bazarr/libs/peewee.py

7815 lines
261 KiB
Python

from bisect import bisect_left
from bisect import bisect_right
from contextlib import contextmanager
from copy import deepcopy
from functools import wraps
from inspect import isclass
import calendar
import collections
import datetime
import decimal
import hashlib
import itertools
import logging
import operator
import re
import socket
import struct
import sys
import threading
import time
import uuid
import warnings
try:
from collections.abc import Mapping
except ImportError:
from collections import Mapping
try:
from pysqlite3 import dbapi2 as pysq3
except ImportError:
try:
from pysqlite2 import dbapi2 as pysq3
except ImportError:
pysq3 = None
try:
import sqlite3
except ImportError:
sqlite3 = pysq3
else:
if pysq3 and pysq3.sqlite_version_info >= sqlite3.sqlite_version_info:
sqlite3 = pysq3
try:
from psycopg2cffi import compat
compat.register()
except ImportError:
pass
try:
import psycopg2
from psycopg2 import extensions as pg_extensions
try:
from psycopg2 import errors as pg_errors
except ImportError:
pg_errors = None
except ImportError:
psycopg2 = pg_errors = None
try:
from psycopg2.extras import register_uuid as pg_register_uuid
pg_register_uuid()
except Exception:
pass
mysql_passwd = False
try:
import pymysql as mysql
except ImportError:
try:
import MySQLdb as mysql
mysql_passwd = True
except ImportError:
mysql = None
__version__ = '3.14.8'
__all__ = [
'AnyField',
'AsIs',
'AutoField',
'BareField',
'BigAutoField',
'BigBitField',
'BigIntegerField',
'BinaryUUIDField',
'BitField',
'BlobField',
'BooleanField',
'Case',
'Cast',
'CharField',
'Check',
'chunked',
'Column',
'CompositeKey',
'Context',
'Database',
'DatabaseError',
'DatabaseProxy',
'DataError',
'DateField',
'DateTimeField',
'DecimalField',
'DeferredForeignKey',
'DeferredThroughModel',
'DJANGO_MAP',
'DoesNotExist',
'DoubleField',
'DQ',
'EXCLUDED',
'Field',
'FixedCharField',
'FloatField',
'fn',
'ForeignKeyField',
'IdentityField',
'ImproperlyConfigured',
'Index',
'IntegerField',
'IntegrityError',
'InterfaceError',
'InternalError',
'IPField',
'JOIN',
'ManyToManyField',
'Model',
'ModelIndex',
'MySQLDatabase',
'NotSupportedError',
'OP',
'OperationalError',
'PostgresqlDatabase',
'PrimaryKeyField', # XXX: Deprecated, change to AutoField.
'prefetch',
'ProgrammingError',
'Proxy',
'QualifiedNames',
'SchemaManager',
'SmallIntegerField',
'Select',
'SQL',
'SqliteDatabase',
'Table',
'TextField',
'TimeField',
'TimestampField',
'Tuple',
'UUIDField',
'Value',
'ValuesList',
'Window',
]
try: # Python 2.7+
from logging import NullHandler
except ImportError:
class NullHandler(logging.Handler):
def emit(self, record):
pass
logger = logging.getLogger('peewee')
logger.addHandler(NullHandler())
if sys.version_info[0] == 2:
text_type = unicode
bytes_type = str
buffer_type = buffer
izip_longest = itertools.izip_longest
callable_ = callable
multi_types = (list, tuple, frozenset, set)
exec('def reraise(tp, value, tb=None): raise tp, value, tb')
def print_(s):
sys.stdout.write(s)
sys.stdout.write('\n')
else:
import builtins
try:
from collections.abc import Callable
except ImportError:
from collections import Callable
from functools import reduce
callable_ = lambda c: isinstance(c, Callable)
text_type = str
bytes_type = bytes
buffer_type = memoryview
basestring = str
long = int
multi_types = (list, tuple, frozenset, set, range)
print_ = getattr(builtins, 'print')
izip_longest = itertools.zip_longest
def reraise(tp, value, tb=None):
if value.__traceback__ is not tb:
raise value.with_traceback(tb)
raise value
if sqlite3:
sqlite3.register_adapter(decimal.Decimal, str)
sqlite3.register_adapter(datetime.date, str)
sqlite3.register_adapter(datetime.time, str)
__sqlite_version__ = sqlite3.sqlite_version_info
else:
__sqlite_version__ = (0, 0, 0)
__date_parts__ = set(('year', 'month', 'day', 'hour', 'minute', 'second'))
# Sqlite does not support the `date_part` SQL function, so we will define an
# implementation in python.
__sqlite_datetime_formats__ = (
'%Y-%m-%d %H:%M:%S',
'%Y-%m-%d %H:%M:%S.%f',
'%Y-%m-%d',
'%H:%M:%S',
'%H:%M:%S.%f',
'%H:%M')
__sqlite_date_trunc__ = {
'year': '%Y-01-01 00:00:00',
'month': '%Y-%m-01 00:00:00',
'day': '%Y-%m-%d 00:00:00',
'hour': '%Y-%m-%d %H:00:00',
'minute': '%Y-%m-%d %H:%M:00',
'second': '%Y-%m-%d %H:%M:%S'}
__mysql_date_trunc__ = __sqlite_date_trunc__.copy()
__mysql_date_trunc__['minute'] = '%Y-%m-%d %H:%i:00'
__mysql_date_trunc__['second'] = '%Y-%m-%d %H:%i:%S'
def _sqlite_date_part(lookup_type, datetime_string):
assert lookup_type in __date_parts__
if not datetime_string:
return
dt = format_date_time(datetime_string, __sqlite_datetime_formats__)
return getattr(dt, lookup_type)
def _sqlite_date_trunc(lookup_type, datetime_string):
assert lookup_type in __sqlite_date_trunc__
if not datetime_string:
return
dt = format_date_time(datetime_string, __sqlite_datetime_formats__)
return dt.strftime(__sqlite_date_trunc__[lookup_type])
def __deprecated__(s):
warnings.warn(s, DeprecationWarning)
class attrdict(dict):
def __getattr__(self, attr):
try:
return self[attr]
except KeyError:
raise AttributeError(attr)
def __setattr__(self, attr, value): self[attr] = value
def __iadd__(self, rhs): self.update(rhs); return self
def __add__(self, rhs): d = attrdict(self); d.update(rhs); return d
SENTINEL = object()
#: Operations for use in SQL expressions.
OP = attrdict(
AND='AND',
OR='OR',
ADD='+',
SUB='-',
MUL='*',
DIV='/',
BIN_AND='&',
BIN_OR='|',
XOR='#',
MOD='%',
EQ='=',
LT='<',
LTE='<=',
GT='>',
GTE='>=',
NE='!=',
IN='IN',
NOT_IN='NOT IN',
IS='IS',
IS_NOT='IS NOT',
LIKE='LIKE',
ILIKE='ILIKE',
BETWEEN='BETWEEN',
REGEXP='REGEXP',
IREGEXP='IREGEXP',
CONCAT='||',
BITWISE_NEGATION='~')
# To support "django-style" double-underscore filters, create a mapping between
# operation name and operation code, e.g. "__eq" == OP.EQ.
DJANGO_MAP = attrdict({
'eq': operator.eq,
'lt': operator.lt,
'lte': operator.le,
'gt': operator.gt,
'gte': operator.ge,
'ne': operator.ne,
'in': operator.lshift,
'is': lambda l, r: Expression(l, OP.IS, r),
'like': lambda l, r: Expression(l, OP.LIKE, r),
'ilike': lambda l, r: Expression(l, OP.ILIKE, r),
'regexp': lambda l, r: Expression(l, OP.REGEXP, r),
})
#: Mapping of field type to the data-type supported by the database. Databases
#: may override or add to this list.
FIELD = attrdict(
AUTO='INTEGER',
BIGAUTO='BIGINT',
BIGINT='BIGINT',
BLOB='BLOB',
BOOL='SMALLINT',
CHAR='CHAR',
DATE='DATE',
DATETIME='DATETIME',
DECIMAL='DECIMAL',
DEFAULT='',
DOUBLE='REAL',
FLOAT='REAL',
INT='INTEGER',
SMALLINT='SMALLINT',
TEXT='TEXT',
TIME='TIME',
UUID='TEXT',
UUIDB='BLOB',
VARCHAR='VARCHAR')
#: Join helpers (for convenience) -- all join types are supported, this object
#: is just to help avoid introducing errors by using strings everywhere.
JOIN = attrdict(
INNER='INNER JOIN',
LEFT_OUTER='LEFT OUTER JOIN',
RIGHT_OUTER='RIGHT OUTER JOIN',
FULL='FULL JOIN',
FULL_OUTER='FULL OUTER JOIN',
CROSS='CROSS JOIN',
NATURAL='NATURAL JOIN',
LATERAL='LATERAL',
LEFT_LATERAL='LEFT JOIN LATERAL')
# Row representations.
ROW = attrdict(
TUPLE=1,
DICT=2,
NAMED_TUPLE=3,
CONSTRUCTOR=4,
MODEL=5)
SCOPE_NORMAL = 1
SCOPE_SOURCE = 2
SCOPE_VALUES = 4
SCOPE_CTE = 8
SCOPE_COLUMN = 16
# Rules for parentheses around subqueries in compound select.
CSQ_PARENTHESES_NEVER = 0
CSQ_PARENTHESES_ALWAYS = 1
CSQ_PARENTHESES_UNNESTED = 2
# Regular expressions used to convert class names to snake-case table names.
# First regex handles acronym followed by word or initial lower-word followed
# by a capitalized word. e.g. APIResponse -> API_Response / fooBar -> foo_Bar.
# Second regex handles the normal case of two title-cased words.
SNAKE_CASE_STEP1 = re.compile('(.)_*([A-Z][a-z]+)')
SNAKE_CASE_STEP2 = re.compile('([a-z0-9])_*([A-Z])')
# Helper functions that are used in various parts of the codebase.
MODEL_BASE = '_metaclass_helper_'
def with_metaclass(meta, base=object):
return meta(MODEL_BASE, (base,), {})
def merge_dict(source, overrides):
merged = source.copy()
if overrides:
merged.update(overrides)
return merged
def quote(path, quote_chars):
if len(path) == 1:
return path[0].join(quote_chars)
return '.'.join([part.join(quote_chars) for part in path])
is_model = lambda o: isclass(o) and issubclass(o, Model)
def ensure_tuple(value):
if value is not None:
return value if isinstance(value, (list, tuple)) else (value,)
def ensure_entity(value):
if value is not None:
return value if isinstance(value, Node) else Entity(value)
def make_snake_case(s):
first = SNAKE_CASE_STEP1.sub(r'\1_\2', s)
return SNAKE_CASE_STEP2.sub(r'\1_\2', first).lower()
def chunked(it, n):
marker = object()
for group in (list(g) for g in izip_longest(*[iter(it)] * n,
fillvalue=marker)):
if group[-1] is marker:
del group[group.index(marker):]
yield group
class _callable_context_manager(object):
def __call__(self, fn):
@wraps(fn)
def inner(*args, **kwargs):
with self:
return fn(*args, **kwargs)
return inner
class Proxy(object):
"""
Create a proxy or placeholder for another object.
"""
__slots__ = ('obj', '_callbacks')
def __init__(self):
self._callbacks = []
self.initialize(None)
def initialize(self, obj):
self.obj = obj
for callback in self._callbacks:
callback(obj)
def attach_callback(self, callback):
self._callbacks.append(callback)
return callback
def passthrough(method):
def inner(self, *args, **kwargs):
if self.obj is None:
raise AttributeError('Cannot use uninitialized Proxy.')
return getattr(self.obj, method)(*args, **kwargs)
return inner
# Allow proxy to be used as a context-manager.
__enter__ = passthrough('__enter__')
__exit__ = passthrough('__exit__')
def __getattr__(self, attr):
if self.obj is None:
raise AttributeError('Cannot use uninitialized Proxy.')
return getattr(self.obj, attr)
def __setattr__(self, attr, value):
if attr not in self.__slots__:
raise AttributeError('Cannot set attribute on proxy.')
return super(Proxy, self).__setattr__(attr, value)
class DatabaseProxy(Proxy):
"""
Proxy implementation specifically for proxying `Database` objects.
"""
def connection_context(self):
return ConnectionContext(self)
def atomic(self, *args, **kwargs):
return _atomic(self, *args, **kwargs)
def manual_commit(self):
return _manual(self)
def transaction(self, *args, **kwargs):
return _transaction(self, *args, **kwargs)
def savepoint(self):
return _savepoint(self)
class ModelDescriptor(object): pass
# SQL Generation.
class AliasManager(object):
__slots__ = ('_counter', '_current_index', '_mapping')
def __init__(self):
# A list of dictionaries containing mappings at various depths.
self._counter = 0
self._current_index = 0
self._mapping = []
self.push()
@property
def mapping(self):
return self._mapping[self._current_index - 1]
def add(self, source):
if source not in self.mapping:
self._counter += 1
self[source] = 't%d' % self._counter
return self.mapping[source]
def get(self, source, any_depth=False):
if any_depth:
for idx in reversed(range(self._current_index)):
if source in self._mapping[idx]:
return self._mapping[idx][source]
return self.add(source)
def __getitem__(self, source):
return self.get(source)
def __setitem__(self, source, alias):
self.mapping[source] = alias
def push(self):
self._current_index += 1
if self._current_index > len(self._mapping):
self._mapping.append({})
def pop(self):
if self._current_index == 1:
raise ValueError('Cannot pop() from empty alias manager.')
self._current_index -= 1
class State(collections.namedtuple('_State', ('scope', 'parentheses',
'settings'))):
def __new__(cls, scope=SCOPE_NORMAL, parentheses=False, **kwargs):
return super(State, cls).__new__(cls, scope, parentheses, kwargs)
def __call__(self, scope=None, parentheses=None, **kwargs):
# Scope and settings are "inherited" (parentheses is not, however).
scope = self.scope if scope is None else scope
# Try to avoid unnecessary dict copying.
if kwargs and self.settings:
settings = self.settings.copy() # Copy original settings dict.
settings.update(kwargs) # Update copy with overrides.
elif kwargs:
settings = kwargs
else:
settings = self.settings
return State(scope, parentheses, **settings)
def __getattr__(self, attr_name):
return self.settings.get(attr_name)
def __scope_context__(scope):
@contextmanager
def inner(self, **kwargs):
with self(scope=scope, **kwargs):
yield self
return inner
class Context(object):
__slots__ = ('stack', '_sql', '_values', 'alias_manager', 'state')
def __init__(self, **settings):
self.stack = []
self._sql = []
self._values = []
self.alias_manager = AliasManager()
self.state = State(**settings)
def as_new(self):
return Context(**self.state.settings)
def column_sort_key(self, item):
return item[0].get_sort_key(self)
@property
def scope(self):
return self.state.scope
@property
def parentheses(self):
return self.state.parentheses
@property
def subquery(self):
return self.state.subquery
def __call__(self, **overrides):
if overrides and overrides.get('scope') == self.scope:
del overrides['scope']
self.stack.append(self.state)
self.state = self.state(**overrides)
return self
scope_normal = __scope_context__(SCOPE_NORMAL)
scope_source = __scope_context__(SCOPE_SOURCE)
scope_values = __scope_context__(SCOPE_VALUES)
scope_cte = __scope_context__(SCOPE_CTE)
scope_column = __scope_context__(SCOPE_COLUMN)
def __enter__(self):
if self.parentheses:
self.literal('(')
return self
def __exit__(self, exc_type, exc_val, exc_tb):
if self.parentheses:
self.literal(')')
self.state = self.stack.pop()
@contextmanager
def push_alias(self):
self.alias_manager.push()
yield
self.alias_manager.pop()
def sql(self, obj):
if isinstance(obj, (Node, Context)):
return obj.__sql__(self)
elif is_model(obj):
return obj._meta.table.__sql__(self)
else:
return self.sql(Value(obj))
def literal(self, keyword):
self._sql.append(keyword)
return self
def value(self, value, converter=None, add_param=True):
if converter:
value = converter(value)
elif converter is None and self.state.converter:
# Explicitly check for None so that "False" can be used to signify
# that no conversion should be applied.
value = self.state.converter(value)
if isinstance(value, Node):
with self(converter=None):
return self.sql(value)
elif is_model(value):
# Under certain circumstances, we could end-up treating a model-
# class itself as a value. This check ensures that we drop the
# table alias into the query instead of trying to parameterize a
# model (for instance, passing a model as a function argument).
with self.scope_column():
return self.sql(value)
self._values.append(value)
return self.literal(self.state.param or '?') if add_param else self
def __sql__(self, ctx):
ctx._sql.extend(self._sql)
ctx._values.extend(self._values)
return ctx
def parse(self, node):
return self.sql(node).query()
def query(self):
return ''.join(self._sql), self._values
def query_to_string(query):
# NOTE: this function is not exported by default as it might be misused --
# and this misuse could lead to sql injection vulnerabilities. This
# function is intended for debugging or logging purposes ONLY.
db = getattr(query, '_database', None)
if db is not None:
ctx = db.get_sql_context()
else:
ctx = Context()
sql, params = ctx.sql(query).query()
if not params:
return sql
param = ctx.state.param or '?'
if param == '?':
sql = sql.replace('?', '%s')
return sql % tuple(map(_query_val_transform, params))
def _query_val_transform(v):
# Interpolate parameters.
if isinstance(v, (text_type, datetime.datetime, datetime.date,
datetime.time)):
v = "'%s'" % v
elif isinstance(v, bytes_type):
try:
v = v.decode('utf8')
except UnicodeDecodeError:
v = v.decode('raw_unicode_escape')
v = "'%s'" % v
elif isinstance(v, int):
v = '%s' % int(v) # Also handles booleans -> 1 or 0.
elif v is None:
v = 'NULL'
else:
v = str(v)
return v
# AST.
class Node(object):
_coerce = True
def clone(self):
obj = self.__class__.__new__(self.__class__)
obj.__dict__ = self.__dict__.copy()
return obj
def __sql__(self, ctx):
raise NotImplementedError
@staticmethod
def copy(method):
def inner(self, *args, **kwargs):
clone = self.clone()
method(clone, *args, **kwargs)
return clone
return inner
def coerce(self, _coerce=True):
if _coerce != self._coerce:
clone = self.clone()
clone._coerce = _coerce
return clone
return self
def is_alias(self):
return False
def unwrap(self):
return self
class ColumnFactory(object):
__slots__ = ('node',)
def __init__(self, node):
self.node = node
def __getattr__(self, attr):
return Column(self.node, attr)
class _DynamicColumn(object):
__slots__ = ()
def __get__(self, instance, instance_type=None):
if instance is not None:
return ColumnFactory(instance) # Implements __getattr__().
