid  # type: ignore[assignment]
        self.cache()
        for attr_name in self.to_save_versioned:
            getattr(self, attr_name).save(db_conn)
        for table, column, attr_name, key_index in self.to_save_relations:
            assert isinstance(self.id_, (int, str))
            db_conn.rewrite_relations(table, column, self.id_,
                                      [[i.id_] for i
                                       in getattr(self, attr_name)], key_index)

    def remove(self, db_conn: DatabaseConnection) -> None:
        """Remove from DB and cache, including dependencies."""
        if self.id_ is None or self._get_cached(self.id_) is None:
            raise HandledException('cannot remove unsaved item')
        for attr_name in self.to_save_versioned:
            getattr(self, attr_name).remove(db_conn)
        for table, column, attr_name, _ in self.to_save_relations:
            db_conn.delete_where(table, column, self.id_)
        self._uncache()
        db_conn.delete_where(self.table_name, 'id', self.id_)
        self._disappear()