plomlompom.com Git - plomtask/blob - plomtask/db.py

   1 """Database management."""
   2 from __future__ import annotations
   3 from os import listdir
   4 from os.path import isfile
   5 from difflib import Differ
   6 from sqlite3 import connect as sql_connect, Cursor, Row
   7 from typing import Any, Self, TypeVar, Generic
   8 from plomtask.exceptions import HandledException, NotFoundException
   9 from plomtask.dating import valid_date
  10
  11 EXPECTED_DB_VERSION = 5
  12 MIGRATIONS_DIR = 'migrations'
  13 FILENAME_DB_SCHEMA = f'init_{EXPECTED_DB_VERSION}.sql'
  14 PATH_DB_SCHEMA = f'{MIGRATIONS_DIR}/{FILENAME_DB_SCHEMA}'
  15
  16
  17 class UnmigratedDbException(HandledException):
  18     """To identify case of unmigrated DB file."""
  19
  20
  21 class DatabaseFile:
  22     """Represents the sqlite3 database's file."""
  23     # pylint: disable=too-few-public-methods
  24
  25     def __init__(self, path: str) -> None:
  26         self.path = path
  27         self._check()
  28
  29     @classmethod
  30     def create_at(cls, path: str) -> DatabaseFile:
  31         """Make new DB file at path."""
  32         with sql_connect(path) as conn:
  33             with open(PATH_DB_SCHEMA, 'r', encoding='utf-8') as f:
  34                 conn.executescript(f.read())
  35             conn.execute(f'PRAGMA user_version = {EXPECTED_DB_VERSION}')
  36         return cls(path)
  37
  38     @classmethod
  39     def migrate(cls, path: str) -> DatabaseFile:
  40         """Apply migrations from_version to EXPECTED_DB_VERSION."""
  41         migrations = cls._available_migrations()
  42         from_version = cls._get_version_of_db(path)
  43         migrations_todo = migrations[from_version+1:]
  44         for j, filename in enumerate(migrations_todo):
  45             with sql_connect(path) as conn:
  46                 with open(f'{MIGRATIONS_DIR}/{filename}', 'r',
  47                           encoding='utf-8') as f:
  48                     conn.executescript(f.read())
  49             user_version = from_version + j + 1
  50             with sql_connect(path) as conn:
  51                 conn.execute(f'PRAGMA user_version = {user_version}')
  52         return cls(path)
  53
  54     def _check(self) -> None:
  55         """Check file exists, and is of proper DB version and schema."""
  56         if not isfile(self.path):
  57             raise NotFoundException
  58         if self._user_version != EXPECTED_DB_VERSION:
  59             raise UnmigratedDbException()
  60         self._validate_schema()
  61
  62     @staticmethod
  63     def _available_migrations() -> list[str]:
  64         """Validate migrations directory and return sorted entries."""
  65         msg_too_big = 'Migration directory points beyond expected DB version.'
  66         msg_bad_entry = 'Migration directory contains unexpected entry: '
  67         msg_missing = 'Migration directory misses migration of number: '
  68         migrations = {}
  69         for entry in listdir(MIGRATIONS_DIR):
  70             if entry == FILENAME_DB_SCHEMA:
  71                 continue
  72             toks = entry.split('_', 1)
  73             if len(toks) < 2:
  74                 raise HandledException(msg_bad_entry + entry)
  75             try:
  76                 i = int(toks[0])
  77             except ValueError as e:
  78                 raise HandledException(msg_bad_entry + entry) from e
  79             if i > EXPECTED_DB_VERSION:
  80                 raise HandledException(msg_too_big)
  81             migrations[i] = toks[1]
  82         migrations_list = []
  83         for i in range(EXPECTED_DB_VERSION + 1):
  84             if i not in migrations:
  85                 raise HandledException(msg_missing + str(i))
  86             migrations_list += [f'{i}_{migrations[i]}']
  87         return migrations_list
  88
  89     @staticmethod
  90     def _get_version_of_db(path: str) -> int:
  91         """Get DB user_version, fail if outside expected range."""
  92         sql_for_db_version = 'PRAGMA user_version'
  93         with sql_connect(path) as conn:
  94             db_version = list(conn.execute(sql_for_db_version))[0][0]
  95         if db_version > EXPECTED_DB_VERSION:
  96             msg = f'Wrong DB version, expected '\
  97                     f'{EXPECTED_DB_VERSION}, got unknown {db_version}.'
  98             raise HandledException(msg)
  99         assert isinstance(db_version, int)
 100         return db_version
 101
 102     @property
 103     def _user_version(self) -> int:
 104         """Get DB user_version."""