return self
class _ExplicitColumn(object):
__slots__ = ()
def __get__(self, instance, instance_type=None):
if instance is not None:
raise AttributeError(
'%s specifies columns explicitly, and does not support '
'dynamic column lookups.' % instance)
return self
class Source(Node):
c = _DynamicColumn()
def __init__(self, alias=None):
super(Source, self).__init__()
self._alias = alias
@Node.copy
def alias(self, name):
self._alias = name
def select(self, *columns):
if not columns:
columns = (SQL('*'),)
return Select((self,), columns)
def join(self, dest, join_type=JOIN.INNER, on=None):
return Join(self, dest, join_type, on)
def left_outer_join(self, dest, on=None):
return Join(self, dest, JOIN.LEFT_OUTER, on)
def cte(self, name, recursive=False, columns=None, materialized=None):
return CTE(name, self, recursive=recursive, columns=columns,
materialized=materialized)
def get_sort_key(self, ctx):
if self._alias:
return (self._alias,)
return (ctx.alias_manager[self],)
def apply_alias(self, ctx):
# If we are defining the source, include the "AS alias" declaration. An
# alias is created for the source if one is not already defined.
if ctx.scope == SCOPE_SOURCE:
if self._alias:
ctx.alias_manager[self] = self._alias
ctx.literal(' AS ').sql(Entity(ctx.alias_manager[self]))
return ctx
def apply_column(self, ctx):
if self._alias:
ctx.alias_manager[self] = self._alias
return ctx.sql(Entity(ctx.alias_manager[self]))
class _HashableSource(object):
def __init__(self, *args, **kwargs):
super(_HashableSource, self).__init__(*args, **kwargs)
self._update_hash()
@Node.copy
def alias(self, name):
self._alias = name
self._update_hash()
def _update_hash(self):
self._hash = self._get_hash()
def _get_hash(self):
return hash((self.__class__, self._path, self._alias))
def __hash__(self):
return self._hash
def __eq__(self, other):
if isinstance(other, _HashableSource):
return self._hash == other._hash
return Expression(self, OP.EQ, other)
def __ne__(self, other):
if isinstance(other, _HashableSource):
return self._hash != other._hash
return Expression(self, OP.NE, other)
def _e(op):
def inner(self, rhs):
return Expression(self, op, rhs)
return inner
__lt__ = _e(OP.LT)
__le__ = _e(OP.LTE)
__gt__ = _e(OP.GT)
__ge__ = _e(OP.GTE)
def __bind_database__(meth):
@wraps(meth)
def inner(self, *args, **kwargs):
result = meth(self, *args, **kwargs)
if self._database:
return result.bind(self._database)
return result
return inner
def __join__(join_type=JOIN.INNER, inverted=False):
def method(self, other):
if inverted:
self, other = other, self
return Join(self, other, join_type=join_type)
return method
class BaseTable(Source):
__and__ = __join__(JOIN.INNER)
__add__ = __join__(JOIN.LEFT_OUTER)
__sub__ = __join__(JOIN.RIGHT_OUTER)
__or__ = __join__(JOIN.FULL_OUTER)
__mul__ = __join__(JOIN.CROSS)
__rand__ = __join__(JOIN.INNER, inverted=True)
__radd__ = __join__(JOIN.LEFT_OUTER, inverted=True)
__rsub__ = __join__(JOIN.RIGHT_OUTER, inverted=True)
__ror__ = __join__(JOIN.FULL_OUTER, inverted=True)
__rmul__ = __join__(JOIN.CROSS, inverted=True)
class _BoundTableContext(_callable_context_manager):
def __init__(self, table, database):
self.table = table
self.database = database
def __enter__(self):
self._orig_database = self.table._database
self.table.bind(self.database)
if self.table._model is not None:
self.table._model.bind(self.database)
return self.table
def __exit__(self, exc_type, exc_val, exc_tb):
self.table.bind(self._orig_database)
if self.table._model is not None:
self.table._model.bind(self._orig_database)
class Table(_HashableSource, BaseTable):
def __init__(self, name, columns=None, primary_key=None, schema=None,
alias=None, _model=None, _database=None):
self.__name__ = name
self._columns = columns
self._primary_key = primary_key
self._schema = schema
self._path = (schema, name) if schema else (name,)
self._model = _model
self._database = _database
super(Table, self).__init__(alias=alias)
# Allow tables to restrict what columns are available.
if columns is not None:
self.c = _ExplicitColumn()
for column in columns:
setattr(self, column, Column(self, column))
if primary_key:
col_src = self if self._columns else self.c
self.primary_key = getattr(col_src, primary_key)
else:
self.primary_key = None
def clone(self):
# Ensure a deep copy of the column instances.
return Table(
self.__name__,
columns=self._columns,
primary_key=self._primary_key,
schema=self._schema,
alias=self._alias,
_model=self._model,
_database=self._database)
def bind(self, database=None):
self._database = database
return self
def bind_ctx(self, database=None):
return _BoundTableContext(self, database)
def _get_hash(self):
return hash((self.__class__, self._path, self._alias, self._model))
@__bind_database__
def select(self, *columns):
if not columns and self._columns:
columns = [Column(self, column) for column in self._columns]
return Select((self,), columns)
@__bind_database__
def insert(self, insert=None, columns=None, **kwargs):
if kwargs:
insert = {} if insert is None else insert
src = self if self._columns else self.c
for key, value in kwargs.items():
insert[getattr(src, key)] = value
return Insert(self, insert=insert, columns=columns)
@__bind_database__
def replace(self, insert=None, columns=None, **kwargs):
return (self
.insert(insert=insert, columns=columns)
.on_conflict('REPLACE'))
@__bind_database__
def update(self, update=None, **kwargs):
if kwargs:
update = {} if update is None else update
for key, value in kwargs.items():
src = self if self._columns else self.c
update[getattr(src, key)] = value
return Update(self, update=update)
@__bind_database__
def delete(self):
return Delete(self)
def __sql__(self, ctx):
if ctx.scope == SCOPE_VALUES:
# Return the quoted table name.
return ctx.sql(Entity(*self._path))
if self._alias:
ctx.alias_manager[self] = self._alias
if ctx.scope == SCOPE_SOURCE:
# Define the table and its alias.
return self.apply_alias(ctx.sql(Entity(*self._path)))
else:
# Refer to the table using the alias.
return self.apply_column(ctx)
class Join(BaseTable):
def __init__(self, lhs, rhs, join_type=JOIN.INNER, on=None, alias=None):
super(Join, self).__init__(alias=alias)
self.lhs = lhs
self.rhs = rhs
self.join_type = join_type
self._on = on
def on(self, predicate):
self._on = predicate
return self
def __sql__(self, ctx):
(ctx
.sql(self.lhs)
.literal(' %s ' % self.join_type)
.sql(self.rhs))
if self._on is not None:
ctx.literal(' ON ').sql(self._on)
return ctx
class ValuesList(_HashableSource, BaseTable):
def __init__(self, values, columns=None, alias=None):
self._values = values
self._columns = columns
super(ValuesList, self).__init__(alias=alias)
def _get_hash(self):
return hash((self.__class__, id(self._values), self._alias))
@Node.copy
def columns(self, *names):
self._columns = names
def __sql__(self, ctx):
if self._alias:
ctx.alias_manager[self] = self._alias
if ctx.scope == SCOPE_SOURCE or ctx.scope == SCOPE_NORMAL:
with ctx(parentheses=not ctx.parentheses):
ctx = (ctx
.literal('VALUES ')
.sql(CommaNodeList([
EnclosedNodeList(row) for row in self._values])))
if ctx.scope == SCOPE_SOURCE:
ctx.literal(' AS ').sql(Entity(ctx.alias_manager[self]))
if self._columns:
entities = [Entity(c) for c in self._columns]
ctx.sql(EnclosedNodeList(entities))
else:
ctx.sql(Entity(ctx.alias_manager[self]))
return ctx
class CTE(_HashableSource, Source):
def __init__(self, name, query, recursive=False, columns=None,
materialized=None):
self._alias = name
self._query = query
self._recursive = recursive
self._materialized = materialized
if columns is not None:
columns = [Entity(c) if isinstance(c, basestring) else c
for c in columns]
self._columns = columns
query._cte_list = ()
super(CTE, self).__init__(alias=name)
def select_from(self, *columns):
if not columns:
raise ValueError('select_from() must specify one or more columns '
'from the CTE to select.')
query = (Select((self,), columns)
.with_cte(self)
.bind(self._query._database))
try:
query = query.objects(self._query.model)
except AttributeError:
pass
return query
def _get_hash(self):
return hash((self.__class__, self._alias, id(self._query)))
def union_all(self, rhs):
clone = self._query.clone()
return CTE(self._alias, clone + rhs, self._recursive, self._columns)
__add__ = union_all
def union(self, rhs):
clone = self._query.clone()
return CTE(self._alias, clone | rhs, self._recursive, self._columns)
__or__ = union
def __sql__(self, ctx):
if ctx.scope != SCOPE_CTE:
return ctx.sql(Entity(self._alias))
with ctx.push_alias():
ctx.alias_manager[self] = self._alias
ctx.sql(Entity(self._alias))
if self._columns:
ctx.literal(' ').sql(EnclosedNodeList(self._columns))
ctx.literal(' AS ')
if self._materialized:
ctx.literal('MATERIALIZED ')
elif self._materialized is False:
ctx.literal('NOT MATERIALIZED ')
with ctx.scope_normal(parentheses=True):
ctx.sql(self._query)
return ctx
class ColumnBase(Node):
_converter = None
@Node.copy
def converter(self, converter=None):
self._converter = converter
def alias(self, alias):
if alias:
return Alias(self, alias)
return self
def unalias(self):
return self
def cast(self, as_type):
return Cast(self, as_type)
def asc(self, collation=None, nulls=None):
return Asc(self, collation=collation, nulls=nulls)
__pos__ = asc
def desc(self, collation=None, nulls=None):
return Desc(self, collation=collation, nulls=nulls)
__neg__ = desc
def __invert__(self):
return Negated(self)
def _e(op, inv=False):
"""
Lightweight factory which returns a method that builds an Expression
consisting of the left-hand and right-hand operands, using `op`.
"""
def inner(self, rhs):
if inv:
return Expression(rhs, op, self)
return Expression(self, op, rhs)
return inner
__and__ = _e(OP.AND)
__or__ = _e(OP.OR)
__add__ = _e(OP.ADD)
__sub__ = _e(OP.SUB)
__mul__ = _e(OP.MUL)
__div__ = __truediv__ = _e(OP.DIV)
__xor__ = _e(OP.XOR)
__radd__ = _e(OP.ADD, inv=True)
__rsub__ = _e(OP.SUB, inv=True)
__rmul__ = _e(OP.MUL, inv=True)
__rdiv__ = __rtruediv__ = _e(OP.DIV, inv=True)
__rand__ = _e(OP.AND, inv=True)
__ror__ = _e(OP.OR, inv=True)
__rxor__ = _e(OP.XOR, inv=True)
def __eq__(self, rhs):
op = OP.IS if rhs is None else OP.EQ
return Expression(self, op, rhs)
def __ne__(self, rhs):
op = OP.IS_NOT if rhs is None else OP.NE
return Expression(self, op, rhs)
__lt__ = _e(OP.LT)
__le__ = _e(OP.LTE)
__gt__ = _e(OP.GT)
__ge__ = _e(OP.GTE)
__lshift__ = _e(OP.IN)
__rshift__ = _e(OP.IS)
__mod__ = _e(OP.LIKE)
__pow__ = _e(OP.ILIKE)
like = _e(OP.LIKE)
ilike = _e(OP.ILIKE)
bin_and = _e(OP.BIN_AND)
bin_or = _e(OP.BIN_OR)
in_ = _e(OP.IN)
not_in = _e(OP.NOT_IN)
regexp = _e(OP.REGEXP)
# Special expressions.
def is_null(self, is_null=True):
op = OP.IS if is_null else OP.IS_NOT
return Expression(self, op, None)
def _escape_like_expr(self, s, template):
if s.find('_') >= 0 or s.find('%') >= 0 or s.find('\\') >= 0:
s = s.replace('\\', '\\\\').replace('_', '\\_').replace('%', '\\%')
return NodeList((template % s, SQL('ESCAPE'), '\\'))
return template % s
def contains(self, rhs):
if isinstance(rhs, Node):
rhs = Expression('%', OP.CONCAT,
Expression(rhs, OP.CONCAT, '%'))
else:
rhs = self._escape_like_expr(rhs, '%%%s%%')
return Expression(self, OP.ILIKE, rhs)
def startswith(self, rhs):
if isinstance(rhs, Node):
rhs = Expression(rhs, OP.CONCAT, '%')
else:
rhs = self._escape_like_expr(rhs, '%s%%')
return Expression(self, OP.ILIKE, rhs)
def endswith(self, rhs):
if isinstance(rhs, Node):
rhs = Expression('%', OP.CONCAT, rhs)
else:
rhs = self._escape_like_expr(rhs, '%%%s')
return Expression(self, OP.ILIKE, rhs)
def between(self, lo, hi):
return Expression(self, OP.BETWEEN, NodeList((lo, SQL('AND'), hi)))
def concat(self, rhs):
return StringExpression(self, OP.CONCAT, rhs)
def regexp(self, rhs):
return Expression(self, OP.REGEXP, rhs)
def iregexp(self, rhs):
return Expression(self, OP.IREGEXP, rhs)
def __getitem__(self, item):
if isinstance(item, slice):
if item.start is None or item.stop is None:
raise ValueError('BETWEEN range must have both a start- and '
'end-point.')
return self.between(item.start, item.stop)
return self == item
def distinct(self):
return NodeList((SQL('DISTINCT'), self))
def collate(self, collation):
return NodeList((self, SQL('COLLATE %s' % collation)))
def get_sort_key(self, ctx):
return ()
class Column(ColumnBase):
def __init__(self, source, name):
self.source = source
self.name = name
def get_sort_key(self, ctx):
if ctx.scope == SCOPE_VALUES:
return (self.name,)
else:
return self.source.get_sort_key(ctx) + (self.name,)
def __hash__(self):
return hash((self.source, self.name))
def __sql__(self, ctx):
if ctx.scope == SCOPE_VALUES:
return ctx.sql(Entity(self.name))
else:
with ctx.scope_column():
return ctx.sql(self.source).literal('.').sql(Entity(self.name))
class WrappedNode(ColumnBase):
def __init__(self, node):
self.node = node
self._coerce = getattr(node, '_coerce', True)
self._converter = getattr(node, '_converter', None)
def is_alias(self):
return self.node.is_alias()
def unwrap(self):
return self.node.unwrap()
class EntityFactory(object):
__slots__ = ('node',)
def __init__(self, node):
self.node = node
def __getattr__(self, attr):
return Entity(self.node, attr)
class _DynamicEntity(object):
__slots__ = ()
def __get__(self, instance, instance_type=None):
if instance is not None:
return EntityFactory(instance._alias) # Implements __getattr__().
return self
class Alias(WrappedNode):
c = _DynamicEntity()
def __init__(self, node, alias):
super(Alias, self).__init__(node)
self._alias = alias
def __hash__(self):
return hash(self._alias)
def alias(self, alias=None):
if alias is None:
return self.node
else:
return Alias(self.node, alias)
def unalias(self):
return self.node
def is_alias(self):
return True
def __sql__(self, ctx):
if ctx.scope == SCOPE_SOURCE:
return (ctx
.sql(self.node)
.literal(' AS ')
.sql(Entity(self._alias)))
else:
return ctx.sql(Entity(self._alias))
class Negated(WrappedNode):
def __invert__(self):
return self.node
def __sql__(self, ctx):
return ctx.literal('NOT ').sql(self.node)
class BitwiseMixin(object):
def __and__(self, other):
return self.bin_and(other)
def __or__(self, other):
return self.bin_or(other)
def __sub__(self, other):
return self.bin_and(other.bin_negated())
def __invert__(self):
return BitwiseNegated(self)
class BitwiseNegated(BitwiseMixin, WrappedNode):
def __invert__(self):
return self.node
def __sql__(self, ctx):
if ctx.state.operations:
op_sql = ctx.state.operations.get(self.op, self.op)
else:
op_sql = self.op
return ctx.literal(op_sql).sql(self.node)
class Value(ColumnBase):
def __init__(self, value, converter=None, unpack=True):
self.value = value
self.converter = converter
self.multi = unpack and isinstance(self.value, multi_types)
if self.multi:
self.values = []
for item in self.value:
if isinstance(item, Node):
self.values.append(item)
else:
self.values.append(Value(item, self.converter))
def __sql__(self, ctx):
if self.multi:
# For multi-part values (e.g. lists of IDs).
return ctx.sql(EnclosedNodeList(self.values))
return ctx.value(self.value, self.converter)
def AsIs(value):
return Value(value, unpack=False)
class Cast(WrappedNode):
def __init__(self, node, cast):
super(Cast, self).__init__(node)
self._cast = cast
self._coerce = False
def __sql__(self, ctx):
return (ctx
.literal('CAST(')
.sql(self.node)
.literal(' AS %s)' % self._cast))
class Ordering(WrappedNode):
def __init__(self, node, direction, collation=None, nulls=None):
super(Ordering, self).__init__(node)
self.direction = direction
self.collation = collation
self.nulls = nulls
if nulls and nulls.lower() not in ('first', 'last'):
raise ValueError('Ordering nulls= parameter must be "first" or '
'"last", got: %s' % nulls)
def collate(self, collation=None):
return Ordering(self.node, self.direction, collation)
def _null_ordering_case(self, nulls):
if nulls.lower() == 'last':
ifnull, notnull = 1, 0
elif nulls.lower() == 'first':
ifnull, notnull = 0, 1
else:
raise ValueError('unsupported value for nulls= ordering.')
return Case(None, ((self.node.is_null(), ifnull),), notnull)
def __sql__(self, ctx):
if self.nulls and not ctx.state.nulls_ordering:
ctx.sql(self._null_ordering_case(self.nulls)).literal(', ')
ctx.sql(self.node).literal(' %s' % self.direction)
if self.collation:
ctx.literal(' COLLATE %s' % self.collation)
if self.nulls and ctx.state.nulls_ordering:
ctx.literal(' NULLS %s' % self.nulls)
return ctx
def Asc(node, collation=None, nulls=None):
return Ordering(node, 'ASC', collation, nulls)
def Desc(node, collation=None, nulls=None):
return Ordering(node, 'DESC', collation, nulls)
class Expression(ColumnBase):
def __init__(self, lhs, op, rhs, flat=False):
self.lhs = lhs
self.op = op
self.rhs = rhs
self.flat = flat
def __sql__(self, ctx):
overrides = {'parentheses': not self.flat, 'in_expr': True}
# First attempt to unwrap the node on the left-hand-side, so that we
# can get at the underlying Field if one is present.
node = raw_node = self.lhs
if isinstance(raw_node, WrappedNode):
node = raw_node.unwrap()
# Set up the appropriate converter if we have a field on the left side.
if isinstance(node, Field) and raw_node._coerce:
overrides['converter'] = node.db_value
overrides['is_fk_expr'] = isinstance(node, ForeignKeyField)
else:
overrides['converter'] = None
if ctx.state.operations:
op_sql = ctx.state.operations.get(self.op, self.op)
else:
op_sql = self.op
with ctx(**overrides):
# Postgresql reports an error for IN/NOT IN (), so convert to
# the equivalent boolean expression.
op_in = self.op == OP.IN or self.op == OP.NOT_IN
if op_in and ctx.as_new().parse(self.rhs)[0] == '()':
return ctx.literal('0 = 1' if self.op == OP.IN else '1 = 1')
return (ctx
.sql(self.lhs)
.literal(' %s ' % op_sql)
.sql(self.rhs))
class StringExpression(Expression):
def __add__(self, rhs):
return self.concat(rhs)
def __radd__(self, lhs):
return StringExpression(lhs, OP.CONCAT, self)
class Entity(ColumnBase):
def __init__(self, *path):
self._path = [part.replace('"', '""') for part in path if part]
def __getattr__(self, attr):
return Entity(*self._path + [attr])
def get_sort_key(self, ctx):
return tuple(self._path)
def __hash__(self):
return hash((self.__class__.__name__, tuple(self._path)))
def __sql__(self, ctx):
return ctx.literal(quote(self._path, ctx.state.quote or '""'))
class SQL(ColumnBase):
def __init__(self, sql, params=None):
self.sql = sql
self.params = params
def __sql__(self, ctx):
ctx.literal(self.sql)
if self.params:
for param in self.params:
ctx.value(param, False, add_param=False)
return ctx
def Check(constraint, name=None):
check = SQL('CHECK (%s)' % constraint)
if not name:
return check
return NodeList((SQL('CONSTRAINT'), Entity(name), check))
class Function(ColumnBase):
def __init__(self, name, arguments, coerce=True, python_value=None):
self.name = name
self.arguments = arguments
self._filter = None
self._order_by = None
self._python_value = python_value
if name and name.lower() in ('sum', 'count', 'cast', 'array_agg'):
self._coerce = False
else:
self._coerce = coerce
def __getattr__(self, attr):
def decorator(*args, **kwargs):
return Function(attr, args, **kwargs)
return decorator
@Node.copy
def filter(self, where=None):
self._filter = where
@Node.copy
def order_by(self, *ordering):
self._order_by = ordering
@Node.copy
def python_value(self, func=None):
self._python_value = func
def over(self, partition_by=None, order_by=None, start=None, end=None,
frame_type=None, window=None, exclude=None):
if isinstance(partition_by, Window) and window is None:
window = partition_by
if window is not None:
node = WindowAlias(window)
else:
node = Window(partition_by=partition_by, order_by=order_by,
start=start, end=end, frame_type=frame_type,
exclude=exclude, _inline=True)
return NodeList((self, SQL('OVER'), node))
def __sql__(self, ctx):
ctx.literal(self.name)
if not len(self.arguments):
ctx.literal('()')
else:
args = self.arguments
# If this is an ordered aggregate, then we will modify the last
# argument to append the ORDER BY ... clause. We do this to avoid
# double-wrapping any expression args in parentheses, as NodeList
# has a special check (hack) in place to work around this.
if self._order_by:
args = list(args)
args[-1] = NodeList((args[-1], SQL('ORDER BY'),
CommaNodeList(self._order_by)))
with ctx(in_function=True, function_arg_count=len(self.arguments)):
ctx.sql(EnclosedNodeList([
(arg if isinstance(arg, Node) else Value(arg, False))
for arg in args]))
if self._filter:
ctx.literal(' FILTER (WHERE ').sql(self._filter).literal(')')
return ctx
fn = Function(None, None)
class Window(Node):
# Frame start/end and frame exclusion.
CURRENT_ROW = SQL('CURRENT ROW')
GROUP = SQL('GROUP')
TIES = SQL('TIES')
NO_OTHERS = SQL('NO OTHERS')
# Frame types.
GROUPS = 'GROUPS'
RANGE = 'RANGE'
ROWS = 'ROWS'
def __init__(self, partition_by=None, order_by=None, start=None, end=None,
frame_type=None, extends=None, exclude=None, alias=None,
_inline=False):
super(Window, self).__init__()
if start is not None and not isinstance(start, SQL):
start = SQL(start)
if end is not None and not isinstance(end, SQL):
end = SQL(end)
self.partition_by = ensure_tuple(partition_by)
self.order_by = ensure_tuple(order_by)
self.start = start
self.end = end
if self.start is None and self.end is not None:
raise ValueError('Cannot specify WINDOW end without start.')
