core.datasets.lmdb_database

core.datasets.lmdb_database#

Copyright (c) Meta, Inc. and its affiliates.

This source code is modified from the ASE db json backend and is thus licensed under the corresponding LGPL2.1 license

The ASE notice for the LGPL2.1 license is available here: ase/ase/-/blob/master/LICENSE

Attributes#

RESERVED_KEYS

Classes#

LMDBDatabase

Base class for all databases.

Module Contents#

core.datasets.lmdb_database.RESERVED_KEYS = ['nextid', 'metadata', 'deleted_ids']#

class core.datasets.lmdb_database.LMDBDatabase(filename: str | pathlib.Path | None = None, create_indices: bool = True, use_lock_file: bool = False, serial: bool = False, readonly: bool = False, *args, **kwargs)#

Bases: ase.db.core.Database

Base class for all databases.

readonly#

ids = []#

deleted_ids = []#

__enter__() → typing_extensions.Self#

__exit__(exc_type, exc_value, tb) → None#

close() → None#

_write(atoms: ase.Atoms | ase.db.row.AtomsRow, key_value_pairs: dict, data: dict | None, idx: int | None = None) → None#

_update(idx: int, key_value_pairs: dict | None = None, data: dict | None = None)#

_write_deleted_ids()#

delete(ids: list[int]) → None#: Delete rows.

_get_row(idx: int, include_data: bool = True)#

_get_row_by_index(index: int, include_data: bool = True)#: Auxiliary function to get the ith entry, rather than a specific id

_select(keys, cmps: list[tuple[str, str, str]], explain: bool = False, verbosity: int = 0, limit: int | None = None, offset: int = 0, sort: str | None = None, include_data: bool = True, columns: str = 'all')#

property metadata#: Load the metadata from the DB if present

property _nextid#: Get the id of the next row to be written

count(selection=None, **kwargs) → int#

Count rows.

See the select() method for the selection syntax. Use db.count() or len(db) to count all rows.

_load_ids() → None#

Load ids from the DB

Since ASE db ids are mostly 1-N integers, but can be missing entries if ids have been deleted. To save space and operating under the assumption that there will probably not be many deletions in most OCP datasets, we just store the deleted ids.