 105         return self._get_version_of_db(self.path)
 106
 107     def _validate_schema(self) -> None:
 108         """Compare found schema with what's stored at PATH_DB_SCHEMA."""
 109
 110         def reformat_rows(rows: list[str]) -> list[str]:
 111             new_rows = []
 112             for row in rows:
 113                 new_row = []
 114                 for subrow in row.split('\n'):
 115                     subrow = subrow.rstrip()
 116                     in_parentheses = 0
 117                     split_at = []
 118                     for i, c in enumerate(subrow):
 119                         if '(' == c:
 120                             in_parentheses += 1
 121                         elif ')' == c:
 122                             in_parentheses -= 1
 123                         elif ',' == c and 0 == in_parentheses:
 124                             split_at += [i + 1]
 125                     prev_split = 0
 126                     for i in split_at:
 127                         segment = subrow[prev_split:i].strip()
 128                         if len(segment) > 0:
 129                             new_row += [f'    {segment}']
 130                         prev_split = i
 131                     segment = subrow[prev_split:].strip()
 132                     if len(segment) > 0:
 133                         new_row += [f'    {segment}']
 134                 new_row[0] = new_row[0].lstrip()
 135                 new_row[-1] = new_row[-1].lstrip()
 136                 if new_row[-1] != ')' and new_row[-3][-1] != ',':
 137                     new_row[-3] = new_row[-3] + ','
 138                     new_row[-2:] = ['    ' + new_row[-1][:-1]] + [')']
 139                 new_rows += ['\n'.join(new_row)]
 140             return new_rows
 141
 142         sql_for_schema = 'SELECT sql FROM sqlite_master ORDER BY sql'
 143         msg_err = 'Database has wrong tables schema. Diff:\n'
 144         with sql_connect(self.path) as conn:
 145             schema_rows = [r[0] for r in conn.execute(sql_for_schema) if r[0]]
 146         schema_rows = reformat_rows(schema_rows)
 147         retrieved_schema = ';\n'.join(schema_rows) + ';'
 148         with open(PATH_DB_SCHEMA, 'r', encoding='utf-8') as f:
 149             stored_schema = f.read().rstrip()
 150         if stored_schema != retrieved_schema:
 151             diff_msg = Differ().compare(retrieved_schema.splitlines(),
 152                                         stored_schema.splitlines())
 153             raise HandledException(msg_err + '\n'.join(diff_msg))
 154
 155
 156 class DatabaseConnection:
 157     """A single connection to the database."""
 158
 159     def __init__(self, db_file: DatabaseFile) -> None:
 160         self.conn = sql_connect(db_file.path)
 161
 162     def commit(self) -> None:
 163         """Commit SQL transaction."""
 164         self.conn.commit()
 165
 166     def exec(self, code: str, inputs: tuple[Any, ...] = tuple()) -> Cursor:
 167         """Add commands to SQL transaction."""
 168         return self.conn.execute(code, inputs)
 169
 170     def exec_on_vals(self, code: str, inputs: tuple[Any, ...]) -> Cursor:
 171         """Wrapper around .exec appending adequate " (?, …)" to code."""
 172         q_marks_from_values = '(' + ','.join(['?'] * len(inputs)) + ')'
 173         return self.exec(f'{code} {q_marks_from_values}', inputs)
 174
 175     def close(self) -> None:
 176         """Close DB connection."""
 177         self.conn.close()
 178
 179     def rewrite_relations(self, table_name: str, key: str, target: int | str,
 180                           rows: list[list[Any]], key_index: int = 0) -> None:
 181         # pylint: disable=too-many-arguments
 182         """Rewrite relations in table_name to target, with rows values.
 183
 184         Note that single rows are expected without the column and value
 185         identified by key and target, which are inserted inside the function
 186         at key_index.
 187         """
 188         self.delete_where(table_name, key, target)
 189         for row in rows:
 190             values = tuple(row[:key_index] + [target] + row[key_index:])
 191             self.exec_on_vals(f'INSERT INTO {table_name} VALUES', values)
 192
 193     def row_where(self, table_name: str, key: str,
 194                   target: int | str) -> list[Row]:
 195         """Return list of Rows at table where key == target."""
 196         return list(self.exec(f'SELECT * FROM {table_name} WHERE {key} = ?',
 197                               (target,)))
 198
 199     # def column_where_pattern(self,
 200     #                          table_name: str,
 201     #                          column: str,
 202     #                          pattern: str,
 203     #                          keys: list[str]) -> list[Any]:
 204     #     """Return column of rows where one of keys matches pattern."""