self._alias = alias or 'w'
self._inline = _inline
self.frame_type = frame_type
self._extends = extends
self._exclude = exclude
def alias(self, alias=None):
self._alias = alias or 'w'
return self
@Node.copy
def as_range(self):
self.frame_type = Window.RANGE
@Node.copy
def as_rows(self):
self.frame_type = Window.ROWS
@Node.copy
def as_groups(self):
self.frame_type = Window.GROUPS
@Node.copy
def extends(self, window=None):
self._extends = window
@Node.copy
def exclude(self, frame_exclusion=None):
if isinstance(frame_exclusion, basestring):
frame_exclusion = SQL(frame_exclusion)
self._exclude = frame_exclusion
@staticmethod
def following(value=None):
if value is None:
return SQL('UNBOUNDED FOLLOWING')
return SQL('%d FOLLOWING' % value)
@staticmethod
def preceding(value=None):
if value is None:
return SQL('UNBOUNDED PRECEDING')
return SQL('%d PRECEDING' % value)
def __sql__(self, ctx):
if ctx.scope != SCOPE_SOURCE and not self._inline:
ctx.literal(self._alias)
ctx.literal(' AS ')
with ctx(parentheses=True):
parts = []
if self._extends is not None:
ext = self._extends
if isinstance(ext, Window):
ext = SQL(ext._alias)
elif isinstance(ext, basestring):
ext = SQL(ext)
parts.append(ext)
if self.partition_by:
parts.extend((
SQL('PARTITION BY'),
CommaNodeList(self.partition_by)))
if self.order_by:
parts.extend((
SQL('ORDER BY'),
CommaNodeList(self.order_by)))
if self.start is not None and self.end is not None:
frame = self.frame_type or 'ROWS'
parts.extend((
SQL('%s BETWEEN' % frame),
self.start,
SQL('AND'),
self.end))
elif self.start is not None:
parts.extend((SQL(self.frame_type or 'ROWS'), self.start))
elif self.frame_type is not None:
parts.append(SQL('%s UNBOUNDED PRECEDING' % self.frame_type))
if self._exclude is not None:
parts.extend((SQL('EXCLUDE'), self._exclude))
ctx.sql(NodeList(parts))
return ctx
class WindowAlias(Node):
def __init__(self, window):
self.window = window
def alias(self, window_alias):
self.window._alias = window_alias
return self
def __sql__(self, ctx):
return ctx.literal(self.window._alias or 'w')
class ForUpdate(Node):
def __init__(self, expr, of=None, nowait=None):
expr = 'FOR UPDATE' if expr is True else expr
if expr.lower().endswith('nowait'):
expr = expr[:-7] # Strip off the "nowait" bit.
nowait = True
self._expr = expr
if of is not None and not isinstance(of, (list, set, tuple)):
of = (of,)
self._of = of
self._nowait = nowait
def __sql__(self, ctx):
ctx.literal(self._expr)
if self._of is not None:
ctx.literal(' OF ').sql(CommaNodeList(self._of))
if self._nowait:
ctx.literal(' NOWAIT')
return ctx
def Case(predicate, expression_tuples, default=None):
clauses = [SQL('CASE')]
if predicate is not None:
clauses.append(predicate)
for expr, value in expression_tuples:
clauses.extend((SQL('WHEN'), expr, SQL('THEN'), value))
if default is not None:
clauses.extend((SQL('ELSE'), default))
clauses.append(SQL('END'))
return NodeList(clauses)
class NodeList(ColumnBase):
def __init__(self, nodes, glue=' ', parens=False):
self.nodes = nodes
self.glue = glue
self.parens = parens
if parens and len(self.nodes) == 1 and \
isinstance(self.nodes[0], Expression) and \
not self.nodes[0].flat:
# Hack to avoid double-parentheses.
self.nodes = (self.nodes[0].clone(),)
self.nodes[0].flat = True
def __sql__(self, ctx):
n_nodes = len(self.nodes)
if n_nodes == 0:
return ctx.literal('()') if self.parens else ctx
with ctx(parentheses=self.parens):
for i in range(n_nodes - 1):
ctx.sql(self.nodes[i])
ctx.literal(self.glue)
ctx.sql(self.nodes[n_nodes - 1])
return ctx
def CommaNodeList(nodes):
return NodeList(nodes, ', ')
def EnclosedNodeList(nodes):
return NodeList(nodes, ', ', True)
class _Namespace(Node):
__slots__ = ('_name',)
def __init__(self, name):
self._name = name
def __getattr__(self, attr):
return NamespaceAttribute(self, attr)
__getitem__ = __getattr__
class NamespaceAttribute(ColumnBase):
def __init__(self, namespace, attribute):
self._namespace = namespace
self._attribute = attribute
def __sql__(self, ctx):
return (ctx
.literal(self._namespace._name + '.')
.sql(Entity(self._attribute)))
EXCLUDED = _Namespace('EXCLUDED')
class DQ(ColumnBase):
def __init__(self, **query):
super(DQ, self).__init__()
self.query = query
self._negated = False
@Node.copy
def __invert__(self):
self._negated = not self._negated
def clone(self):
node = DQ(**self.query)
node._negated = self._negated
return node
#: Represent a row tuple.
Tuple = lambda *a: EnclosedNodeList(a)
class QualifiedNames(WrappedNode):
def __sql__(self, ctx):
with ctx.scope_column():
return ctx.sql(self.node)
def qualify_names(node):
# Search a node heirarchy to ensure that any column-like objects are
# referenced using fully-qualified names.
if isinstance(node, Expression):
return node.__class__(qualify_names(node.lhs), node.op,
qualify_names(node.rhs), node.flat)
elif isinstance(node, ColumnBase):
return QualifiedNames(node)
return node
class OnConflict(Node):
def __init__(self, action=None, update=None, preserve=None, where=None,
conflict_target=None, conflict_where=None,
conflict_constraint=None):
self._action = action
self._update = update
self._preserve = ensure_tuple(preserve)
self._where = where
if conflict_target is not None and conflict_constraint is not None:
raise ValueError('only one of "conflict_target" and '
'"conflict_constraint" may be specified.')
self._conflict_target = ensure_tuple(conflict_target)
self._conflict_where = conflict_where
self._conflict_constraint = conflict_constraint
def get_conflict_statement(self, ctx, query):
return ctx.state.conflict_statement(self, query)
def get_conflict_update(self, ctx, query):
return ctx.state.conflict_update(self, query)
@Node.copy
def preserve(self, *columns):
self._preserve = columns
@Node.copy
def update(self, _data=None, **kwargs):
if _data and kwargs and not isinstance(_data, dict):
raise ValueError('Cannot mix data with keyword arguments in the '
'OnConflict update method.')
_data = _data or {}
if kwargs:
_data.update(kwargs)
self._update = _data
@Node.copy
def where(self, *expressions):
if self._where is not None:
expressions = (self._where,) + expressions
self._where = reduce(operator.and_, expressions)
@Node.copy
def conflict_target(self, *constraints):
self._conflict_constraint = None
self._conflict_target = constraints
@Node.copy
def conflict_where(self, *expressions):
if self._conflict_where is not None:
expressions = (self._conflict_where,) + expressions
self._conflict_where = reduce(operator.and_, expressions)
@Node.copy
def conflict_constraint(self, constraint):
self._conflict_constraint = constraint
self._conflict_target = None
def database_required(method):
@wraps(method)
def inner(self, database=None, *args, **kwargs):
database = self._database if database is None else database
if not database:
raise InterfaceError('Query must be bound to a database in order '
'to call "%s".' % method.__name__)
return method(self, database, *args, **kwargs)
return inner
# BASE QUERY INTERFACE.
class BaseQuery(Node):
default_row_type = ROW.DICT
def __init__(self, _database=None, **kwargs):
self._database = _database
self._cursor_wrapper = None
self._row_type = None
self._constructor = None
super(BaseQuery, self).__init__(**kwargs)
def bind(self, database=None):
self._database = database
return self
def clone(self):
query = super(BaseQuery, self).clone()
query._cursor_wrapper = None
return query
@Node.copy
def dicts(self, as_dict=True):
self._row_type = ROW.DICT if as_dict else None
return self
@Node.copy
def tuples(self, as_tuple=True):
self._row_type = ROW.TUPLE if as_tuple else None
return self
@Node.copy
def namedtuples(self, as_namedtuple=True):
self._row_type = ROW.NAMED_TUPLE if as_namedtuple else None
return self
@Node.copy
def objects(self, constructor=None):
self._row_type = ROW.CONSTRUCTOR if constructor else None
self._constructor = constructor
return self
def _get_cursor_wrapper(self, cursor):
row_type = self._row_type or self.default_row_type
if row_type == ROW.DICT:
return DictCursorWrapper(cursor)
elif row_type == ROW.TUPLE:
return CursorWrapper(cursor)
elif row_type == ROW.NAMED_TUPLE:
return NamedTupleCursorWrapper(cursor)
elif row_type == ROW.CONSTRUCTOR:
return ObjectCursorWrapper(cursor, self._constructor)
else:
raise ValueError('Unrecognized row type: "%s".' % row_type)
def __sql__(self, ctx):
raise NotImplementedError
def sql(self):
if self._database:
context = self._database.get_sql_context()
else:
context = Context()
return context.parse(self)
@database_required
def execute(self, database):
return self._execute(database)
def _execute(self, database):
raise NotImplementedError
def iterator(self, database=None):
return iter(self.execute(database).iterator())
def _ensure_execution(self):
if not self._cursor_wrapper:
if not self._database:
raise ValueError('Query has not been executed.')
self.execute()
def __iter__(self):
self._ensure_execution()
return iter(self._cursor_wrapper)
def __getitem__(self, value):
self._ensure_execution()
if isinstance(value, slice):
index = value.stop
else:
index = value
if index is not None:
index = index + 1 if index >= 0 else 0
self._cursor_wrapper.fill_cache(index)
return self._cursor_wrapper.row_cache[value]
def __len__(self):
self._ensure_execution()
return len(self._cursor_wrapper)
def __str__(self):
return query_to_string(self)
class RawQuery(BaseQuery):
def __init__(self, sql=None, params=None, **kwargs):
super(RawQuery, self).__init__(**kwargs)
self._sql = sql
self._params = params
def __sql__(self, ctx):
ctx.literal(self._sql)
if self._params:
for param in self._params:
ctx.value(param, add_param=False)
return ctx
def _execute(self, database):
if self._cursor_wrapper is None:
cursor = database.execute(self)
self._cursor_wrapper = self._get_cursor_wrapper(cursor)
return self._cursor_wrapper
class Query(BaseQuery):
def __init__(self, where=None, order_by=None, limit=None, offset=None,
**kwargs):
super(Query, self).__init__(**kwargs)
self._where = where
self._order_by = order_by
self._limit = limit
self._offset = offset
self._cte_list = None
@Node.copy
def with_cte(self, *cte_list):
self._cte_list = cte_list
@Node.copy
def where(self, *expressions):
if self._where is not None:
expressions = (self._where,) + expressions
self._where = reduce(operator.and_, expressions)
@Node.copy
def orwhere(self, *expressions):
if self._where is not None:
expressions = (self._where,) + expressions
self._where = reduce(operator.or_, expressions)
@Node.copy
def order_by(self, *values):
self._order_by = values
@Node.copy
def order_by_extend(self, *values):
self._order_by = ((self._order_by or ()) + values) or None
@Node.copy
def limit(self, value=None):
self._limit = value
@Node.copy
def offset(self, value=None):
self._offset = value
@Node.copy
def paginate(self, page, paginate_by=20):
if page > 0:
page -= 1
self._limit = paginate_by
self._offset = page * paginate_by
def _apply_ordering(self, ctx):
if self._order_by:
(ctx
.literal(' ORDER BY ')
.sql(CommaNodeList(self._order_by)))
if self._limit is not None or (self._offset is not None and
ctx.state.limit_max):
limit = ctx.state.limit_max if self._limit is None else self._limit
ctx.literal(' LIMIT ').sql(limit)
if self._offset is not None:
ctx.literal(' OFFSET ').sql(self._offset)
return ctx
def __sql__(self, ctx):
if self._cte_list:
# The CTE scope is only used at the very beginning of the query,
# when we are describing the various CTEs we will be using.
recursive = any(cte._recursive for cte in self._cte_list)
# Explicitly disable the "subquery" flag here, so as to avoid
# unnecessary parentheses around subsequent selects.
with ctx.scope_cte(subquery=False):
(ctx
.literal('WITH RECURSIVE ' if recursive else 'WITH ')
.sql(CommaNodeList(self._cte_list))
.literal(' '))
return ctx
def __compound_select__(operation, inverted=False):
@__bind_database__
def method(self, other):
if inverted:
self, other = other, self
return CompoundSelectQuery(self, operation, other)
return method
class SelectQuery(Query):
union_all = __add__ = __compound_select__('UNION ALL')
union = __or__ = __compound_select__('UNION')
intersect = __and__ = __compound_select__('INTERSECT')
except_ = __sub__ = __compound_select__('EXCEPT')
__radd__ = __compound_select__('UNION ALL', inverted=True)
__ror__ = __compound_select__('UNION', inverted=True)
__rand__ = __compound_select__('INTERSECT', inverted=True)
__rsub__ = __compound_select__('EXCEPT', inverted=True)
def select_from(self, *columns):
if not columns:
raise ValueError('select_from() must specify one or more columns.')
query = (Select((self,), columns)
.bind(self._database))
if getattr(self, 'model', None) is not None:
# Bind to the sub-select's model type, if defined.
query = query.objects(self.model)
return query
class SelectBase(_HashableSource, Source, SelectQuery):
def _get_hash(self):
return hash((self.__class__, self._alias or id(self)))
def _execute(self, database):
if self._cursor_wrapper is None:
cursor = database.execute(self)
self._cursor_wrapper = self._get_cursor_wrapper(cursor)
return self._cursor_wrapper
@database_required
def peek(self, database, n=1):
rows = self.execute(database)[:n]
if rows:
return rows[0] if n == 1 else rows
@database_required
def first(self, database, n=1):
if self._limit != n:
self._limit = n
self._cursor_wrapper = None
return self.peek(database, n=n)
@database_required
def scalar(self, database, as_tuple=False):
row = self.tuples().peek(database)
return row[0] if row and not as_tuple else row
@database_required
def count(self, database, clear_limit=False):
clone = self.order_by().alias('_wrapped')
if clear_limit:
clone._limit = clone._offset = None
try:
if clone._having is None and clone._group_by is None and \
clone._windows is None and clone._distinct is None and \
clone._simple_distinct is not True:
clone = clone.select(SQL('1'))
except AttributeError:
pass
return Select([clone], [fn.COUNT(SQL('1'))]).scalar(database)
@database_required
def exists(self, database):
clone = self.columns(SQL('1'))
clone._limit = 1
clone._offset = None
return bool(clone.scalar())
@database_required
def get(self, database):
self._cursor_wrapper = None
try:
return self.execute(database)[0]
except IndexError:
pass
# QUERY IMPLEMENTATIONS.
class CompoundSelectQuery(SelectBase):
def __init__(self, lhs, op, rhs):
super(CompoundSelectQuery, self).__init__()
self.lhs = lhs
self.op = op
self.rhs = rhs
@property
def _returning(self):
return self.lhs._returning
@database_required
def exists(self, database):
query = Select((self.limit(1),), (SQL('1'),)).bind(database)
return bool(query.scalar())
def _get_query_key(self):
return (self.lhs.get_query_key(), self.rhs.get_query_key())
def _wrap_parens(self, ctx, subq):
csq_setting = ctx.state.compound_select_parentheses
if not csq_setting or csq_setting == CSQ_PARENTHESES_NEVER:
return False
elif csq_setting == CSQ_PARENTHESES_ALWAYS:
return True
elif csq_setting == CSQ_PARENTHESES_UNNESTED:
if ctx.state.in_expr or ctx.state.in_function:
# If this compound select query is being used inside an
# expression, e.g., an IN or EXISTS().
return False
# If the query on the left or right is itself a compound select
# query, then we do not apply parentheses. However, if it is a
# regular SELECT query, we will apply parentheses.
return not isinstance(subq, CompoundSelectQuery)
def __sql__(self, ctx):
if ctx.scope == SCOPE_COLUMN:
return self.apply_column(ctx)
# Call parent method to handle any CTEs.
super(CompoundSelectQuery, self).__sql__(ctx)
outer_parens = ctx.subquery or (ctx.scope == SCOPE_SOURCE)
with ctx(parentheses=outer_parens):
# Should the left-hand query be wrapped in parentheses?
lhs_parens = self._wrap_parens(ctx, self.lhs)
with ctx.scope_normal(parentheses=lhs_parens, subquery=False):
ctx.sql(self.lhs)
ctx.literal(' %s ' % self.op)
with ctx.push_alias():
# Should the right-hand query be wrapped in parentheses?
rhs_parens = self._wrap_parens(ctx, self.rhs)
with ctx.scope_normal(parentheses=rhs_parens, subquery=False):
ctx.sql(self.rhs)
# Apply ORDER BY, LIMIT, OFFSET. We use the "values" scope so that
# entity names are not fully-qualified. This is a bit of a hack, as
# we're relying on the logic in Column.__sql__() to not fully
# qualify column names.
with ctx.scope_values():
self._apply_ordering(ctx)
return self.apply_alias(ctx)
class Select(SelectBase):
def __init__(self, from_list=None, columns=None, group_by=None,
having=None, distinct=None, windows=None, for_update=None,
for_update_of=None, nowait=None, lateral=None, **kwargs):
super(Select, self).__init__(**kwargs)
self._from_list = (list(from_list) if isinstance(from_list, tuple)
else from_list) or []
self._returning = columns
self._group_by = group_by
self._having = having
self._windows = None
self._for_update = for_update # XXX: consider reorganizing.
self._for_update_of = for_update_of
self._for_update_nowait = nowait
self._lateral = lateral
self._distinct = self._simple_distinct = None
if distinct:
if isinstance(distinct, bool):
self._simple_distinct = distinct
else:
self._distinct = distinct
self._cursor_wrapper = None
def clone(self):
clone = super(Select, self).clone()
if clone._from_list:
clone._from_list = list(clone._from_list)
return clone
@Node.copy
def columns(self, *columns, **kwargs):
self._returning = columns
select = columns
@Node.copy
def select_extend(self, *columns):
self._returning = tuple(self._returning) + columns
@Node.copy
def from_(self, *sources):
self._from_list = list(sources)
@Node.copy
def join(self, dest, join_type=JOIN.INNER, on=None):
if not self._from_list:
raise ValueError('No sources to join on.')
item = self._from_list.pop()
self._from_list.append(Join(item, dest, join_type, on))
@Node.copy
def group_by(self, *columns):
grouping = []
for column in columns:
if isinstance(column, Table):
if not column._columns:
raise ValueError('Cannot pass a table to group_by() that '
'does not have columns explicitly '
'declared.')
grouping.extend([getattr(column, col_name)
for col_name in column._columns])
else:
grouping.append(column)
self._group_by = grouping
def group_by_extend(self, *values):
"""@Node.copy used from group_by() call"""
group_by = tuple(self._group_by or ()) + values
return self.group_by(*group_by)
@Node.copy
def having(self, *expressions):
if self._having is not None:
expressions = (self._having,) + expressions
self._having = reduce(operator.and_, expressions)
@Node.copy
def distinct(self, *columns):
if len(columns) == 1 and (columns[0] is True or columns[0] is False):
self._simple_distinct = columns[0]
else:
self._simple_distinct = False
self._distinct = columns
@Node.copy
def window(self, *windows):
self._windows = windows if windows else None
@Node.copy
def for_update(self, for_update=True, of=None, nowait=None):
if not for_update and (of is not None or nowait):
for_update = True
self._for_update = for_update
self._for_update_of = of
self._for_update_nowait = nowait
@Node.copy
def lateral(self, lateral=True):
self._lateral = lateral
def _get_query_key(self):
return self._alias
def __sql_selection__(self, ctx, is_subquery=False):
return ctx.sql(CommaNodeList(self._returning))
def __sql__(self, ctx):
if ctx.scope == SCOPE_COLUMN:
return self.apply_column(ctx)
if self._lateral and ctx.scope == SCOPE_SOURCE:
ctx.literal('LATERAL ')
is_subquery = ctx.subquery
state = {
'converter': None,
'in_function': False,
'parentheses': is_subquery or (ctx.scope == SCOPE_SOURCE),
'subquery': True,
}
if ctx.state.in_function and ctx.state.function_arg_count == 1:
state['parentheses'] = False
with ctx.scope_normal(**state):
# Defer calling parent SQL until here. This ensures that any CTEs
# for this query will be properly nested if this query is a
# sub-select or is used in an expression. See GH#1809 for example.
super(Select, self).__sql__(ctx)
ctx.literal('SELECT ')
if self._simple_distinct or self._distinct is not None:
ctx.literal('DISTINCT ')
if self._distinct:
(ctx
.literal('ON ')
.sql(EnclosedNodeList(self._distinct))
.literal(' '))
with ctx.scope_source():
ctx = self.__sql_selection__(ctx, is_subquery)
if self._from_list:
with ctx.scope_source(parentheses=False):
ctx.literal(' FROM ').sql(CommaNodeList(self._from_list))
if self._where is not None:
ctx.literal(' WHERE ').sql(self._where)
if self._group_by:
ctx.literal(' GROUP BY ').sql(CommaNodeList(self._group_by))
if self._having is not None:
ctx.literal(' HAVING ').sql(self._having)
if self._windows is not None:
ctx.literal(' WINDOW ')
ctx.sql(CommaNodeList(self._windows))
# Apply ORDER BY, LIMIT, OFFSET.
self._apply_ordering(ctx)
if self._for_update:
if not ctx.state.for_update:
raise ValueError('FOR UPDATE specified but not supported '
'by database.')
ctx.literal(' ')
ctx.sql(ForUpdate(self._for_update, self._for_update_of,
self._for_update_nowait))
# If the subquery is inside a function -or- we are evaluating a
# subquery on either side of an expression w/o an explicit alias, do
# not generate an alias + AS clause.
if ctx.state.in_function or (ctx.state.in_expr and
self._alias is None):
return ctx
return self.apply_alias(ctx)
class _WriteQuery(Query):
def __init__(self, table, returning=None, **kwargs):
self.table = table
self._returning = returning
self._return_cursor = True if returning else False
super(_WriteQuery, self).__init__(**kwargs)
@Node.copy
def returning(self, *returning):
self._returning = returning
self._return_cursor = True if returning else False
def apply_returning(self, ctx):
if self._returning:
with ctx.scope_source():
ctx.literal(' RETURNING ').sql(CommaNodeList(self._returning))
return ctx
def _execute(self, database):
if self._returning:
cursor = self.execute_returning(database)
else:
cursor = database.execute(self)
return self.handle_result(database, cursor)
def execute_returning(self, database):
if self._cursor_wrapper is None:
cursor = database.execute(self)
self._cursor_wrapper = self._get_cursor_wrapper(cursor)
return self._cursor_wrapper
def handle_result(self, database, cursor):
if self._return_cursor:
return cursor
return database.rows_affected(cursor)
def _set_table_alias(self, ctx):
ctx.alias_manager[self.table] = self.table.__name__
def __sql__(self, ctx):
super(_WriteQuery, self).__sql__(ctx)
# We explicitly set the table alias to the table's name, which ensures
# that if a sub-select references a column on the outer table, we won't
# assign it a new alias (e.g. t2) but will refer to it as table.column.
self._set_table_alias(ctx)
return ctx
class Update(_WriteQuery):
def __init__(self, table, update=None, **kwargs):
super(Update, self).__init__(table, **kwargs)
self._update = update
self._from = None
@Node.copy
def from_(self, *sources):
self._from = sources
def __sql__(self, ctx):
super(Update, self).__sql__(ctx)
with ctx.scope_values(subquery=True):
ctx.literal('UPDATE ')
expressions = []
for k, v in sorted(self._update.items(), key=ctx.column_sort_key):
if not isinstance(v, Node):
if isinstance(k, Field):
v = k.to_value(v)
else:
v = Value(v, unpack=False)
elif isinstance(v, Model) and isinstance(k, ForeignKeyField):
# NB: we want to ensure that when passed a model instance
# in the context of a foreign-key, we apply the fk-specific
# adaptation of the model.
v = k.to_value(v)
if not isinstance(v, Value):
v = qualify_names(v)
expressions.append(NodeList((k, SQL('='), v)))
(ctx
.sql(self.table)
.literal(' SET ')
.sql(CommaNodeList(expressions)))
if self._from:
with ctx.scope_source(parentheses=False):
ctx.literal(' FROM ').sql(CommaNodeList(self._from))
if self._where:
with ctx.scope_normal():
ctx.literal(' WHERE ').sql(self._where)
self._apply_ordering(ctx)
return self.apply_returning(ctx)
class Insert(_WriteQuery):
SIMPLE = 0
QUERY = 1
MULTI = 2
class DefaultValuesException(Exception): pass
def __init__(self, table, insert=None, columns=None, on_conflict=None,
**kwargs):
super(Insert, self).__init__(table, **kwargs)
self._insert = insert
self._columns = columns
self._on_conflict = on_conflict
self._query_type = None
def where(self, *expressions):
raise NotImplementedError('INSERT queries cannot have a WHERE clause.')
@Node.copy
def on_conflict_ignore(self, ignore=True):
self._on_conflict = OnConflict('IGNORE') if ignore else None
@Node.copy
def on_conflict_replace(self, replace=True):
self._on_conflict = OnConflict('REPLACE') if replace else None
@Node.copy
def on_conflict(self, *args, **kwargs):
self._on_conflict = (OnConflict(*args, **kwargs) if (args or kwargs)
else None)
def _simple_insert(self, ctx):
if not self._insert:
raise self.DefaultValuesException('Error: no data to insert.')
return self._generate_insert((self._insert,), ctx)
def get_default_data(self):
return {}
def get_default_columns(self):
if self.table._columns:
return [getattr(self.table, col) for col in self.table._columns
if col != self.table._primary_key]
def _generate_insert(self, insert, ctx):
rows_iter = iter(insert)
columns = self._columns
# Load and organize column defaults (if provided).
defaults = self.get_default_data()
# First figure out what columns are being inserted (if they weren't
# specified explicitly). Resulting columns are normalized and ordered.
if not columns:
try:
row = next(rows_iter)
except StopIteration:
raise self.DefaultValuesException('Error: no rows to insert.')
if not isinstance(row, Mapping):
columns = self.get_default_columns()
if columns is None:
raise ValueError('Bulk insert must specify columns.')
else:
# Infer column names from the dict of data being inserted.
accum = []
for column in row:
if isinstance(column, basestring):
column = getattr(self.table, column)
accum.append(column)
# Add any columns present in the default data that are not
# accounted for by the dictionary of row data.
column_set = set(accum)
for col in (set(defaults) - column_set):
accum.append(col)
columns = sorted(accum, key=lambda obj: obj.get_sort_key(ctx))
rows_iter = itertools.chain(iter((row,)), rows_iter)
else:
clean_columns = []
seen = set()
for column in columns:
if isinstance(column, basestring):
column_obj = getattr(self.table, column)
else:
column_obj = column
clean_columns.append(column_obj)
seen.add(column_obj)
columns = clean_columns
for col in sorted(defaults, key=lambda obj: obj.get_sort_key(ctx)):
if col not in seen:
columns.append(col)
fk_fields = set()
nullable_columns = set()
value_lookups = {}
for column in columns:
lookups = [column, column.name]
if isinstance(column, Field):
if column.name != column.column_name:
lookups.append(column.column_name)
if column.null:
nullable_columns.add(column)
if isinstance(column, ForeignKeyField):
fk_fields.add(column)
value_lookups[column] = lookups
ctx.sql(EnclosedNodeList(columns)).literal(' VALUES ')
columns_converters = [
(column, column.db_value if isinstance(column, Field) else None)
for column in columns]
all_values = []
for row in rows_iter:
values = []
is_dict = isinstance(row, Mapping)
for i, (column, converter) in enumerate(columns_converters):
try:
if is_dict:
# The logic is a bit convoluted, but in order to be
# flexible in what we accept (dict keyed by
# column/field, field name, or underlying column name),
# we try accessing the row data dict using each
# possible key. If no match is found, throw an error.
for lookup in value_lookups[column]:
try:
val = row[lookup]
except KeyError: pass
else: break
else:
raise KeyError
else:
val = row[i]
except (KeyError, IndexError):
if column in defaults:
val = defaults[column]
if callable_(val):
val = val()
elif column in nullable_columns:
val = None
else:
raise ValueError('Missing value for %s.' % column.name)
if not isinstance(val, Node) or (isinstance(val, Model) and
column in fk_fields):
val = Value(val, converter=converter, unpack=False)
values.append(val)
all_values.append(EnclosedNodeList(values))
if not all_values:
raise self.DefaultValuesException('Error: no data to insert.')
with ctx.scope_values(subquery=True):
return ctx.sql(CommaNodeList(all_values))
def _query_insert(self, ctx):
return (ctx
.sql(EnclosedNodeList(self._columns))
.literal(' ')
.sql(self._insert))
def _default_values(self, ctx):
if not self._database:
return ctx.literal('DEFAULT VALUES')
return self._database.default_values_insert(ctx)
def __sql__(self, ctx):
super(Insert, self).__sql__(ctx)
with ctx.scope_values():
stmt = None
if self._on_conflict is not None:
stmt = self._on_conflict.get_conflict_statement(ctx, self)
(ctx
.sql(stmt or SQL('INSERT'))
.literal(' INTO ')
.sql(self.table)
.literal(' '))
if isinstance(self._insert, Mapping) and not self._columns:
try:
self._simple_insert(ctx)
except self.DefaultValuesException:
self._default_values(ctx)
self._query_type = Insert.SIMPLE
elif isinstance(self._insert, (SelectQuery, SQL)):
self._query_insert(ctx)
self._query_type = Insert.QUERY
else:
self._generate_insert(self._insert, ctx)
self._query_type = Insert.MULTI
if self._on_conflict is not None:
update = self._on_conflict.get_conflict_update(ctx, self)
if update is not None:
ctx.literal(' ').sql(update)
return self.apply_returning(ctx)
def _execute(self, database):
if self._returning is None and database.returning_clause \
and self.table._primary_key:
self._returning = (self.table._primary_key,)
try:
return super(Insert, self)._execute(database)
except self.DefaultValuesException:
pass
def handle_result(self, database, cursor):
if self._return_cursor:
return cursor
if self._query_type != Insert.SIMPLE and not self._returning:
return database.rows_affected(cursor)
return database.last_insert_id(cursor, self._query_type)
class Delete(_WriteQuery):
def __sql__(self, ctx):
super(Delete, self).__sql__(ctx)
with ctx.scope_values(subquery=True):
ctx.literal('DELETE FROM ').sql(self.table)
if self._where is not None:
with ctx.scope_normal():
ctx.literal(' WHERE ').sql(self._where)
self._apply_ordering(ctx)
return self.apply_returning(ctx)
class Index(Node):
def __init__(self, name, table, expressions, unique=False, safe=False,
where=None, using=None):
self._name = name
self._table = Entity(table) if not isinstance(table, Table) else table
self._expressions = expressions
self._where = where
self._unique = unique
self._safe = safe
self._using = using
@Node.copy
def safe(self, _safe=True):
self._safe = _safe
@Node.copy
def where(self, *expressions):
if self._where is not None:
expressions = (self._where,) + expressions
self._where = reduce(operator.and_, expressions)
@Node.copy
def using(self, _using=None):
self._using = _using
def __sql__(self, ctx):
statement = 'CREATE UNIQUE INDEX ' if self._unique else 'CREATE INDEX '
with ctx.scope_values(subquery=True):
ctx.literal(statement)
if self._safe:
ctx.literal('IF NOT EXISTS ')
# Sqlite uses CREATE INDEX <schema>.<name> ON <table>, whereas most
# others use: CREATE INDEX <name> ON <schema>.<table>.
if ctx.state.index_schema_prefix and \
isinstance(self._table, Table) and self._table._schema:
index_name = Entity(self._table._schema, self._name)
table_name = Entity(self._table.__name__)
else:
index_name = Entity(self._name)
table_name = self._table
ctx.sql(index_name)
if self._using is not None and \
ctx.state.index_using_precedes_table:
ctx.literal(' USING %s' % self._using) # MySQL style.
(ctx
.literal(' ON ')
.sql(table_name)
.literal(' '))
if self._using is not None and not \
ctx.state.index_using_precedes_table:
ctx.literal('USING %s ' % self._using) # Postgres/default.
ctx.sql(EnclosedNodeList([
SQL(expr) if isinstance(expr, basestring) else expr
for expr in self._expressions]))
if self._where is not None:
ctx.literal(' WHERE ').sql(self._where)
return ctx
class ModelIndex(Index):
def __init__(self, model, fields, unique=False, safe=True, where=None,
using=None, name=None):
self._model = model
if name is None:
name = self._generate_name_from_fields(model, fields)
if using is None:
for field in fields:
if isinstance(field, Field) and hasattr(field, 'index_type'):
using = field.index_type
super(ModelIndex, self).__init__(
name=name,
table=model._meta.table,
expressions=fields,
unique=unique,
safe=safe,
where=where,
using=using)
def _generate_name_from_fields(self, model, fields):
accum = []
for field in fields:
if isinstance(field, basestring):
accum.append(field.split()[0])
else:
if isinstance(field, Node) and not isinstance(field, Field):
field = field.unwrap()
if isinstance(field, Field):
accum.append(field.column_name)
if not accum:
raise ValueError('Unable to generate a name for the index, please '
'explicitly specify a name.')
clean_field_names = re.sub(r'[^\w]+', '', '_'.join(accum))
meta = model._meta
prefix = meta.name if meta.legacy_table_names else meta.table_name
return _truncate_constraint_name('_'.join((prefix, clean_field_names)))
def _truncate_constraint_name(constraint, maxlen=64):
if len(constraint) > maxlen:
name_hash = hashlib.md5(constraint.encode('utf-8')).hexdigest()
constraint = '%s_%s' % (constraint[:(maxlen - 8)], name_hash[:7])
return constraint
# DB-API 2.0 EXCEPTIONS.
class PeeweeException(Exception):
def __init__(self, *args):
if args and isinstance(args[0], Exception):
self.orig, args = args[0], args[1:]
super(PeeweeException, self).__init__(*args)
class ImproperlyConfigured(PeeweeException): pass
class DatabaseError(PeeweeException): pass
class DataError(DatabaseError): pass
class IntegrityError(DatabaseError): pass
class InterfaceError(PeeweeException): pass
class InternalError(DatabaseError): pass
class NotSupportedError(DatabaseError): pass
class OperationalError(DatabaseError): pass
class ProgrammingError(DatabaseError): pass
class ExceptionWrapper(object):
__slots__ = ('exceptions',)
def __init__(self, exceptions):
self.exceptions = exceptions
def __enter__(self): pass
def __exit__(self, exc_type, exc_value, traceback):
if exc_type is None:
return
# psycopg2.8 shits out a million cute error types. Try to catch em all.
if pg_errors is not None and exc_type.__name__ not in self.exceptions \
and issubclass(exc_type, pg_errors.Error):
exc_type = exc_type.__bases__[0]
if exc_type.__name__ in self.exceptions:
new_type = self.exceptions[exc_type.__name__]
exc_args = exc_value.args
reraise(new_type, new_type(exc_value, *exc_args), traceback)
EXCEPTIONS = {
'ConstraintError': IntegrityError,
'DatabaseError': DatabaseError,
'DataError': DataError,
'IntegrityError': IntegrityError,
'InterfaceError': InterfaceError,
'InternalError': InternalError,
'NotSupportedError': NotSupportedError,
'OperationalError': OperationalError,
'ProgrammingError': ProgrammingError,
'TransactionRollbackError': OperationalError}
__exception_wrapper__ = ExceptionWrapper(EXCEPTIONS)
# DATABASE INTERFACE AND CONNECTION MANAGEMENT.
IndexMetadata = collections.namedtuple(
'IndexMetadata',
('name', 'sql', 'columns', 'unique', 'table'))
ColumnMetadata = collections.namedtuple(
'ColumnMetadata',
('name', 'data_type', 'null', 'primary_key', 'table', 'default'))
ForeignKeyMetadata = collections.namedtuple(
'ForeignKeyMetadata',
('column', 'dest_table', 'dest_column', 'table'))
ViewMetadata = collections.namedtuple('ViewMetadata', ('name', 'sql'))
class _ConnectionState(object):
def __init__(self, **kwargs):
super(_ConnectionState, self).__init__(**kwargs)
self.reset()
def reset(self):
self.closed = True
self.conn = None
self.ctx = []
self.transactions = []
def set_connection(self, conn):
self.conn = conn
self.closed = False
self.ctx = []
self.transactions = []
class _ConnectionLocal(_ConnectionState, threading.local): pass
class _NoopLock(object):
__slots__ = ()
def __enter__(self): return self
def __exit__(self, exc_type, exc_val, exc_tb): pass
class ConnectionContext(_callable_context_manager):
__slots__ = ('db',)
def __init__(self, db): self.db = db
def __enter__(self):
if self.db.is_closed():
self.db.connect()
def __exit__(self, exc_type, exc_val, exc_tb): self.db.close()
class Database(_callable_context_manager):
context_class = Context
field_types = {}
operations = {}
param = '?'
quote = '""'
server_version = None
# Feature toggles.
commit_select = False
compound_select_parentheses = CSQ_PARENTHESES_NEVER
for_update = False
index_schema_prefix = False
index_using_precedes_table = False
limit_max = None
nulls_ordering = False
returning_clause = False
safe_create_index = True
safe_drop_index = True
sequences = False
truncate_table = True
def __init__(self, database, thread_safe=True, autorollback=False,
field_types=None, operations=None, autocommit=None,
autoconnect=True, **kwargs):
self._field_types = merge_dict(FIELD, self.field_types)
self._operations = merge_dict(OP, self.operations)
if field_types:
self._field_types.update(field_types)
if operations:
self._operations.update(operations)
self.autoconnect = autoconnect
self.autorollback = autorollback
self.thread_safe = thread_safe
if thread_safe:
self._state = _ConnectionLocal()
self._lock = threading.RLock()
else:
self._state = _ConnectionState()
self._lock = _NoopLock()
if autocommit is not None:
__deprecated__('Peewee no longer uses the "autocommit" option, as '
'the semantics now require it to always be True. '
'Because some database-drivers also use the '
'"autocommit" parameter, you are receiving a '
'warning so you may update your code and remove '
'the parameter, as in the future, specifying '
'autocommit could impact the behavior of the '
'database driver you are using.')
self.connect_params = {}
self.init(database, **kwargs)
def init(self, database, **kwargs):
if not self.is_closed():
self.close()
self.database = database
self.connect_params.update(kwargs)
self.deferred = not bool(database)
def __enter__(self):
if self.is_closed():
self.connect()
ctx = self.atomic()
self._state.ctx.append(ctx)
ctx.__enter__()
return self
def __exit__(self, exc_type, exc_val, exc_tb):
ctx = self._state.ctx.pop()
try:
ctx.__exit__(exc_type, exc_val, exc_tb)
finally:
if not self._state.ctx:
self.close()
def connection_context(self):
return ConnectionContext(self)
def _connect(self):
raise NotImplementedError
def connect(self, reuse_if_open=False):
with self._lock:
if self.deferred:
raise InterfaceError('Error, database must be initialized '
'before opening a connection.')
if not self._state.closed:
if reuse_if_open:
return False
raise OperationalError('Connection already opened.')
self._state.reset()
with __exception_wrapper__:
self._state.set_connection(self._connect())
if self.server_version is None:
self._set_server_version(self._state.conn)
self._initialize_connection(self._state.conn)
return True
def _initialize_connection(self, conn):
pass
def _set_server_version(self, conn):
self.server_version = 0
def close(self):
with self._lock:
if self.deferred:
raise InterfaceError('Error, database must be initialized '
'before opening a connection.')
if self.in_transaction():
raise OperationalError('Attempting to close database while '
'transaction is open.')
is_open = not self._state.closed
try:
if is_open:
with __exception_wrapper__:
self._close(self._state.conn)
finally:
self._state.reset()
return is_open
def _close(self, conn):
conn.close()
def is_closed(self):
return self._state.closed
def is_connection_usable(self):
return not self._state.closed
def connection(self):
if self.is_closed():
self.connect()
return self._state.conn
def cursor(self, commit=None):
if self.is_closed():
if self.autoconnect:
self.connect()
else:
raise InterfaceError('Error, database connection not opened.')
return self._state.conn.cursor()
def execute_sql(self, sql, params=None, commit=SENTINEL):
logger.debug((sql, params))
if commit is SENTINEL:
if self.in_transaction():
commit = False
elif self.commit_select:
commit = True
else:
commit = not sql[:6].lower().startswith('select')
with __exception_wrapper__:
cursor = self.cursor(commit)
try:
cursor.execute(sql, params or ())
except Exception:
if self.autorollback and not self.in_transaction():
self.rollback()
raise
else:
if commit and not self.in_transaction():
self.commit()
return cursor
def execute(self, query, commit=SENTINEL, **context_options):
ctx = self.get_sql_context(**context_options)
sql, params = ctx.sql(query).query()
return self.execute_sql(sql, params, commit=commit)
def get_context_options(self):
return {
'field_types': self._field_types,
'operations': self._operations,
'param': self.param,
'quote': self.quote,
'compound_select_parentheses': self.compound_select_parentheses,
'conflict_statement': self.conflict_statement,
'conflict_update': self.conflict_update,
'for_update': self.for_update,
'index_schema_prefix': self.index_schema_prefix,
'index_using_precedes_table': self.index_using_precedes_table,
'limit_max': self.limit_max,
'nulls_ordering': self.nulls_ordering,
}
def get_sql_context(self, **context_options):
context = self.get_context_options()
if context_options:
context.update(context_options)
return self.context_class(**context)
def conflict_statement(self, on_conflict, query):
raise NotImplementedError
def conflict_update(self, on_conflict, query):
raise NotImplementedError
def _build_on_conflict_update(self, on_conflict, query):
if on_conflict._conflict_target:
stmt = SQL('ON CONFLICT')
target = EnclosedNodeList([
Entity(col) if isinstance(col, basestring) else col
for col in on_conflict._conflict_target])
if on_conflict._conflict_where is not None:
target = NodeList([target, SQL('WHERE'),
on_conflict._conflict_where])
else:
stmt = SQL('ON CONFLICT ON CONSTRAINT')
target = on_conflict._conflict_constraint
if isinstance(target, basestring):
target = Entity(target)
updates = []
if on_conflict._preserve:
for column in on_conflict._preserve:
excluded = NodeList((SQL('EXCLUDED'), ensure_entity(column)),
glue='.')