 205     #     targets = tuple([f'%{pattern}%'] * len(keys))
 206     #     haystack = ' OR '.join([f'{k} LIKE ?' for k in keys])
 207     #     sql = f'SELECT {column} FROM {table_name} WHERE {haystack}'
 208     #     return [row[0] for row in self.exec(sql, targets)]
 209
 210     def column_where(self, table_name: str, column: str, key: str,
 211                      target: int | str) -> list[Any]:
 212         """Return column of table where key == target."""
 213         return [row[0] for row in
 214                 self.exec(f'SELECT {column} FROM {table_name} '
 215                           f'WHERE {key} = ?', (target,))]
 216
 217     def column_all(self, table_name: str, column: str) -> list[Any]:
 218         """Return complete column of table."""
 219         return [row[0] for row in
 220                 self.exec(f'SELECT {column} FROM {table_name}')]
 221
 222     def delete_where(self, table_name: str, key: str,
 223                      target: int | str) -> None:
 224         """Delete from table where key == target."""
 225         self.exec(f'DELETE FROM {table_name} WHERE {key} = ?', (target,))
 226
 227
 228 BaseModelId = TypeVar('BaseModelId', int, str)
 229 BaseModelInstance = TypeVar('BaseModelInstance', bound='BaseModel[Any]')
 230
 231
 232 class BaseModel(Generic[BaseModelId]):
 233     """Template for most of the models we use/derive from the DB."""
 234     table_name = ''
 235     to_save: list[str] = []
 236     to_save_versioned: list[str] = []
 237     to_save_relations: list[tuple[str, str, str, int]] = []
 238     id_: None | BaseModelId
 239     cache_: dict[BaseModelId, Self]
 240     to_search: list[str] = []
 241     can_create_by_id = False
 242     _exists = True
 243
 244     def __init__(self, id_: BaseModelId | None) -> None:
 245         if isinstance(id_, int) and id_ < 1:
 246             msg = f'illegal {self.__class__.__name__} ID, must be >=1: {id_}'
 247             raise HandledException(msg)
 248         if isinstance(id_, str) and "" == id_:
 249             msg = f'illegal {self.__class__.__name__} ID, must be non-empty'
 250             raise HandledException(msg)
 251         self.id_ = id_
 252
 253     def __hash__(self) -> int:
 254         hashable = [self.id_] + [getattr(self, name) for name in self.to_save]
 255         for definition in self.to_save_relations:
 256             attr = getattr(self, definition[2])
 257             hashable += [tuple(rel.id_ for rel in attr)]
 258         for name in self.to_save_versioned:
 259             hashable += [hash(getattr(self, name))]
 260         return hash(tuple(hashable))
 261
 262     def __eq__(self, other: object) -> bool:
 263         if not isinstance(other, self.__class__):
 264             return False
 265         return hash(self) == hash(other)
 266
 267     def __lt__(self, other: Any) -> bool:
 268         if not isinstance(other, self.__class__):
 269             msg = 'cannot compare to object of different class'
 270             raise HandledException(msg)
 271         assert isinstance(self.id_, int)
 272         assert isinstance(other.id_, int)
 273         return self.id_ < other.id_
 274
 275     @property
 276     def as_dict(self) -> dict[str, object]:
 277         """Return self as (json.dumps-coompatible) dict."""
 278         d: dict[str, object] = {'id': self.id_}
 279         if len(self.to_save_versioned) > 0:
 280             d['_versioned'] = {}
 281         for k in self.to_save:
 282             attr = getattr(self, k)
 283             if hasattr(attr, 'as_dict'):
 284                 d[k] = attr.as_dict
 285             d[k] = attr
 286         for k in self.to_save_versioned:
 287             attr = getattr(self, k)
 288             assert isinstance(d['_versioned'], dict)
 289             d['_versioned'][k] = attr.history
 290         for r in self.to_save_relations:
 291             attr_name = r[2]
 292             d[attr_name] = [x.as_dict for x in getattr(self, attr_name)]
 293         return d
 294
 295     # cache management
 296     # (we primarily use the cache to ensure we work on the same object in
 297     # memory no matter where and how we retrieve it, e.g. we don't want
 298     # .by_id() calls to create a new object each time, but rather a pointer
 299     # to the one already instantiated)
 300
 301     def __getattribute__(self, name: str) -> Any:
 302         """Ensure fail if ._disappear() was called, except to check ._exists"""
 303         if name != '_exists' and not super().__getattribute__('_exists'):
 304             raise HandledException('Object does not exist.')