expression = NodeList((ensure_entity(column), SQL('='),
excluded))
updates.append(expression)
if on_conflict._update:
for k, v in on_conflict._update.items():
if not isinstance(v, Node):
# Attempt to resolve string field-names to their respective
# field object, to apply data-type conversions.
if isinstance(k, basestring):
k = getattr(query.table, k)
if isinstance(k, Field):
v = k.to_value(v)
else:
v = Value(v, unpack=False)
else:
v = QualifiedNames(v)
updates.append(NodeList((ensure_entity(k), SQL('='), v)))
parts = [stmt, target, SQL('DO UPDATE SET'), CommaNodeList(updates)]
if on_conflict._where:
parts.extend((SQL('WHERE'), QualifiedNames(on_conflict._where)))
return NodeList(parts)
def last_insert_id(self, cursor, query_type=None):
return cursor.lastrowid
def rows_affected(self, cursor):
return cursor.rowcount
def default_values_insert(self, ctx):
return ctx.literal('DEFAULT VALUES')
def session_start(self):
with self._lock:
return self.transaction().__enter__()
def session_commit(self):
with self._lock:
try:
txn = self.pop_transaction()
except IndexError:
return False
txn.commit(begin=self.in_transaction())
return True
def session_rollback(self):
with self._lock:
try:
txn = self.pop_transaction()
except IndexError:
return False
txn.rollback(begin=self.in_transaction())
return True
def in_transaction(self):
return bool(self._state.transactions)
def push_transaction(self, transaction):
self._state.transactions.append(transaction)
def pop_transaction(self):
return self._state.transactions.pop()
def transaction_depth(self):
return len(self._state.transactions)
def top_transaction(self):
if self._state.transactions:
return self._state.transactions[-1]
def atomic(self, *args, **kwargs):
return _atomic(self, *args, **kwargs)
def manual_commit(self):
return _manual(self)
def transaction(self, *args, **kwargs):
return _transaction(self, *args, **kwargs)
def savepoint(self):
return _savepoint(self)
def begin(self):
if self.is_closed():
self.connect()
def commit(self):
with __exception_wrapper__:
return self._state.conn.commit()
def rollback(self):
with __exception_wrapper__:
return self._state.conn.rollback()
def batch_commit(self, it, n):
for group in chunked(it, n):
with self.atomic():
for obj in group:
yield obj
def table_exists(self, table_name, schema=None):
return table_name in self.get_tables(schema=schema)
def get_tables(self, schema=None):
raise NotImplementedError
def get_indexes(self, table, schema=None):
raise NotImplementedError
def get_columns(self, table, schema=None):
raise NotImplementedError
def get_primary_keys(self, table, schema=None):
raise NotImplementedError
def get_foreign_keys(self, table, schema=None):
raise NotImplementedError
def sequence_exists(self, seq):
raise NotImplementedError
def create_tables(self, models, **options):
for model in sort_models(models):
model.create_table(**options)
def drop_tables(self, models, **kwargs):
for model in reversed(sort_models(models)):
model.drop_table(**kwargs)
def extract_date(self, date_part, date_field):
raise NotImplementedError
def truncate_date(self, date_part, date_field):
raise NotImplementedError
def to_timestamp(self, date_field):
raise NotImplementedError
def from_timestamp(self, date_field):
raise NotImplementedError
def random(self):
return fn.random()
def bind(self, models, bind_refs=True, bind_backrefs=True):
for model in models:
model.bind(self, bind_refs=bind_refs, bind_backrefs=bind_backrefs)
def bind_ctx(self, models, bind_refs=True, bind_backrefs=True):
return _BoundModelsContext(models, self, bind_refs, bind_backrefs)
def get_noop_select(self, ctx):
return ctx.sql(Select().columns(SQL('0')).where(SQL('0')))
def __pragma__(name):
def __get__(self):
return self.pragma(name)
def __set__(self, value):
return self.pragma(name, value)
return property(__get__, __set__)
class SqliteDatabase(Database):
field_types = {
'BIGAUTO': FIELD.AUTO,
'BIGINT': FIELD.INT,
'BOOL': FIELD.INT,
'DOUBLE': FIELD.FLOAT,
'SMALLINT': FIELD.INT,
'UUID': FIELD.TEXT}
operations = {
'LIKE': 'GLOB',
'ILIKE': 'LIKE'}
index_schema_prefix = True
limit_max = -1
server_version = __sqlite_version__
truncate_table = False
def __init__(self, database, *args, **kwargs):
self._pragmas = kwargs.pop('pragmas', ())
super(SqliteDatabase, self).__init__(database, *args, **kwargs)
self._aggregates = {}
self._collations = {}
self._functions = {}
self._window_functions = {}
self._table_functions = []
self._extensions = set()
self._attached = {}
self.register_function(_sqlite_date_part, 'date_part', 2)
self.register_function(_sqlite_date_trunc, 'date_trunc', 2)
self.nulls_ordering = self.server_version >= (3, 30, 0)
def init(self, database, pragmas=None, timeout=5, **kwargs):
if pragmas is not None:
self._pragmas = pragmas
if isinstance(self._pragmas, dict):
self._pragmas = list(self._pragmas.items())
self._timeout = timeout
super(SqliteDatabase, self).init(database, **kwargs)
def _set_server_version(self, conn):
pass
def _connect(self):
if sqlite3 is None:
raise ImproperlyConfigured('SQLite driver not installed!')
conn = sqlite3.connect(self.database, timeout=self._timeout,
isolation_level=None, **self.connect_params)
try:
self._add_conn_hooks(conn)
except:
conn.close()
raise
return conn
def _add_conn_hooks(self, conn):
if self._attached:
self._attach_databases(conn)
if self._pragmas:
self._set_pragmas(conn)
self._load_aggregates(conn)
self._load_collations(conn)
self._load_functions(conn)
if self.server_version >= (3, 25, 0):
self._load_window_functions(conn)
if self._table_functions:
for table_function in self._table_functions:
table_function.register(conn)
if self._extensions:
self._load_extensions(conn)
def _set_pragmas(self, conn):
cursor = conn.cursor()
for pragma, value in self._pragmas:
cursor.execute('PRAGMA %s = %s;' % (pragma, value))
cursor.close()
def _attach_databases(self, conn):
cursor = conn.cursor()
for name, db in self._attached.items():
cursor.execute('ATTACH DATABASE "%s" AS "%s"' % (db, name))
cursor.close()
def pragma(self, key, value=SENTINEL, permanent=False, schema=None):
if schema is not None:
key = '"%s".%s' % (schema, key)
sql = 'PRAGMA %s' % key
if value is not SENTINEL:
sql += ' = %s' % (value or 0)
if permanent:
pragmas = dict(self._pragmas or ())
pragmas[key] = value
self._pragmas = list(pragmas.items())
elif permanent:
raise ValueError('Cannot specify a permanent pragma without value')
row = self.execute_sql(sql).fetchone()
if row:
return row[0]
cache_size = __pragma__('cache_size')
foreign_keys = __pragma__('foreign_keys')
journal_mode = __pragma__('journal_mode')
journal_size_limit = __pragma__('journal_size_limit')
mmap_size = __pragma__('mmap_size')
page_size = __pragma__('page_size')
read_uncommitted = __pragma__('read_uncommitted')
synchronous = __pragma__('synchronous')
wal_autocheckpoint = __pragma__('wal_autocheckpoint')
@property
def timeout(self):
return self._timeout
@timeout.setter
def timeout(self, seconds):
if self._timeout == seconds:
return
self._timeout = seconds
if not self.is_closed():
# PySQLite multiplies user timeout by 1000, but the unit of the
# timeout PRAGMA is actually milliseconds.
self.execute_sql('PRAGMA busy_timeout=%d;' % (seconds * 1000))
def _load_aggregates(self, conn):
for name, (klass, num_params) in self._aggregates.items():
conn.create_aggregate(name, num_params, klass)
def _load_collations(self, conn):
for name, fn in self._collations.items():
conn.create_collation(name, fn)
def _load_functions(self, conn):
for name, (fn, num_params) in self._functions.items():
conn.create_function(name, num_params, fn)
def _load_window_functions(self, conn):
for name, (klass, num_params) in self._window_functions.items():
conn.create_window_function(name, num_params, klass)
def register_aggregate(self, klass, name=None, num_params=-1):
self._aggregates[name or klass.__name__.lower()] = (klass, num_params)
if not self.is_closed():
self._load_aggregates(self.connection())
def aggregate(self, name=None, num_params=-1):
def decorator(klass):
self.register_aggregate(klass, name, num_params)
return klass
return decorator
def register_collation(self, fn, name=None):
name = name or fn.__name__
def _collation(*args):
expressions = args + (SQL('collate %s' % name),)
return NodeList(expressions)
fn.collation = _collation
self._collations[name] = fn
if not self.is_closed():
self._load_collations(self.connection())
def collation(self, name=None):
def decorator(fn):
self.register_collation(fn, name)
return fn
return decorator
def register_function(self, fn, name=None, num_params=-1):
self._functions[name or fn.__name__] = (fn, num_params)
if not self.is_closed():
self._load_functions(self.connection())
def func(self, name=None, num_params=-1):
def decorator(fn):
self.register_function(fn, name, num_params)
return fn
return decorator
def register_window_function(self, klass, name=None, num_params=-1):
name = name or klass.__name__.lower()
self._window_functions[name] = (klass, num_params)
if not self.is_closed():
self._load_window_functions(self.connection())
def window_function(self, name=None, num_params=-1):
def decorator(klass):
self.register_window_function(klass, name, num_params)
return klass
return decorator
def register_table_function(self, klass, name=None):
if name is not None:
klass.name = name
self._table_functions.append(klass)
if not self.is_closed():
klass.register(self.connection())
def table_function(self, name=None):
def decorator(klass):
self.register_table_function(klass, name)
return klass
return decorator
def unregister_aggregate(self, name):
del(self._aggregates[name])
def unregister_collation(self, name):
del(self._collations[name])
def unregister_function(self, name):
del(self._functions[name])
def unregister_window_function(self, name):
del(self._window_functions[name])
def unregister_table_function(self, name):
for idx, klass in enumerate(self._table_functions):
if klass.name == name:
break
else:
return False
self._table_functions.pop(idx)
return True
def _load_extensions(self, conn):
conn.enable_load_extension(True)
for extension in self._extensions:
conn.load_extension(extension)
def load_extension(self, extension):
self._extensions.add(extension)
if not self.is_closed():
conn = self.connection()
conn.enable_load_extension(True)
conn.load_extension(extension)
def unload_extension(self, extension):
self._extensions.remove(extension)
def attach(self, filename, name):
if name in self._attached:
if self._attached[name] == filename:
return False
raise OperationalError('schema "%s" already attached.' % name)
self._attached[name] = filename
if not self.is_closed():
self.execute_sql('ATTACH DATABASE "%s" AS "%s"' % (filename, name))
return True
def detach(self, name):
if name not in self._attached:
return False
del self._attached[name]
if not self.is_closed():
self.execute_sql('DETACH DATABASE "%s"' % name)
return True
def begin(self, lock_type=None):
statement = 'BEGIN %s' % lock_type if lock_type else 'BEGIN'
self.execute_sql(statement, commit=False)
def get_tables(self, schema=None):
schema = schema or 'main'
cursor = self.execute_sql('SELECT name FROM "%s".sqlite_master WHERE '
'type=? ORDER BY name' % schema, ('table',))
return [row for row, in cursor.fetchall()]
def get_views(self, schema=None):
sql = ('SELECT name, sql FROM "%s".sqlite_master WHERE type=? '
'ORDER BY name') % (schema or 'main')
return [ViewMetadata(*row) for row in self.execute_sql(sql, ('view',))]
def get_indexes(self, table, schema=None):
schema = schema or 'main'
query = ('SELECT name, sql FROM "%s".sqlite_master '
'WHERE tbl_name = ? AND type = ? ORDER BY name') % schema
cursor = self.execute_sql(query, (table, 'index'))
index_to_sql = dict(cursor.fetchall())
# Determine which indexes have a unique constraint.
unique_indexes = set()
cursor = self.execute_sql('PRAGMA "%s".index_list("%s")' %
(schema, table))
for row in cursor.fetchall():
name = row[1]
is_unique = int(row[2]) == 1
if is_unique:
unique_indexes.add(name)
# Retrieve the indexed columns.
index_columns = {}
for index_name in sorted(index_to_sql):
cursor = self.execute_sql('PRAGMA "%s".index_info("%s")' %
(schema, index_name))
index_columns[index_name] = [row[2] for row in cursor.fetchall()]
return [
IndexMetadata(
name,
index_to_sql[name],
index_columns[name],
name in unique_indexes,
table)
for name in sorted(index_to_sql)]
def get_columns(self, table, schema=None):
cursor = self.execute_sql('PRAGMA "%s".table_info("%s")' %
(schema or 'main', table))
return [ColumnMetadata(r[1], r[2], not r[3], bool(r[5]), table, r[4])
for r in cursor.fetchall()]
def get_primary_keys(self, table, schema=None):
cursor = self.execute_sql('PRAGMA "%s".table_info("%s")' %
(schema or 'main', table))
return [row[1] for row in filter(lambda r: r[-1], cursor.fetchall())]
def get_foreign_keys(self, table, schema=None):
cursor = self.execute_sql('PRAGMA "%s".foreign_key_list("%s")' %
(schema or 'main', table))
return [ForeignKeyMetadata(row[3], row[2], row[4], table)
for row in cursor.fetchall()]
def get_binary_type(self):
return sqlite3.Binary
def conflict_statement(self, on_conflict, query):
action = on_conflict._action.lower() if on_conflict._action else ''
if action and action not in ('nothing', 'update'):
return SQL('INSERT OR %s' % on_conflict._action.upper())
def conflict_update(self, oc, query):
# Sqlite prior to 3.24.0 does not support Postgres-style upsert.
if self.server_version < (3, 24, 0) and \
any((oc._preserve, oc._update, oc._where, oc._conflict_target,
oc._conflict_constraint)):
raise ValueError('SQLite does not support specifying which values '
'to preserve or update.')
action = oc._action.lower() if oc._action else ''
if action and action not in ('nothing', 'update', ''):
return
if action == 'nothing':
return SQL('ON CONFLICT DO NOTHING')
elif not oc._update and not oc._preserve:
raise ValueError('If you are not performing any updates (or '
'preserving any INSERTed values), then the '
'conflict resolution action should be set to '
'"NOTHING".')
elif oc._conflict_constraint:
raise ValueError('SQLite does not support specifying named '
'constraints for conflict resolution.')
elif not oc._conflict_target:
raise ValueError('SQLite requires that a conflict target be '
'specified when doing an upsert.')
return self._build_on_conflict_update(oc, query)
def extract_date(self, date_part, date_field):
return fn.date_part(date_part, date_field, python_value=int)
def truncate_date(self, date_part, date_field):
return fn.date_trunc(date_part, date_field,
python_value=simple_date_time)
def to_timestamp(self, date_field):
return fn.strftime('%s', date_field).cast('integer')
def from_timestamp(self, date_field):
return fn.datetime(date_field, 'unixepoch')
class PostgresqlDatabase(Database):
field_types = {
'AUTO': 'SERIAL',
'BIGAUTO': 'BIGSERIAL',
'BLOB': 'BYTEA',
'BOOL': 'BOOLEAN',
'DATETIME': 'TIMESTAMP',
'DECIMAL': 'NUMERIC',
'DOUBLE': 'DOUBLE PRECISION',
'UUID': 'UUID',
'UUIDB': 'BYTEA'}
operations = {'REGEXP': '~', 'IREGEXP': '~*'}
param = '%s'
commit_select = True
compound_select_parentheses = CSQ_PARENTHESES_ALWAYS
for_update = True
nulls_ordering = True
returning_clause = True
safe_create_index = False
sequences = True
def init(self, database, register_unicode=True, encoding=None,
isolation_level=None, **kwargs):
self._register_unicode = register_unicode
self._encoding = encoding
self._isolation_level = isolation_level
super(PostgresqlDatabase, self).init(database, **kwargs)
def _connect(self):
if psycopg2 is None:
raise ImproperlyConfigured('Postgres driver not installed!')
# Handle connection-strings nicely, since psycopg2 will accept them,
# and they may be easier when lots of parameters are specified.
params = self.connect_params.copy()
if self.database.startswith('postgresql://'):
params.setdefault('dsn', self.database)
else:
params.setdefault('dbname', self.database)
conn = psycopg2.connect(**params)
if self._register_unicode:
pg_extensions.register_type(pg_extensions.UNICODE, conn)
pg_extensions.register_type(pg_extensions.UNICODEARRAY, conn)
if self._encoding:
conn.set_client_encoding(self._encoding)
if self._isolation_level:
conn.set_isolation_level(self._isolation_level)
return conn
def _set_server_version(self, conn):
self.server_version = conn.server_version
if self.server_version >= 90600:
self.safe_create_index = True
def is_connection_usable(self):
if self._state.closed:
return False
# Returns True if we are idle, running a command, or in an active
# connection. If the connection is in an error state or the connection
# is otherwise unusable, return False.
txn_status = self._state.conn.get_transaction_status()
return txn_status < pg_extensions.TRANSACTION_STATUS_INERROR
def last_insert_id(self, cursor, query_type=None):
try:
return cursor if query_type != Insert.SIMPLE else cursor[0][0]
except (IndexError, KeyError, TypeError):
pass
def get_tables(self, schema=None):
query = ('SELECT tablename FROM pg_catalog.pg_tables '
'WHERE schemaname = %s ORDER BY tablename')
cursor = self.execute_sql(query, (schema or 'public',))
return [table for table, in cursor.fetchall()]
def get_views(self, schema=None):
query = ('SELECT viewname, definition FROM pg_catalog.pg_views '
'WHERE schemaname = %s ORDER BY viewname')
cursor = self.execute_sql(query, (schema or 'public',))
return [ViewMetadata(view_name, sql.strip(' \t;'))
for (view_name, sql) in cursor.fetchall()]
def get_indexes(self, table, schema=None):
query = """
SELECT
i.relname, idxs.indexdef, idx.indisunique,
array_to_string(ARRAY(
SELECT pg_get_indexdef(idx.indexrelid, k + 1, TRUE)
FROM generate_subscripts(idx.indkey, 1) AS k
ORDER BY k), ',')
FROM pg_catalog.pg_class AS t
INNER JOIN pg_catalog.pg_index AS idx ON t.oid = idx.indrelid
INNER JOIN pg_catalog.pg_class AS i ON idx.indexrelid = i.oid
INNER JOIN pg_catalog.pg_indexes AS idxs ON
(idxs.tablename = t.relname AND idxs.indexname = i.relname)
WHERE t.relname = %s AND t.relkind = %s AND idxs.schemaname = %s
ORDER BY idx.indisunique DESC, i.relname;"""
cursor = self.execute_sql(query, (table, 'r', schema or 'public'))
return [IndexMetadata(name, sql.rstrip(' ;'), columns.split(','),
is_unique, table)
for name, sql, is_unique, columns in cursor.fetchall()]
def get_columns(self, table, schema=None):
query = """
SELECT column_name, is_nullable, data_type, column_default
FROM information_schema.columns
WHERE table_name = %s AND table_schema = %s
ORDER BY ordinal_position"""
cursor = self.execute_sql(query, (table, schema or 'public'))
pks = set(self.get_primary_keys(table, schema))
return [ColumnMetadata(name, dt, null == 'YES', name in pks, table, df)
for name, null, dt, df in cursor.fetchall()]
def get_primary_keys(self, table, schema=None):
query = """
SELECT kc.column_name
FROM information_schema.table_constraints AS tc
INNER JOIN information_schema.key_column_usage AS kc ON (
tc.table_name = kc.table_name AND
tc.table_schema = kc.table_schema AND
tc.constraint_name = kc.constraint_name)
WHERE
tc.constraint_type = %s AND
tc.table_name = %s AND
tc.table_schema = %s"""
ctype = 'PRIMARY KEY'
cursor = self.execute_sql(query, (ctype, table, schema or 'public'))
return [pk for pk, in cursor.fetchall()]
def get_foreign_keys(self, table, schema=None):
sql = """
SELECT DISTINCT
kcu.column_name, ccu.table_name, ccu.column_name
FROM information_schema.table_constraints AS tc
JOIN information_schema.key_column_usage AS kcu
ON (tc.constraint_name = kcu.constraint_name AND
tc.constraint_schema = kcu.constraint_schema AND
tc.table_name = kcu.table_name AND
tc.table_schema = kcu.table_schema)
JOIN information_schema.constraint_column_usage AS ccu
ON (ccu.constraint_name = tc.constraint_name AND
ccu.constraint_schema = tc.constraint_schema)
WHERE
tc.constraint_type = 'FOREIGN KEY' AND
tc.table_name = %s AND
tc.table_schema = %s"""
cursor = self.execute_sql(sql, (table, schema or 'public'))
return [ForeignKeyMetadata(row[0], row[1], row[2], table)
for row in cursor.fetchall()]
def sequence_exists(self, sequence):
res = self.execute_sql("""
SELECT COUNT(*) FROM pg_class, pg_namespace
WHERE relkind='S'
AND pg_class.relnamespace = pg_namespace.oid
AND relname=%s""", (sequence,))
return bool(res.fetchone()[0])
def get_binary_type(self):
return psycopg2.Binary
def conflict_statement(self, on_conflict, query):
return
def conflict_update(self, oc, query):
action = oc._action.lower() if oc._action else ''
if action in ('ignore', 'nothing'):
parts = [SQL('ON CONFLICT')]
if oc._conflict_target:
parts.append(EnclosedNodeList([
Entity(col) if isinstance(col, basestring) else col
for col in oc._conflict_target]))
parts.append(SQL('DO NOTHING'))
return NodeList(parts)
elif action and action != 'update':
raise ValueError('The only supported actions for conflict '
'resolution with Postgresql are "ignore" or '
'"update".')
elif not oc._update and not oc._preserve:
raise ValueError('If you are not performing any updates (or '
'preserving any INSERTed values), then the '
'conflict resolution action should be set to '
'"IGNORE".')
elif not (oc._conflict_target or oc._conflict_constraint):
raise ValueError('Postgres requires that a conflict target be '
'specified when doing an upsert.')
return self._build_on_conflict_update(oc, query)
def extract_date(self, date_part, date_field):
return fn.EXTRACT(NodeList((date_part, SQL('FROM'), date_field)))
def truncate_date(self, date_part, date_field):
return fn.DATE_TRUNC(date_part, date_field)
def to_timestamp(self, date_field):
return self.extract_date('EPOCH', date_field)
def from_timestamp(self, date_field):
# Ironically, here, Postgres means "to the Postgresql timestamp type".
return fn.to_timestamp(date_field)
def get_noop_select(self, ctx):
return ctx.sql(Select().columns(SQL('0')).where(SQL('false')))
def set_time_zone(self, timezone):
self.execute_sql('set time zone "%s";' % timezone)
class MySQLDatabase(Database):
field_types = {
'AUTO': 'INTEGER AUTO_INCREMENT',
'BIGAUTO': 'BIGINT AUTO_INCREMENT',
'BOOL': 'BOOL',
'DECIMAL': 'NUMERIC',
'DOUBLE': 'DOUBLE PRECISION',
'FLOAT': 'FLOAT',
'UUID': 'VARCHAR(40)',
'UUIDB': 'VARBINARY(16)'}
operations = {
'LIKE': 'LIKE BINARY',
'ILIKE': 'LIKE',
'REGEXP': 'REGEXP BINARY',
'IREGEXP': 'REGEXP',
'XOR': 'XOR'}
param = '%s'
quote = '``'
commit_select = True
compound_select_parentheses = CSQ_PARENTHESES_UNNESTED
for_update = True
index_using_precedes_table = True
limit_max = 2 ** 64 - 1
safe_create_index = False
safe_drop_index = False
sql_mode = 'PIPES_AS_CONCAT'
def init(self, database, **kwargs):
params = {
'charset': 'utf8',
'sql_mode': self.sql_mode,
'use_unicode': True}
params.update(kwargs)
if 'password' in params and mysql_passwd:
params['passwd'] = params.pop('password')
super(MySQLDatabase, self).init(database, **params)
def _connect(self):
if mysql is None:
raise ImproperlyConfigured('MySQL driver not installed!')