 305         return super().__getattribute__(name)
 306
 307     def _disappear(self) -> None:
 308         """Invalidate object, make future use raise exceptions."""
 309         assert self.id_ is not None
 310         if self._get_cached(self.id_):
 311             self._uncache()
 312         to_kill = list(self.__dict__.keys())
 313         for attr in to_kill:
 314             delattr(self, attr)
 315         self._exists = False
 316
 317     @classmethod
 318     def empty_cache(cls) -> None:
 319         """Empty class's cache, and disappear all former inhabitants."""
 320         # pylint: disable=protected-access
 321         # (cause we remain within the class)
 322         if hasattr(cls, 'cache_'):
 323             to_disappear = list(cls.cache_.values())
 324             for item in to_disappear:
 325                 item._disappear()
 326         cls.cache_ = {}
 327
 328     @classmethod
 329     def get_cache(cls: type[BaseModelInstance]) -> dict[Any, BaseModel[Any]]:
 330         """Get cache dictionary, create it if not yet existing."""
 331         if not hasattr(cls, 'cache_'):
 332             d: dict[Any, BaseModel[Any]] = {}
 333             cls.cache_ = d
 334         return cls.cache_
 335
 336     @classmethod
 337     def _get_cached(cls: type[BaseModelInstance],
 338                     id_: BaseModelId) -> BaseModelInstance | None:
 339         """Get object of id_ from class's cache, or None if not found."""
 340         cache = cls.get_cache()
 341         if id_ in cache:
 342             obj = cache[id_]
 343             assert isinstance(obj, cls)
 344             return obj
 345         return None
 346
 347     def cache(self) -> None:
 348         """Update object in class's cache.
 349
 350         Also calls ._disappear if cache holds older reference to object of same
 351         ID, but different memory address, to avoid doing anything with
 352         dangling leftovers.
 353         """
 354         if self.id_ is None:
 355             raise HandledException('Cannot cache object without ID.')
 356         cache = self.get_cache()
 357         old_cached = self._get_cached(self.id_)
 358         if old_cached and id(old_cached) != id(self):
 359             # pylint: disable=protected-access
 360             # (cause we remain within the class)
 361             old_cached._disappear()
 362         cache[self.id_] = self
 363
 364     def _uncache(self) -> None:
 365         """Remove self from cache."""
 366         if self.id_ is None:
 367             raise HandledException('Cannot un-cache object without ID.')
 368         cache = self.get_cache()
 369         del cache[self.id_]
 370
 371     # object retrieval and generation
 372
 373     @classmethod
 374     def from_table_row(cls: type[BaseModelInstance],
 375                        # pylint: disable=unused-argument
 376                        db_conn: DatabaseConnection,
 377                        row: Row | list[Any]) -> BaseModelInstance:
 378         """Make from DB row (sans relations), update DB cache with it."""
 379         obj = cls(*row)
 380         assert obj.id_ is not None
 381         for attr_name in cls.to_save_versioned:
 382             attr = getattr(obj, attr_name)
 383             table_name = attr.table_name
 384             for row_ in db_conn.row_where(table_name, 'parent', obj.id_):
 385                 attr.history_from_row(row_)
 386         obj.cache()
 387         return obj
 388
 389     @classmethod
 390     def by_id(cls, db_conn: DatabaseConnection, id_: BaseModelId) -> Self:
 391         """Retrieve by id_, on failure throw NotFoundException.
 392
 393         First try to get from cls.cache_, only then check DB; if found,
 394         put into cache.
 395         """
 396         obj = None
 397         if id_ is not None:
 398             obj = cls._get_cached(id_)
 399             if not obj:
 400                 for row in db_conn.row_where(cls.table_name, 'id', id_):
 401                     obj = cls.from_table_row(db_conn, row)
 402                     break
 403         if obj:
 404             return obj
 405         raise NotFoundException(f'found no object of ID {id_}')
 406
 407     @classmethod
 408     def by_id_or_create(cls, db_conn: DatabaseConnection,
 409                         id_: BaseModelId | None
 410                         ) -> Self:
 411         """Wrapper around .by_id, creating (not caching/saving) if not find."""
 412         if not cls.can_create_by_id:
 413             raise HandledException('Class cannot .by_id_or_create.')
 414         if id_ is None:
 415             return cls(None)
 416         try:
 417             return cls.by_id(db_conn, id_)
 418         except NotFoundException:
 419             return cls(id_)
 420
 421     @classmethod
 422     def all(cls: type[BaseModelInstance],
 423             db_conn: DatabaseConnection) -> list[BaseModelInstance]:
 424         """Collect all objects of class into list.