conn = mysql.connect(db=self.database, **self.connect_params)
return conn
def _set_server_version(self, conn):
try:
version_raw = conn.server_version
except AttributeError:
version_raw = conn.get_server_info()
self.server_version = self._extract_server_version(version_raw)
def _extract_server_version(self, version):
version = version.lower()
if 'maria' in version:
match_obj = re.search(r'(1\d\.\d+\.\d+)', version)
else:
match_obj = re.search(r'(\d\.\d+\.\d+)', version)
if match_obj is not None:
return tuple(int(num) for num in match_obj.groups()[0].split('.'))
warnings.warn('Unable to determine MySQL version: "%s"' % version)
return (0, 0, 0) # Unable to determine version!
def default_values_insert(self, ctx):
return ctx.literal('() VALUES ()')
def get_tables(self, schema=None):
query = ('SELECT table_name FROM information_schema.tables '
'WHERE table_schema = DATABASE() AND table_type != %s '
'ORDER BY table_name')
return [table for table, in self.execute_sql(query, ('VIEW',))]
def get_views(self, schema=None):
query = ('SELECT table_name, view_definition '
'FROM information_schema.views '
'WHERE table_schema = DATABASE() ORDER BY table_name')
cursor = self.execute_sql(query)
return [ViewMetadata(*row) for row in cursor.fetchall()]
def get_indexes(self, table, schema=None):
cursor = self.execute_sql('SHOW INDEX FROM `%s`' % table)
unique = set()
indexes = {}
for row in cursor.fetchall():
if not row[1]:
unique.add(row[2])
indexes.setdefault(row[2], [])
indexes[row[2]].append(row[4])
return [IndexMetadata(name, None, indexes[name], name in unique, table)
for name in indexes]
def get_columns(self, table, schema=None):
sql = """
SELECT column_name, is_nullable, data_type, column_default
FROM information_schema.columns
WHERE table_name = %s AND table_schema = DATABASE()"""
cursor = self.execute_sql(sql, (table,))
pks = set(self.get_primary_keys(table))
return [ColumnMetadata(name, dt, null == 'YES', name in pks, table, df)
for name, null, dt, df in cursor.fetchall()]
def get_primary_keys(self, table, schema=None):
cursor = self.execute_sql('SHOW INDEX FROM `%s`' % table)
return [row[4] for row in
filter(lambda row: row[2] == 'PRIMARY', cursor.fetchall())]
def get_foreign_keys(self, table, schema=None):
query = """
SELECT column_name, referenced_table_name, referenced_column_name
FROM information_schema.key_column_usage
WHERE table_name = %s
AND table_schema = DATABASE()
AND referenced_table_name IS NOT NULL
AND referenced_column_name IS NOT NULL"""
cursor = self.execute_sql(query, (table,))
return [
ForeignKeyMetadata(column, dest_table, dest_column, table)
for column, dest_table, dest_column in cursor.fetchall()]
def get_binary_type(self):
return mysql.Binary
def conflict_statement(self, on_conflict, query):
if not on_conflict._action: return
action = on_conflict._action.lower()
if action == 'replace':
return SQL('REPLACE')
elif action == 'ignore':
return SQL('INSERT IGNORE')
elif action != 'update':
raise ValueError('Un-supported action for conflict resolution. '
'MySQL supports REPLACE, IGNORE and UPDATE.')
def conflict_update(self, on_conflict, query):
if on_conflict._where or on_conflict._conflict_target or \
on_conflict._conflict_constraint:
raise ValueError('MySQL does not support the specification of '
'where clauses or conflict targets for conflict '
'resolution.')
updates = []
if on_conflict._preserve:
# Here we need to determine which function to use, which varies
# depending on the MySQL server version. MySQL and MariaDB prior to
# 10.3.3 use "VALUES", while MariaDB 10.3.3+ use "VALUE".
version = self.server_version or (0,)
if version[0] == 10 and version >= (10, 3, 3):
VALUE_FN = fn.VALUE
else:
VALUE_FN = fn.VALUES
for column in on_conflict._preserve:
entity = ensure_entity(column)
expression = NodeList((
ensure_entity(column),
SQL('='),
VALUE_FN(entity)))
updates.append(expression)
if on_conflict._update:
for k, v in on_conflict._update.items():
if not isinstance(v, Node):
# Attempt to resolve string field-names to their respective
# field object, to apply data-type conversions.
if isinstance(k, basestring):
k = getattr(query.table, k)
if isinstance(k, Field):
v = k.to_value(v)
else:
v = Value(v, unpack=False)
updates.append(NodeList((ensure_entity(k), SQL('='), v)))
if updates:
return NodeList((SQL('ON DUPLICATE KEY UPDATE'),
CommaNodeList(updates)))
def extract_date(self, date_part, date_field):
return fn.EXTRACT(NodeList((SQL(date_part), SQL('FROM'), date_field)))
def truncate_date(self, date_part, date_field):
return fn.DATE_FORMAT(date_field, __mysql_date_trunc__[date_part],
python_value=simple_date_time)
def to_timestamp(self, date_field):
return fn.UNIX_TIMESTAMP(date_field)
def from_timestamp(self, date_field):
return fn.FROM_UNIXTIME(date_field)
def random(self):
return fn.rand()
def get_noop_select(self, ctx):
return ctx.literal('DO 0')
# TRANSACTION CONTROL.
class _manual(_callable_context_manager):
def __init__(self, db):
self.db = db
def __enter__(self):
top = self.db.top_transaction()
if top is not None and not isinstance(top, _manual):
raise ValueError('Cannot enter manual commit block while a '
'transaction is active.')
self.db.push_transaction(self)
def __exit__(self, exc_type, exc_val, exc_tb):
if self.db.pop_transaction() is not self:
raise ValueError('Transaction stack corrupted while exiting '
'manual commit block.')
class _atomic(_callable_context_manager):
def __init__(self, db, *args, **kwargs):
self.db = db
self._transaction_args = (args, kwargs)
def __enter__(self):
if self.db.transaction_depth() == 0:
args, kwargs = self._transaction_args
self._helper = self.db.transaction(*args, **kwargs)
elif isinstance(self.db.top_transaction(), _manual):
raise ValueError('Cannot enter atomic commit block while in '
'manual commit mode.')
else:
self._helper = self.db.savepoint()
return self._helper.__enter__()
def __exit__(self, exc_type, exc_val, exc_tb):
return self._helper.__exit__(exc_type, exc_val, exc_tb)
class _transaction(_callable_context_manager):
def __init__(self, db, *args, **kwargs):
self.db = db
self._begin_args = (args, kwargs)
def _begin(self):
args, kwargs = self._begin_args
self.db.begin(*args, **kwargs)
def commit(self, begin=True):
self.db.commit()
if begin:
self._begin()
def rollback(self, begin=True):
self.db.rollback()
if begin:
self._begin()
def __enter__(self):
if self.db.transaction_depth() == 0:
self._begin()
self.db.push_transaction(self)
return self
def __exit__(self, exc_type, exc_val, exc_tb):
try:
if exc_type:
self.rollback(False)
elif self.db.transaction_depth() == 1:
try:
self.commit(False)
except:
self.rollback(False)
raise
finally:
self.db.pop_transaction()
class _savepoint(_callable_context_manager):
def __init__(self, db, sid=None):
self.db = db
self.sid = sid or 's' + uuid.uuid4().hex
self.quoted_sid = self.sid.join(self.db.quote)
def _begin(self):
self.db.execute_sql('SAVEPOINT %s;' % self.quoted_sid)
def commit(self, begin=True):
self.db.execute_sql('RELEASE SAVEPOINT %s;' % self.quoted_sid)
if begin: self._begin()
def rollback(self):
self.db.execute_sql('ROLLBACK TO SAVEPOINT %s;' % self.quoted_sid)
def __enter__(self):
self._begin()
return self
def __exit__(self, exc_type, exc_val, exc_tb):
if exc_type:
self.rollback()
else:
try:
self.commit(begin=False)
except:
self.rollback()
raise
# CURSOR REPRESENTATIONS.
class CursorWrapper(object):
def __init__(self, cursor):
self.cursor = cursor
self.count = 0
self.index = 0
self.initialized = False
self.populated = False
self.row_cache = []
def __iter__(self):
if self.populated:
return iter(self.row_cache)
return ResultIterator(self)
def __getitem__(self, item):
if isinstance(item, slice):
stop = item.stop
if stop is None or stop < 0:
self.fill_cache()
else:
self.fill_cache(stop)
return self.row_cache[item]
elif isinstance(item, int):
self.fill_cache(item if item > 0 else 0)
return self.row_cache[item]
else:
raise ValueError('CursorWrapper only supports integer and slice '
'indexes.')
def __len__(self):
self.fill_cache()
return self.count
def initialize(self):
pass
def iterate(self, cache=True):
row = self.cursor.fetchone()
if row is None:
self.populated = True
self.cursor.close()
raise StopIteration
elif not self.initialized:
self.initialize() # Lazy initialization.
self.initialized = True
self.count += 1
result = self.process_row(row)
if cache:
self.row_cache.append(result)
return result
def process_row(self, row):
return row
def iterator(self):
"""Efficient one-pass iteration over the result set."""
while True:
try:
yield self.iterate(False)
except StopIteration:
return
def fill_cache(self, n=0):
n = n or float('Inf')
if n < 0:
raise ValueError('Negative values are not supported.')
iterator = ResultIterator(self)
iterator.index = self.count
while not self.populated and (n > self.count):
try:
iterator.next()
except StopIteration:
break
class DictCursorWrapper(CursorWrapper):
def _initialize_columns(self):
description = self.cursor.description
self.columns = [t[0][t[0].rfind('.') + 1:].strip('()"`')
for t in description]
self.ncols = len(description)
initialize = _initialize_columns
def _row_to_dict(self, row):
result = {}
for i in range(self.ncols):
result.setdefault(self.columns[i], row[i]) # Do not overwrite.
return result
process_row = _row_to_dict
class NamedTupleCursorWrapper(CursorWrapper):
def initialize(self):
description = self.cursor.description
self.tuple_class = collections.namedtuple('Row', [
t[0][t[0].rfind('.') + 1:].strip('()"`') for t in description])
def process_row(self, row):
return self.tuple_class(*row)
class ObjectCursorWrapper(DictCursorWrapper):
def __init__(self, cursor, constructor):
super(ObjectCursorWrapper, self).__init__(cursor)
self.constructor = constructor
def process_row(self, row):
row_dict = self._row_to_dict(row)
return self.constructor(**row_dict)
class ResultIterator(object):
def __init__(self, cursor_wrapper):
self.cursor_wrapper = cursor_wrapper
self.index = 0
def __iter__(self):
return self
def next(self):
if self.index < self.cursor_wrapper.count:
obj = self.cursor_wrapper.row_cache[self.index]
elif not self.cursor_wrapper.populated:
self.cursor_wrapper.iterate()
obj = self.cursor_wrapper.row_cache[self.index]
else:
raise StopIteration
self.index += 1
return obj
__next__ = next
# FIELDS
class FieldAccessor(object):
def __init__(self, model, field, name):
self.model = model
self.field = field
self.name = name
def __get__(self, instance, instance_type=None):
if instance is not None:
return instance.__data__.get(self.name)
return self.field
def __set__(self, instance, value):
instance.__data__[self.name] = value
instance._dirty.add(self.name)
class ForeignKeyAccessor(FieldAccessor):
def __init__(self, model, field, name):
super(ForeignKeyAccessor, self).__init__(model, field, name)
self.rel_model = field.rel_model
def get_rel_instance(self, instance):
value = instance.__data__.get(self.name)
if value is not None or self.name in instance.__rel__:
if self.name not in instance.__rel__ and self.field.lazy_load:
obj = self.rel_model.get(self.field.rel_field == value)
instance.__rel__[self.name] = obj
return instance.__rel__.get(self.name, value)
elif not self.field.null and self.field.lazy_load:
raise self.rel_model.DoesNotExist
return value
def __get__(self, instance, instance_type=None):
if instance is not None:
return self.get_rel_instance(instance)
return self.field
def __set__(self, instance, obj):
if isinstance(obj, self.rel_model):
instance.__data__[self.name] = getattr(obj, self.field.rel_field.name)
instance.__rel__[self.name] = obj
else:
fk_value = instance.__data__.get(self.name)
instance.__data__[self.name] = obj
if (obj != fk_value or obj is None) and \
self.name in instance.__rel__:
del instance.__rel__[self.name]
instance._dirty.add(self.name)
class BackrefAccessor(object):
def __init__(self, field):
self.field = field
self.model = field.rel_model
self.rel_model = field.model
def __get__(self, instance, instance_type=None):
if instance is not None:
dest = self.field.rel_field.name
return (self.rel_model
.select()
.where(self.field == getattr(instance, dest)))
return self
class ObjectIdAccessor(object):
"""Gives direct access to the underlying id"""
def __init__(self, field):
self.field = field
def __get__(self, instance, instance_type=None):
if instance is not None:
value = instance.__data__.get(self.field.name)
# Pull the object-id from the related object if it is not set.
if value is None and self.field.name in instance.__rel__:
rel_obj = instance.__rel__[self.field.name]
value = getattr(rel_obj, self.field.rel_field.name)
return value
return self.field
def __set__(self, instance, value):
setattr(instance, self.field.name, value)
class Field(ColumnBase):
_field_counter = 0
_order = 0
accessor_class = FieldAccessor
auto_increment = False
default_index_type = None
field_type = 'DEFAULT'
unpack = True
def __init__(self, null=False, index=False, unique=False, column_name=None,
default=None, primary_key=False, constraints=None,
sequence=None, collation=None, unindexed=False, choices=None,
help_text=None, verbose_name=None, index_type=None,
db_column=None, _hidden=False):
if db_column is not None:
__deprecated__('"db_column" has been deprecated in favor of '
'"column_name" for Field objects.')
column_name = db_column
self.null = null
self.index = index
self.unique = unique
self.column_name = column_name
self.default = default
self.primary_key = primary_key
self.constraints = constraints # List of column constraints.
self.sequence = sequence # Name of sequence, e.g. foo_id_seq.
self.collation = collation
self.unindexed = unindexed
self.choices = choices
self.help_text = help_text
self.verbose_name = verbose_name
self.index_type = index_type or self.default_index_type
self._hidden = _hidden
# Used internally for recovering the order in which Fields were defined
# on the Model class.
Field._field_counter += 1
self._order = Field._field_counter
self._sort_key = (self.primary_key and 1 or 2), self._order
def __hash__(self):
return hash(self.name + '.' + self.model.__name__)
def __repr__(self):
if hasattr(self, 'model') and getattr(self, 'name', None):
return '<%s: %s.%s>' % (type(self).__name__,
self.model.__name__,
self.name)
return '<%s: (unbound)>' % type(self).__name__
def bind(self, model, name, set_attribute=True):
self.model = model
self.name = self.safe_name = name
self.column_name = self.column_name or name
if set_attribute:
setattr(model, name, self.accessor_class(model, self, name))
@property
def column(self):
return Column(self.model._meta.table, self.column_name)
def adapt(self, value):
return value
def db_value(self, value):
return value if value is None else self.adapt(value)
def python_value(self, value):
return value if value is None else self.adapt(value)
def to_value(self, value):
return Value(value, self.db_value, unpack=False)
def get_sort_key(self, ctx):
return self._sort_key
def __sql__(self, ctx):
return ctx.sql(self.column)
def get_modifiers(self):
pass
def ddl_datatype(self, ctx):
if ctx and ctx.state.field_types:
column_type = ctx.state.field_types.get(self.field_type,
self.field_type)
else:
column_type = self.field_type
modifiers = self.get_modifiers()
if column_type and modifiers:
modifier_literal = ', '.join([str(m) for m in modifiers])
return SQL('%s(%s)' % (column_type, modifier_literal))
else:
return SQL(column_type)
def ddl(self, ctx):
accum = [Entity(self.column_name)]
data_type = self.ddl_datatype(ctx)
if data_type:
accum.append(data_type)
if self.unindexed:
accum.append(SQL('UNINDEXED'))
if not self.null:
accum.append(SQL('NOT NULL'))
if self.primary_key:
accum.append(SQL('PRIMARY KEY'))
if self.sequence:
accum.append(SQL("DEFAULT NEXTVAL('%s')" % self.sequence))
if self.constraints:
accum.extend(self.constraints)
if self.collation:
accum.append(SQL('COLLATE %s' % self.collation))
return NodeList(accum)
class AnyField(Field):
field_type = 'ANY'
class IntegerField(Field):
field_type = 'INT'
def adapt(self, value):
try:
return int(value)
except ValueError:
return value
class BigIntegerField(IntegerField):
field_type = 'BIGINT'
class SmallIntegerField(IntegerField):
field_type = 'SMALLINT'
class AutoField(IntegerField):
auto_increment = True
field_type = 'AUTO'
def __init__(self, *args, **kwargs):
if kwargs.get('primary_key') is False:
raise ValueError('%s must always be a primary key.' % type(self))
kwargs['primary_key'] = True
super(AutoField, self).__init__(*args, **kwargs)
class BigAutoField(AutoField):
field_type = 'BIGAUTO'
class IdentityField(AutoField):
field_type = 'INT GENERATED BY DEFAULT AS IDENTITY'
def __init__(self, generate_always=False, **kwargs):
if generate_always:
self.field_type = 'INT GENERATED ALWAYS AS IDENTITY'
super(IdentityField, self).__init__(**kwargs)
class PrimaryKeyField(AutoField):
def __init__(self, *args, **kwargs):
__deprecated__('"PrimaryKeyField" has been renamed to "AutoField". '
'Please update your code accordingly as this will be '
'completely removed in a subsequent release.')
super(PrimaryKeyField, self).__init__(*args, **kwargs)
class FloatField(Field):
field_type = 'FLOAT'
def adapt(self, value):
try:
return float(value)
except ValueError:
return value
class DoubleField(FloatField):
field_type = 'DOUBLE'
class DecimalField(Field):
field_type = 'DECIMAL'
def __init__(self, max_digits=10, decimal_places=5, auto_round=False,
rounding=None, *args, **kwargs):
self.max_digits = max_digits
self.decimal_places = decimal_places
self.auto_round = auto_round
self.rounding = rounding or decimal.DefaultContext.rounding
self._exp = decimal.Decimal(10) ** (-self.decimal_places)
super(DecimalField, self).__init__(*args, **kwargs)
def get_modifiers(self):
return [self.max_digits, self.decimal_places]
def db_value(self, value):
D = decimal.Decimal
if not value:
return value if value is None else D(0)
if self.auto_round:
decimal_value = D(text_type(value))
return decimal_value.quantize(self._exp, rounding=self.rounding)
return value
def python_value(self, value):
if value is not None:
if isinstance(value, decimal.Decimal):
return value
return decimal.Decimal(text_type(value))
class _StringField(Field):
def adapt(self, value):
if isinstance(value, text_type):
return value
elif isinstance(value, bytes_type):
return value.decode('utf-8')
return text_type(value)
def __add__(self, other): return StringExpression(self, OP.CONCAT, other)
def __radd__(self, other): return StringExpression(other, OP.CONCAT, self)
class CharField(_StringField):
field_type = 'VARCHAR'
def __init__(self, max_length=255, *args, **kwargs):
self.max_length = max_length
super(CharField, self).__init__(*args, **kwargs)
def get_modifiers(self):
return self.max_length and [self.max_length] or None
class FixedCharField(CharField):
field_type = 'CHAR'
def python_value(self, value):
value = super(FixedCharField, self).python_value(value)
if value:
value = value.strip()
return value
class TextField(_StringField):
field_type = 'TEXT'
class BlobField(Field):
field_type = 'BLOB'
def _db_hook(self, database):
if database is None:
self._constructor = bytearray
else:
self._constructor = database.get_binary_type()
def bind(self, model, name, set_attribute=True):
self._constructor = bytearray
if model._meta.database:
if isinstance(model._meta.database, Proxy):
model._meta.database.attach_callback(self._db_hook)
else:
self._db_hook(model._meta.database)
# Attach a hook to the model metadata; in the event the database is
# changed or set at run-time, we will be sure to apply our callback and
# use the proper data-type for our database driver.
model._meta._db_hooks.append(self._db_hook)
return super(BlobField, self).bind(model, name, set_attribute)
def db_value(self, value):
if isinstance(value, text_type):
value = value.encode('raw_unicode_escape')
if isinstance(value, bytes_type):
return self._constructor(value)
return value
class BitField(BitwiseMixin, BigIntegerField):
def __init__(self, *args, **kwargs):
kwargs.setdefault('default', 0)
super(BitField, self).__init__(*args, **kwargs)
self.__current_flag = 1
def flag(self, value=None):
if value is None:
value = self.__current_flag
self.__current_flag <<= 1
else:
self.__current_flag = value << 1
class FlagDescriptor(ColumnBase):
def __init__(self, field, value):
self._field = field
self._value = value
super(FlagDescriptor, self).__init__()
def clear(self):
return self._field.bin_and(~self._value)
def set(self):
return self._field.bin_or(self._value)
def __get__(self, instance, instance_type=None):
if instance is None:
return self
value = getattr(instance, self._field.name) or 0
return (value & self._value) != 0
def __set__(self, instance, is_set):
if is_set not in (True, False):
raise ValueError('Value must be either True or False')
value = getattr(instance, self._field.name) or 0
if is_set:
value |= self._value
else:
value &= ~self._value
setattr(instance, self._field.name, value)
def __sql__(self, ctx):
return ctx.sql(self._field.bin_and(self._value) != 0)
return FlagDescriptor(self, value)
class BigBitFieldData(object):
def __init__(self, instance, name):
self.instance = instance
self.name = name
value = self.instance.__data__.get(self.name)
if not value:
value = bytearray()
elif not isinstance(value, bytearray):
value = bytearray(value)
self._buffer = self.instance.__data__[self.name] = value
def _ensure_length(self, idx):
byte_num, byte_offset = divmod(idx, 8)
cur_size = len(self._buffer)
if cur_size <= byte_num:
self._buffer.extend(b'\x00' * ((byte_num + 1) - cur_size))
return byte_num, byte_offset
def set_bit(self, idx):
byte_num, byte_offset = self._ensure_length(idx)
self._buffer[byte_num] |= (1 << byte_offset)
def clear_bit(self, idx):
byte_num, byte_offset = self._ensure_length(idx)
self._buffer[byte_num] &= ~(1 << byte_offset)
def toggle_bit(self, idx):
byte_num, byte_offset = self._ensure_length(idx)
self._buffer[byte_num] ^= (1 << byte_offset)
return bool(self._buffer[byte_num] & (1 << byte_offset))
def is_set(self, idx):
byte_num, byte_offset = self._ensure_length(idx)
return bool(self._buffer[byte_num] & (1 << byte_offset))
def __repr__(self):
return repr(self._buffer)
class BigBitFieldAccessor(FieldAccessor):
def __get__(self, instance, instance_type=None):
if instance is None:
return self.field
return BigBitFieldData(instance, self.name)
def __set__(self, instance, value):
if isinstance(value, memoryview):
value = value.tobytes()
elif isinstance(value, buffer_type):
value = bytes(value)
elif isinstance(value, bytearray):
value = bytes_type(value)
elif isinstance(value, BigBitFieldData):
value = bytes_type(value._buffer)
elif isinstance(value, text_type):
value = value.encode('utf-8')
elif not isinstance(value, bytes_type):
raise ValueError('Value must be either a bytes, memoryview or '
'BigBitFieldData instance.')