 425
 426         Note that this primarily returns the contents of the cache, and only
 427         _expands_ that by additional findings in the DB. This assumes the
 428         cache is always instantly cleaned of any items that would be removed
 429         from the DB.
 430         """
 431         items: dict[BaseModelId, BaseModelInstance] = {}
 432         for k, v in cls.get_cache().items():
 433             assert isinstance(v, cls)
 434             items[k] = v
 435         already_recorded = items.keys()
 436         for id_ in db_conn.column_all(cls.table_name, 'id'):
 437             if id_ not in already_recorded:
 438                 item = cls.by_id(db_conn, id_)
 439                 assert item.id_ is not None
 440                 items[item.id_] = item
 441         return list(items.values())
 442
 443     @classmethod
 444     def by_date_range_with_limits(cls: type[BaseModelInstance],
 445                                   db_conn: DatabaseConnection,
 446                                   date_range: tuple[str, str],
 447                                   date_col: str = 'day'
 448                                   ) -> tuple[list[BaseModelInstance], str,
 449                                              str]:
 450         """Return list of items in database within (open) date_range interval.
 451
 452         If no range values provided, defaults them to 'yesterday' and
 453         'tomorrow'. Knows to properly interpret these and 'today' as value.
 454         """
 455         start_str = date_range[0] if date_range[0] else 'yesterday'
 456         end_str = date_range[1] if date_range[1] else 'tomorrow'
 457         start_date = valid_date(start_str)
 458         end_date = valid_date(end_str)
 459         items = []
 460         sql = f'SELECT id FROM {cls.table_name} '
 461         sql += f'WHERE {date_col} >= ? AND {date_col} <= ?'
 462         for row in db_conn.exec(sql, (start_date, end_date)):
 463             items += [cls.by_id(db_conn, row[0])]
 464         return items, start_date, end_date
 465
 466     @classmethod
 467     def matching(cls: type[BaseModelInstance], db_conn: DatabaseConnection,
 468                  pattern: str) -> list[BaseModelInstance]:
 469         """Return all objects whose .to_search match pattern."""
 470         items = cls.all(db_conn)
 471         if pattern:
 472             filtered = []
 473             for item in items:
 474                 for attr_name in cls.to_search:
 475                     toks = attr_name.split('.')
 476                     parent = item
 477                     for tok in toks:
 478                         attr = getattr(parent, tok)
 479                         parent = attr
 480                     if pattern in attr:
 481                         filtered += [item]
 482                         break
 483             return filtered
 484         return items
 485
 486     # database writing
 487
 488     def save(self, db_conn: DatabaseConnection) -> None:
 489         """Write self to DB and cache and ensure .id_.
 490
 491         Write both to DB, and to cache. To DB, write .id_ and attributes
 492         listed in cls.to_save[_versioned|_relations].
 493
 494         Ensure self.id_ by setting it to what the DB command returns as the
 495         last saved row's ID (cursor.lastrowid), EXCEPT if self.id_ already
 496         exists as a 'str', which implies we do our own ID creation (so far
 497         only the case with the Day class, where it's to be a date string.
 498         """
 499         values = tuple([self.id_] + [getattr(self, key)
 500                                      for key in self.to_save])
 501         table_name = self.table_name
 502         cursor = db_conn.exec_on_vals(f'REPLACE INTO {table_name} VALUES',
 503                                       values)
 504         if not isinstance(self.id_, str):
 505             self.id_ = cursor.lastrowid  # type: ignore[assignment]
 506         self.cache()
 507         for attr_name in self.to_save_versioned:
 508             getattr(self, attr_name).save(db_conn)
 509         for table, column, attr_name, key_index in self.to_save_relations:
 510             assert isinstance(self.id_, (int, str))
 511             db_conn.rewrite_relations(table, column, self.id_,
 512                                       [[i.id_] for i
 513                                        in getattr(self, attr_name)], key_index)
 514
 515     def remove(self, db_conn: DatabaseConnection) -> None:
 516         """Remove from DB and cache, including dependencies."""
 517         if self.id_ is None or self._get_cached(self.id_) is None:
 518             raise HandledException('cannot remove unsaved item')
 519         for attr_name in self.to_save_versioned:
 520             getattr(self, attr_name).remove(db_conn)
 521         for table, column, attr_name, _ in self.to_save_relations:
 522             db_conn.delete_where(table, column, self.id_)
 523         self._uncache()
 524         db_conn.delete_where(self.table_name, 'id', self.id_)
 525         self._disappear()