super(BigBitFieldAccessor, self).__set__(instance, value)
class BigBitField(BlobField):
accessor_class = BigBitFieldAccessor
def __init__(self, *args, **kwargs):
kwargs.setdefault('default', bytes_type)
super(BigBitField, self).__init__(*args, **kwargs)
def db_value(self, value):
return bytes_type(value) if value is not None else value
class UUIDField(Field):
field_type = 'UUID'
def db_value(self, value):
if isinstance(value, basestring) and len(value) == 32:
# Hex string. No transformation is necessary.
return value
elif isinstance(value, bytes) and len(value) == 16:
# Allow raw binary representation.
value = uuid.UUID(bytes=value)
if isinstance(value, uuid.UUID):
return value.hex
try:
return uuid.UUID(value).hex
except:
return value
def python_value(self, value):
if isinstance(value, uuid.UUID):
return value
return uuid.UUID(value) if value is not None else None
class BinaryUUIDField(BlobField):
field_type = 'UUIDB'
def db_value(self, value):
if isinstance(value, bytes) and len(value) == 16:
# Raw binary value. No transformation is necessary.
return self._constructor(value)
elif isinstance(value, basestring) and len(value) == 32:
# Allow hex string representation.
value = uuid.UUID(hex=value)
if isinstance(value, uuid.UUID):
return self._constructor(value.bytes)
elif value is not None:
raise ValueError('value for binary UUID field must be UUID(), '
'a hexadecimal string, or a bytes object.')
def python_value(self, value):
if isinstance(value, uuid.UUID):
return value
elif isinstance(value, memoryview):
value = value.tobytes()
elif value and not isinstance(value, bytes):
value = bytes(value)
return uuid.UUID(bytes=value) if value is not None else None
def _date_part(date_part):
def dec(self):
return self.model._meta.database.extract_date(date_part, self)
return dec
def format_date_time(value, formats, post_process=None):
post_process = post_process or (lambda x: x)
for fmt in formats:
try:
return post_process(datetime.datetime.strptime(value, fmt))
except ValueError:
pass
return value
def simple_date_time(value):
try:
return datetime.datetime.strptime(value, '%Y-%m-%d %H:%M:%S')
except (TypeError, ValueError):
return value
class _BaseFormattedField(Field):
formats = None
def __init__(self, formats=None, *args, **kwargs):
if formats is not None:
self.formats = formats
super(_BaseFormattedField, self).__init__(*args, **kwargs)
class DateTimeField(_BaseFormattedField):
field_type = 'DATETIME'
formats = [
'%Y-%m-%d %H:%M:%S.%f',
'%Y-%m-%d %H:%M:%S',
'%Y-%m-%d',
]
def adapt(self, value):
if value and isinstance(value, basestring):
return format_date_time(value, self.formats)
return value
def to_timestamp(self):
return self.model._meta.database.to_timestamp(self)
def truncate(self, part):
return self.model._meta.database.truncate_date(part, self)
year = property(_date_part('year'))
month = property(_date_part('month'))
day = property(_date_part('day'))
hour = property(_date_part('hour'))
minute = property(_date_part('minute'))
second = property(_date_part('second'))
class DateField(_BaseFormattedField):
field_type = 'DATE'
formats = [
'%Y-%m-%d',
'%Y-%m-%d %H:%M:%S',
'%Y-%m-%d %H:%M:%S.%f',
]
def adapt(self, value):
if value and isinstance(value, basestring):
pp = lambda x: x.date()
return format_date_time(value, self.formats, pp)
elif value and isinstance(value, datetime.datetime):
return value.date()
return value
def to_timestamp(self):
return self.model._meta.database.to_timestamp(self)
def truncate(self, part):
return self.model._meta.database.truncate_date(part, self)
year = property(_date_part('year'))
month = property(_date_part('month'))
day = property(_date_part('day'))
class TimeField(_BaseFormattedField):
field_type = 'TIME'
formats = [
'%H:%M:%S.%f',
'%H:%M:%S',
'%H:%M',
'%Y-%m-%d %H:%M:%S.%f',
'%Y-%m-%d %H:%M:%S',
]
def adapt(self, value):
if value:
if isinstance(value, basestring):
pp = lambda x: x.time()
return format_date_time(value, self.formats, pp)
elif isinstance(value, datetime.datetime):
return value.time()
if value is not None and isinstance(value, datetime.timedelta):
return (datetime.datetime.min + value).time()
return value
hour = property(_date_part('hour'))
minute = property(_date_part('minute'))
second = property(_date_part('second'))
def _timestamp_date_part(date_part):
def dec(self):
db = self.model._meta.database
expr = ((self / Value(self.resolution, converter=False))
if self.resolution > 1 else self)
return db.extract_date(date_part, db.from_timestamp(expr))
return dec
class TimestampField(BigIntegerField):
# Support second -> microsecond resolution.
valid_resolutions = [10**i for i in range(7)]
def __init__(self, *args, **kwargs):
self.resolution = kwargs.pop('resolution', None)
if not self.resolution:
self.resolution = 1
elif self.resolution in range(2, 7):
self.resolution = 10 ** self.resolution
elif self.resolution not in self.valid_resolutions:
raise ValueError('TimestampField resolution must be one of: %s' %
', '.join(str(i) for i in self.valid_resolutions))
self.ticks_to_microsecond = 1000000 // self.resolution
self.utc = kwargs.pop('utc', False) or False
dflt = datetime.datetime.utcnow if self.utc else datetime.datetime.now
kwargs.setdefault('default', dflt)
super(TimestampField, self).__init__(*args, **kwargs)
def local_to_utc(self, dt):
# Convert naive local datetime into naive UTC, e.g.:
# 2019-03-01T12:00:00 (local=US/Central) -> 2019-03-01T18:00:00.
# 2019-05-01T12:00:00 (local=US/Central) -> 2019-05-01T17:00:00.
# 2019-03-01T12:00:00 (local=UTC) -> 2019-03-01T12:00:00.
return datetime.datetime(*time.gmtime(time.mktime(dt.timetuple()))[:6])
def utc_to_local(self, dt):
# Convert a naive UTC datetime into local time, e.g.:
# 2019-03-01T18:00:00 (local=US/Central) -> 2019-03-01T12:00:00.
# 2019-05-01T17:00:00 (local=US/Central) -> 2019-05-01T12:00:00.
# 2019-03-01T12:00:00 (local=UTC) -> 2019-03-01T12:00:00.
ts = calendar.timegm(dt.utctimetuple())
return datetime.datetime.fromtimestamp(ts)
def get_timestamp(self, value):
if self.utc:
# If utc-mode is on, then we assume all naive datetimes are in UTC.
return calendar.timegm(value.utctimetuple())
else:
return time.mktime(value.timetuple())
def db_value(self, value):
if value is None:
return
if isinstance(value, datetime.datetime):
pass
elif isinstance(value, datetime.date):
value = datetime.datetime(value.year, value.month, value.day)
else:
return int(round(value * self.resolution))
timestamp = self.get_timestamp(value)
if self.resolution > 1:
timestamp += (value.microsecond * .000001)
timestamp *= self.resolution
return int(round(timestamp))
def python_value(self, value):
if value is not None and isinstance(value, (int, float, long)):
if self.resolution > 1:
value, ticks = divmod(value, self.resolution)
microseconds = int(ticks * self.ticks_to_microsecond)
else:
microseconds = 0
if self.utc:
value = datetime.datetime.utcfromtimestamp(value)
else:
value = datetime.datetime.fromtimestamp(value)
if microseconds:
value = value.replace(microsecond=microseconds)
return value
def from_timestamp(self):
expr = ((self / Value(self.resolution, converter=False))
if self.resolution > 1 else self)
return self.model._meta.database.from_timestamp(expr)
year = property(_timestamp_date_part('year'))
month = property(_timestamp_date_part('month'))
day = property(_timestamp_date_part('day'))
hour = property(_timestamp_date_part('hour'))
minute = property(_timestamp_date_part('minute'))
second = property(_timestamp_date_part('second'))
class IPField(BigIntegerField):
def db_value(self, val):
if val is not None:
return struct.unpack('!I', socket.inet_aton(val))[0]
def python_value(self, val):
if val is not None:
return socket.inet_ntoa(struct.pack('!I', val))
class BooleanField(Field):
field_type = 'BOOL'
adapt = bool
class BareField(Field):
def __init__(self, adapt=None, *args, **kwargs):
super(BareField, self).__init__(*args, **kwargs)
if adapt is not None:
self.adapt = adapt
def ddl_datatype(self, ctx):
return
class ForeignKeyField(Field):
accessor_class = ForeignKeyAccessor
backref_accessor_class = BackrefAccessor
def __init__(self, model, field=None, backref=None, on_delete=None,
on_update=None, deferrable=None, _deferred=None,
rel_model=None, to_field=None, object_id_name=None,
lazy_load=True, constraint_name=None, related_name=None,
*args, **kwargs):
kwargs.setdefault('index', True)
super(ForeignKeyField, self).__init__(*args, **kwargs)
if rel_model is not None:
__deprecated__('"rel_model" has been deprecated in favor of '
'"model" for ForeignKeyField objects.')
model = rel_model
if to_field is not None:
__deprecated__('"to_field" has been deprecated in favor of '
'"field" for ForeignKeyField objects.')
field = to_field
if related_name is not None:
__deprecated__('"related_name" has been deprecated in favor of '
'"backref" for Field objects.')
backref = related_name
self._is_self_reference = model == 'self'
self.rel_model = model
self.rel_field = field
self.declared_backref = backref
self.backref = None
self.on_delete = on_delete
self.on_update = on_update
self.deferrable = deferrable
self.deferred = _deferred
self.object_id_name = object_id_name
self.lazy_load = lazy_load
self.constraint_name = constraint_name
@property
def field_type(self):
if not isinstance(self.rel_field, AutoField):
return self.rel_field.field_type
elif isinstance(self.rel_field, BigAutoField):
return BigIntegerField.field_type
return IntegerField.field_type
def get_modifiers(self):
if not isinstance(self.rel_field, AutoField):
return self.rel_field.get_modifiers()
return super(ForeignKeyField, self).get_modifiers()
def adapt(self, value):
return self.rel_field.adapt(value)
def db_value(self, value):
if isinstance(value, self.rel_model):
value = getattr(value, self.rel_field.name)
return self.rel_field.db_value(value)
def python_value(self, value):
if isinstance(value, self.rel_model):
return value
return self.rel_field.python_value(value)
def bind(self, model, name, set_attribute=True):
if not self.column_name:
self.column_name = name if name.endswith('_id') else name + '_id'
if not self.object_id_name:
self.object_id_name = self.column_name
if self.object_id_name == name:
self.object_id_name += '_id'
elif self.object_id_name == name:
raise ValueError('ForeignKeyField "%s"."%s" specifies an '
'object_id_name that conflicts with its field '
'name.' % (model._meta.name, name))
if self._is_self_reference:
self.rel_model = model
if isinstance(self.rel_field, basestring):
self.rel_field = getattr(self.rel_model, self.rel_field)
elif self.rel_field is None:
self.rel_field = self.rel_model._meta.primary_key
# Bind field before assigning backref, so field is bound when
# calling declared_backref() (if callable).
super(ForeignKeyField, self).bind(model, name, set_attribute)
self.safe_name = self.object_id_name
if callable_(self.declared_backref):
self.backref = self.declared_backref(self)
else:
self.backref, self.declared_backref = self.declared_backref, None
if not self.backref:
self.backref = '%s_set' % model._meta.name
if set_attribute:
setattr(model, self.object_id_name, ObjectIdAccessor(self))
if self.backref not in '!+':
setattr(self.rel_model, self.backref,
self.backref_accessor_class(self))
def foreign_key_constraint(self):
parts = []
if self.constraint_name:
parts.extend((SQL('CONSTRAINT'), Entity(self.constraint_name)))
parts.extend([
SQL('FOREIGN KEY'),
EnclosedNodeList((self,)),
SQL('REFERENCES'),
self.rel_model,
EnclosedNodeList((self.rel_field,))])
if self.on_delete:
parts.append(SQL('ON DELETE %s' % self.on_delete))
if self.on_update:
parts.append(SQL('ON UPDATE %s' % self.on_update))
if self.deferrable:
parts.append(SQL('DEFERRABLE %s' % self.deferrable))
return NodeList(parts)
def __getattr__(self, attr):
if attr.startswith('__'):
# Prevent recursion error when deep-copying.
raise AttributeError('Cannot look-up non-existant "__" methods.')
if attr in self.rel_model._meta.fields:
return self.rel_model._meta.fields[attr]
raise AttributeError('Foreign-key has no attribute %s, nor is it a '
'valid field on the related model.' % attr)
class DeferredForeignKey(Field):
_unresolved = set()
def __init__(self, rel_model_name, **kwargs):
self.field_kwargs = kwargs
self.rel_model_name = rel_model_name.lower()
DeferredForeignKey._unresolved.add(self)
super(DeferredForeignKey, self).__init__(
column_name=kwargs.get('column_name'),
null=kwargs.get('null'),
primary_key=kwargs.get('primary_key'))
__hash__ = object.__hash__
def __deepcopy__(self, memo=None):
return DeferredForeignKey(self.rel_model_name, **self.field_kwargs)
def set_model(self, rel_model):
field = ForeignKeyField(rel_model, _deferred=True, **self.field_kwargs)
if field.primary_key:
# NOTE: this calls add_field() under-the-hood.
self.model._meta.set_primary_key(self.name, field)
else:
self.model._meta.add_field(self.name, field)
@staticmethod
def resolve(model_cls):
unresolved = sorted(DeferredForeignKey._unresolved,
key=operator.attrgetter('_order'))
for dr in unresolved:
if dr.rel_model_name == model_cls.__name__.lower():
dr.set_model(model_cls)
DeferredForeignKey._unresolved.discard(dr)
class DeferredThroughModel(object):
def __init__(self):
self._refs = []
def set_field(self, model, field, name):
self._refs.append((model, field, name))
def set_model(self, through_model):
for src_model, m2mfield, name in self._refs:
m2mfield.through_model = through_model
src_model._meta.add_field(name, m2mfield)
class MetaField(Field):
column_name = default = model = name = None
primary_key = False
class ManyToManyFieldAccessor(FieldAccessor):
def __init__(self, model, field, name):
super(ManyToManyFieldAccessor, self).__init__(model, field, name)
self.model = field.model
self.rel_model = field.rel_model
self.through_model = field.through_model
src_fks = self.through_model._meta.model_refs[self.model]
dest_fks = self.through_model._meta.model_refs[self.rel_model]
if not src_fks:
raise ValueError('Cannot find foreign-key to "%s" on "%s" model.' %
(self.model, self.through_model))
elif not dest_fks:
raise ValueError('Cannot find foreign-key to "%s" on "%s" model.' %
(self.rel_model, self.through_model))
self.src_fk = src_fks[0]
self.dest_fk = dest_fks[0]
def __get__(self, instance, instance_type=None, force_query=False):
if instance is not None:
if not force_query and self.src_fk.backref != '+':
backref = getattr(instance, self.src_fk.backref)
if isinstance(backref, list):
return [getattr(obj, self.dest_fk.name) for obj in backref]
src_id = getattr(instance, self.src_fk.rel_field.name)
return (ManyToManyQuery(instance, self, self.rel_model)
.join(self.through_model)
.join(self.model)
.where(self.src_fk == src_id))
return self.field
def __set__(self, instance, value):
query = self.__get__(instance, force_query=True)
query.add(value, clear_existing=True)
class ManyToManyField(MetaField):
accessor_class = ManyToManyFieldAccessor
def __init__(self, model, backref=None, through_model=None, on_delete=None,
on_update=None, _is_backref=False):
if through_model is not None:
if not (isinstance(through_model, DeferredThroughModel) or
is_model(through_model)):
raise TypeError('Unexpected value for through_model. Expected '
'Model or DeferredThroughModel.')
if not _is_backref and (on_delete is not None or on_update is not None):
raise ValueError('Cannot specify on_delete or on_update when '
'through_model is specified.')
self.rel_model = model
self.backref = backref
self._through_model = through_model
self._on_delete = on_delete
self._on_update = on_update
self._is_backref = _is_backref
def _get_descriptor(self):
return ManyToManyFieldAccessor(self)
def bind(self, model, name, set_attribute=True):
if isinstance(self._through_model, DeferredThroughModel):
self._through_model.set_field(model, self, name)
return
super(ManyToManyField, self).bind(model, name, set_attribute)
if not self._is_backref:
many_to_many_field = ManyToManyField(
self.model,
backref=name,
through_model=self.through_model,
on_delete=self._on_delete,
on_update=self._on_update,
_is_backref=True)
self.backref = self.backref or model._meta.name + 's'
self.rel_model._meta.add_field(self.backref, many_to_many_field)
def get_models(self):
return [model for _, model in sorted((
(self._is_backref, self.model),
(not self._is_backref, self.rel_model)))]
@property
def through_model(self):
if self._through_model is None:
self._through_model = self._create_through_model()
return self._through_model
@through_model.setter
def through_model(self, value):
self._through_model = value
def _create_through_model(self):
lhs, rhs = self.get_models()
tables = [model._meta.table_name for model in (lhs, rhs)]
class Meta:
database = self.model._meta.database
schema = self.model._meta.schema
table_name = '%s_%s_through' % tuple(tables)
indexes = (
((lhs._meta.name, rhs._meta.name),
True),)
params = {'on_delete': self._on_delete, 'on_update': self._on_update}
attrs = {
lhs._meta.name: ForeignKeyField(lhs, **params),
rhs._meta.name: ForeignKeyField(rhs, **params),
'Meta': Meta}
klass_name = '%s%sThrough' % (lhs.__name__, rhs.__name__)
return type(klass_name, (Model,), attrs)
def get_through_model(self):
# XXX: Deprecated. Just use the "through_model" property.
return self.through_model
class VirtualField(MetaField):
field_class = None
def __init__(self, field_class=None, *args, **kwargs):
Field = field_class if field_class is not None else self.field_class
self.field_instance = Field() if Field is not None else None
super(VirtualField, self).__init__(*args, **kwargs)
def db_value(self, value):
if self.field_instance is not None:
return self.field_instance.db_value(value)
return value
def python_value(self, value):
if self.field_instance is not None:
return self.field_instance.python_value(value)
return value
def bind(self, model, name, set_attribute=True):
self.model = model
self.column_name = self.name = self.safe_name = name
setattr(model, name, self.accessor_class(model, self, name))
class CompositeKey(MetaField):
sequence = None
def __init__(self, *field_names):
self.field_names = field_names
self._safe_field_names = None
@property
def safe_field_names(self):
if self._safe_field_names is None:
if self.model is None:
return self.field_names
self._safe_field_names = [self.model._meta.fields[f].safe_name
for f in self.field_names]
return self._safe_field_names
def __get__(self, instance, instance_type=None):
if instance is not None:
return tuple([getattr(instance, f) for f in self.safe_field_names])
return self
def __set__(self, instance, value):
if not isinstance(value, (list, tuple)):
raise TypeError('A list or tuple must be used to set the value of '
'a composite primary key.')
if len(value) != len(self.field_names):
raise ValueError('The length of the value must equal the number '
'of columns of the composite primary key.')
for idx, field_value in enumerate(value):
setattr(instance, self.field_names[idx], field_value)
def __eq__(self, other):
expressions = [(self.model._meta.fields[field] == value)
for field, value in zip(self.field_names, other)]
return reduce(operator.and_, expressions)
def __ne__(self, other):
return ~(self == other)
def __hash__(self):
return hash((self.model.__name__, self.field_names))
def __sql__(self, ctx):
# If the composite PK is being selected, do not use parens. Elsewhere,
# such as in an expression, we want to use parentheses and treat it as
# a row value.
parens = ctx.scope != SCOPE_SOURCE
return ctx.sql(NodeList([self.model._meta.fields[field]
for field in self.field_names], ', ', parens))
def bind(self, model, name, set_attribute=True):
self.model = model
self.column_name = self.name = self.safe_name = name
setattr(model, self.name, self)
class _SortedFieldList(object):
__slots__ = ('_keys', '_items')
def __init__(self):
self._keys = []
self._items = []
def __getitem__(self, i):
return self._items[i]
def __iter__(self):
return iter(self._items)
def __contains__(self, item):
k = item._sort_key
i = bisect_left(self._keys, k)
j = bisect_right(self._keys, k)
return item in self._items[i:j]
def index(self, field):
return self._keys.index(field._sort_key)
def insert(self, item):
k = item._sort_key
i = bisect_left(self._keys, k)
self._keys.insert(i, k)
self._items.insert(i, item)
def remove(self, item):
idx = self.index(item)
del self._items[idx]
del self._keys[idx]
# MODELS
class SchemaManager(object):
def __init__(self, model, database=None, **context_options):
self.model = model
self._database = database
context_options.setdefault('scope', SCOPE_VALUES)
self.context_options = context_options
@property
def database(self):
db = self._database or self.model._meta.database
if db is None:
raise ImproperlyConfigured('database attribute does not appear to '
'be set on the model: %s' % self.model)
return db
@database.setter
def database(self, value):
self._database = value
def _create_context(self):
return self.database.get_sql_context(**self.context_options)
def _create_table(self, safe=True, **options):
is_temp = options.pop('temporary', False)
ctx = self._create_context()
ctx.literal('CREATE TEMPORARY TABLE ' if is_temp else 'CREATE TABLE ')
if safe:
ctx.literal('IF NOT EXISTS ')
ctx.sql(self.model).literal(' ')
columns = []
constraints = []
meta = self.model._meta
if meta.composite_key:
pk_columns = [meta.fields[field_name].column
for field_name in meta.primary_key.field_names]
constraints.append(NodeList((SQL('PRIMARY KEY'),
EnclosedNodeList(pk_columns))))
for field in meta.sorted_fields:
columns.append(field.ddl(ctx))
if isinstance(field, ForeignKeyField) and not field.deferred:
constraints.append(field.foreign_key_constraint())
if meta.constraints:
constraints.extend(meta.constraints)
constraints.extend(self._create_table_option_sql(options))
ctx.sql(EnclosedNodeList(columns + constraints))
if meta.table_settings is not None:
table_settings = ensure_tuple(meta.table_settings)
for setting in table_settings:
if not isinstance(setting, basestring):
raise ValueError('table_settings must be strings')
ctx.literal(' ').literal(setting)
extra_opts = []
if meta.strict_tables: extra_opts.append('STRICT')
if meta.without_rowid: extra_opts.append('WITHOUT ROWID')
if extra_opts:
ctx.literal(' %s' % ', '.join(extra_opts))
return ctx
def _create_table_option_sql(self, options):
accum = []
options = merge_dict(self.model._meta.options or {}, options)
if not options:
return accum
for key, value in sorted(options.items()):
if not isinstance(value, Node):
if is_model(value):
value = value._meta.table
else:
value = SQL(str(value))
accum.append(NodeList((SQL(key), value), glue='='))
return accum
def create_table(self, safe=True, **options):
self.database.execute(self._create_table(safe=safe, **options))
def _create_table_as(self, table_name, query, safe=True, **meta):
ctx = (self._create_context()
.literal('CREATE TEMPORARY TABLE '
if meta.get('temporary') else 'CREATE TABLE '))
if safe:
ctx.literal('IF NOT EXISTS ')
return (ctx
.sql(Entity(*ensure_tuple(table_name)))
.literal(' AS ')
.sql(query))
def create_table_as(self, table_name, query, safe=True, **meta):
ctx = self._create_table_as(table_name, query, safe=safe, **meta)
self.database.execute(ctx)
def _drop_table(self, safe=True, **options):
ctx = (self._create_context()
.literal('DROP TABLE IF EXISTS ' if safe else 'DROP TABLE ')
.sql(self.model))
if options.get('cascade'):
ctx = ctx.literal(' CASCADE')
elif options.get('restrict'):
ctx = ctx.literal(' RESTRICT')
return ctx
def drop_table(self, safe=True, **options):
self.database.execute(self._drop_table(safe=safe, **options))
def _truncate_table(self, restart_identity=False, cascade=False):
db = self.database
if not db.truncate_table:
return (self._create_context()
.literal('DELETE FROM ').sql(self.model))
ctx = self._create_context().literal('TRUNCATE TABLE ').sql(self.model)
if restart_identity:
ctx = ctx.literal(' RESTART IDENTITY')
if cascade:
ctx = ctx.literal(' CASCADE')
return ctx
def truncate_table(self, restart_identity=False, cascade=False):
self.database.execute(self._truncate_table(restart_identity, cascade))
def _create_indexes(self, safe=True):
return [self._create_index(index, safe)
for index in self.model._meta.fields_to_index()]
def _create_index(self, index, safe=True):
if isinstance(index, Index):
if not self.database.safe_create_index:
index = index.safe(False)
elif index._safe != safe:
index = index.safe(safe)
return self._create_context().sql(index)
def create_indexes(self, safe=True):
for query in self._create_indexes(safe=safe):
self.database.execute(query)
def _drop_indexes(self, safe=True):
return [self._drop_index(index, safe)
for index in self.model._meta.fields_to_index()
if isinstance(index, Index)]
def _drop_index(self, index, safe):
statement = 'DROP INDEX '
if safe and self.database.safe_drop_index:
statement += 'IF EXISTS '
if isinstance(index._table, Table) and index._table._schema:
index_name = Entity(index._table._schema, index._name)
else:
index_name = Entity(index._name)
return (self
._create_context()
.literal(statement)
.sql(index_name))
def drop_indexes(self, safe=True):
for query in self._drop_indexes(safe=safe):
self.database.execute(query)
def _check_sequences(self, field):
if not field.sequence or not self.database.sequences:
raise ValueError('Sequences are either not supported, or are not '
'defined for "%s".' % field.name)
def _sequence_for_field(self, field):
if field.model._meta.schema:
return Entity(field.model._meta.schema, field.sequence)
else:
return Entity(field.sequence)
def _create_sequence(self, field):
self._check_sequences(field)
if not self.database.sequence_exists(field.sequence):
return (self
._create_context()
.literal('CREATE SEQUENCE ')
.sql(self._sequence_for_field(field)))
def create_sequence(self, field):
seq_ctx = self._create_sequence(field)
if seq_ctx is not None:
self.database.execute(seq_ctx)
def _drop_sequence(self, field):
self._check_sequences(field)
if self.database.sequence_exists(field.sequence):
return (self
._create_context()
.literal('DROP SEQUENCE ')
.sql(self._sequence_for_field(field)))
def drop_sequence(self, field):
seq_ctx = self._drop_sequence(field)
if seq_ctx is not None:
self.database.execute(seq_ctx)
def _create_foreign_key(self, field):
name = 'fk_%s_%s_refs_%s' % (field.model._meta.table_name,
field.column_name,
field.rel_model._meta.table_name)
return (self
._create_context()
.literal('ALTER TABLE ')
.sql(field.model)
.literal(' ADD CONSTRAINT ')
.sql(Entity(_truncate_constraint_name(name)))
.literal(' ')
.sql(field.foreign_key_constraint()))
def create_foreign_key(self, field):
self.database.execute(self._create_foreign_key(field))
def create_sequences(self):
if self.database.sequences:
for field in self.model._meta.sorted_fields:
if field.sequence:
self.create_sequence(field)
def create_all(self, safe=True, **table_options):
self.create_sequences()
self.create_table(safe, **table_options)
self.create_indexes(safe=safe)
def drop_sequences(self):
if self.database.sequences:
for field in self.model._meta.sorted_fields:
if field.sequence:
self.drop_sequence(field)
def drop_all(self, safe=True, drop_sequences=True, **options):
self.drop_table(safe, **options)
if drop_sequences:
self.drop_sequences()
class Metadata(object):
def __init__(self, model, database=None, table_name=None, indexes=None,
primary_key=None, constraints=None, schema=None,
only_save_dirty=False, depends_on=None, options=None,
db_table=None, table_function=None, table_settings=None,
without_rowid=False, temporary=False, strict_tables=None,
legacy_table_names=True, **kwargs):
if db_table is not None:
__deprecated__('"db_table" has been deprecated in favor of '
'"table_name" for Models.')
table_name = db_table
self.model = model
self.database = database
self.fields = {}
self.columns = {}
self.combined = {}
self._sorted_field_list = _SortedFieldList()
self.sorted_fields = []
self.sorted_field_names = []
self.defaults = {}
self._default_by_name = {}
self._default_dict = {}
self._default_callables = {}
self._default_callable_list = []
self.name = model.__name__.lower()
self.table_function = table_function
self.legacy_table_names = legacy_table_names
if not table_name:
table_name = (self.table_function(model)
if self.table_function
else self.make_table_name())
self.table_name = table_name
self._table = None
self.indexes = list(indexes) if indexes else []
self.constraints = constraints
self._schema = schema
self.primary_key = primary_key
self.composite_key = self.auto_increment = None
self.only_save_dirty = only_save_dirty
self.depends_on = depends_on
self.table_settings = table_settings
self.without_rowid = without_rowid
self.strict_tables = strict_tables
self.temporary = temporary
self.refs = {}
self.backrefs = {}
self.model_refs = collections.defaultdict(list)
self.model_backrefs = collections.defaultdict(list)
self.manytomany = {}
self.options = options or {}
for key, value in kwargs.items():
setattr(self, key, value)
self._additional_keys = set(kwargs.keys())
# Allow objects to register hooks that are called if the model is bound
# to a different database. For example, BlobField uses a different
# Python data-type depending on the db driver / python version. When
# the database changes, we need to update any BlobField so they can use
# the appropriate data-type.
self._db_hooks = []
def make_table_name(self):
if self.legacy_table_names:
return re.sub(r'[^\w]+', '_', self.name)
return make_snake_case(self.model.__name__)
def model_graph(self, refs=True, backrefs=True, depth_first=True):
if not refs and not backrefs:
raise ValueError('One of `refs` or `backrefs` must be True.')
accum = [(None, self.model, None)]
seen = set()
queue = collections.deque((self,))
method = queue.pop if depth_first else queue.popleft
while queue:
curr = method()
if curr in seen: continue
seen.add(curr)
if refs:
for fk, model in curr.refs.items():
accum.append((fk, model, False))
queue.append(model._meta)
if backrefs:
for fk, model in curr.backrefs.items():
accum.append((fk, model, True))
queue.append(model._meta)
return accum
def add_ref(self, field):
rel = field.rel_model
self.refs[field] = rel
self.model_refs[rel].append(field)
rel._meta.backrefs[field] = self.model
rel._meta.model_backrefs[self.model].append(field)
def remove_ref(self, field):
rel = field.rel_model
del self.refs[field]
self.model_refs[rel].remove(field)
del rel._meta.backrefs[field]
rel._meta.model_backrefs[self.model].remove(field)
def add_manytomany(self, field):
self.manytomany[field.name] = field
def remove_manytomany(self, field):
del self.manytomany[field.name]
@property
def table(self):
if self._table is None:
self._table = Table(
self.table_name,
[field.column_name for field in self.sorted_fields],
schema=self.schema,
_model=self.model,
_database=self.database)
return self._table
@table.setter
def table(self, value):
raise AttributeError('Cannot set the "table".')
@table.deleter
def table(self):
self._table = None
@property
def schema(self):
return self._schema
@schema.setter
def schema(self, value):
self._schema = value
del self.table
@property
def entity(self):
if self._schema:
return Entity(self._schema, self.table_name)
else:
return Entity(self.table_name)
def _update_sorted_fields(self):
self.sorted_fields = list(self._sorted_field_list)
self.sorted_field_names = [f.name for f in self.sorted_fields]
def get_rel_for_model(self, model):
if isinstance(model, ModelAlias):
model = model.model
forwardrefs = self.model_refs.get(model, [])
backrefs = self.model_backrefs.get(model, [])
return (forwardrefs, backrefs)
def add_field(self, field_name, field, set_attribute=True):
if field_name in self.fields:
self.remove_field(field_name)
elif field_name in self.manytomany:
self.remove_manytomany(self.manytomany[field_name])
if not isinstance(field, MetaField):
del self.table
field.bind(self.model, field_name, set_attribute)
self.fields[field.name] = field
self.columns[field.column_name] = field
self.combined[field.name] = field
self.combined[field.column_name] = field
self._sorted_field_list.insert(field)
self._update_sorted_fields()
if field.default is not None:
# This optimization helps speed up model instance construction.
self.defaults[field] = field.default
if callable_(field.default):
self._default_callables[field] = field.default
self._default_callable_list.append((field.name,
field.default))
else:
self._default_dict[field] = field.default
self._default_by_name[field.name] = field.default
else:
field.bind(self.model, field_name, set_attribute)
if isinstance(field, ForeignKeyField):
self.add_ref(field)
elif isinstance(field, ManyToManyField) and field.name:
self.add_manytomany(field)
def remove_field(self, field_name):
if field_name not in self.fields:
return
del self.table
original = self.fields.pop(field_name)
del self.columns[original.column_name]
del self.combined[field_name]
try:
del self.combined[original.column_name]
except KeyError:
pass
self._sorted_field_list.remove(original)
self._update_sorted_fields()
if original.default is not None:
del self.defaults[original]
if self._default_callables.pop(original, None):
for i, (name, _) in enumerate(self._default_callable_list):
if name == field_name:
self._default_callable_list.pop(i)
break
else:
self._default_dict.pop(original, None)
self._default_by_name.pop(original.name, None)
if isinstance(original, ForeignKeyField):
self.remove_ref(original)
def set_primary_key(self, name, field):
self.composite_key = isinstance(field, CompositeKey)
self.add_field(name, field)
self.primary_key = field
self.auto_increment = (
field.auto_increment or
bool(field.sequence))
def get_primary_keys(self):
if self.composite_key:
return tuple([self.fields[field_name]
for field_name in self.primary_key.field_names])
else:
return (self.primary_key,) if self.primary_key is not False else ()
def get_default_dict(self):
dd = self._default_by_name.copy()
for field_name, default in self._default_callable_list:
dd[field_name] = default()
return dd
def fields_to_index(self):
indexes = []
for f in self.sorted_fields:
if f.primary_key:
continue
if f.index or f.unique:
indexes.append(ModelIndex(self.model, (f,), unique=f.unique,
using=f.index_type))
for index_obj in self.indexes:
if isinstance(index_obj, Node):
indexes.append(index_obj)
elif isinstance(index_obj, (list, tuple)):
index_parts, unique = index_obj
fields = []
for part in index_parts:
if isinstance(part, basestring):
fields.append(self.combined[part])
elif isinstance(part, Node):
fields.append(part)
else:
raise ValueError('Expected either a field name or a '
'subclass of Node. Got: %s' % part)
indexes.append(ModelIndex(self.model, fields, unique=unique))
return indexes
def set_database(self, database):
self.database = database
self.model._schema._database = database
del self.table
# Apply any hooks that have been registered.
for hook in self._db_hooks:
hook(database)
def set_table_name(self, table_name):
self.table_name = table_name
del self.table
class SubclassAwareMetadata(Metadata):
models = []
def __init__(self, model, *args, **kwargs):
super(SubclassAwareMetadata, self).__init__(model, *args, **kwargs)
self.models.append(model)
def map_models(self, fn):
for model in self.models:
fn(model)
class DoesNotExist(Exception): pass
class ModelBase(type):
inheritable = set(['constraints', 'database', 'indexes', 'primary_key',
'options', 'schema', 'table_function', 'temporary',
'only_save_dirty', 'legacy_table_names',
'table_settings', 'strict_tables'])
def __new__(cls, name, bases, attrs):
if name == MODEL_BASE or bases[0].__name__ == MODEL_BASE:
return super(ModelBase, cls).__new__(cls, name, bases, attrs)
meta_options = {}
meta = attrs.pop('Meta', None)
if meta:
for k, v in meta.__dict__.items():
if not k.startswith('_'):
meta_options[k] = v
pk = getattr(meta, 'primary_key', None)
pk_name = parent_pk = None
# Inherit any field descriptors by deep copying the underlying field
# into the attrs of the new model, additionally see if the bases define
# inheritable model options and swipe them.
for b in bases:
if not hasattr(b, '_meta'):
continue
base_meta = b._meta
if parent_pk is None:
parent_pk = deepcopy(base_meta.primary_key)
all_inheritable = cls.inheritable | base_meta._additional_keys
for k in base_meta.__dict__:
if k in all_inheritable and k not in meta_options:
meta_options[k] = base_meta.__dict__[k]
meta_options.setdefault('database', base_meta.database)
meta_options.setdefault('schema', base_meta.schema)
for (k, v) in b.__dict__.items():
if k in attrs: continue
if isinstance(v, FieldAccessor) and not v.field.primary_key:
attrs[k] = deepcopy(v.field)
sopts = meta_options.pop('schema_options', None) or {}
Meta = meta_options.get('model_metadata_class', Metadata)
Schema = meta_options.get('schema_manager_class', SchemaManager)
# Construct the new class.
cls = super(ModelBase, cls).__new__(cls, name, bases, attrs)
cls.__data__ = cls.__rel__ = None
cls._meta = Meta(cls, **meta_options)
cls._schema = Schema(cls, **sopts)
fields = []
for key, value in cls.__dict__.items():
if isinstance(value, Field):
if value.primary_key and pk:
raise ValueError('over-determined primary key %s.' % name)
elif value.primary_key:
pk, pk_name = value, key
else:
fields.append((key, value))
if pk is None:
if parent_pk is not False:
pk, pk_name = ((parent_pk, parent_pk.name)
if parent_pk is not None else
(AutoField(), 'id'))
else:
pk = False
elif isinstance(pk, CompositeKey):
pk_name = '__composite_key__'
cls._meta.composite_key = True
if pk is not False:
cls._meta.set_primary_key(pk_name, pk)
for name, field in fields:
cls._meta.add_field(name, field)
# Create a repr and error class before finalizing.
if hasattr(cls, '__str__') and '__repr__' not in attrs:
setattr(cls, '__repr__', lambda self: '<%s: %s>' % (
cls.__name__, self.__str__()))
exc_name = '%sDoesNotExist' % cls.__name__
exc_attrs = {'__module__': cls.__module__}
exception_class = type(exc_name, (DoesNotExist,), exc_attrs)
cls.DoesNotExist = exception_class
# Call validation hook, allowing additional model validation.
cls.validate_model()
DeferredForeignKey.resolve(cls)
return cls
def __repr__(self):
return '<Model: %s>' % self.__name__
def __iter__(self):
return iter(self.select())
def __getitem__(self, key):
return self.get_by_id(key)
def __setitem__(self, key, value):
self.set_by_id(key, value)
def __delitem__(self, key):
self.delete_by_id(key)
def __contains__(self, key):
try:
self.get_by_id(key)
except self.DoesNotExist:
return False
else:
return True
def __len__(self):
return self.select().count()
def __bool__(self): return True
__nonzero__ = __bool__ # Python 2.
def __sql__(self, ctx):
return ctx.sql(self._meta.table)
class _BoundModelsContext(_callable_context_manager):
def __init__(self, models, database, bind_refs, bind_backrefs):
self.models = models
self.database = database
self.bind_refs = bind_refs
self.bind_backrefs = bind_backrefs
def __enter__(self):
self._orig_database = []
for model in self.models:
self._orig_database.append(model._meta.database)
model.bind(self.database, self.bind_refs, self.bind_backrefs,
_exclude=set(self.models))
return self.models
def __exit__(self, exc_type, exc_val, exc_tb):
for model, db in zip(self.models, self._orig_database):
model.bind(db, self.bind_refs, self.bind_backrefs,
_exclude=set(self.models))
class Model(with_metaclass(ModelBase, Node)):
def __init__(self, *args, **kwargs):
if kwargs.pop('__no_default__', None):
self.__data__ = {}
else:
self.__data__ = self._meta.get_default_dict()
self._dirty = set(self.__data__)
self.__rel__ = {}
for k in kwargs:
setattr(self, k, kwargs[k])
def __str__(self):
return str(self._pk) if self._meta.primary_key is not False else 'n/a'
@classmethod
def validate_model(cls):
pass
@classmethod
def alias(cls, alias=None):
return ModelAlias(cls, alias)
@classmethod
def select(cls, *fields):
is_default = not fields
if not fields:
fields = cls._meta.sorted_fields
return ModelSelect(cls, fields, is_default=is_default)
@classmethod
def _normalize_data(cls, data, kwargs):
normalized = {}
if data:
if not isinstance(data, dict):
if kwargs:
raise ValueError('Data cannot be mixed with keyword '
'arguments: %s' % data)
return data
for key in data:
try:
field = (key if isinstance(key, Field)
else cls._meta.combined[key])
except KeyError:
if not isinstance(key, Node):
raise ValueError('Unrecognized field name: "%s" in %s.'
% (key, data))
field = key
normalized[field] = data[key]
if kwargs:
for key in kwargs:
try:
normalized[cls._meta.combined[key]] = kwargs[key]
except KeyError:
normalized[getattr(cls, key)] = kwargs[key]
return normalized
@classmethod
def update(cls, __data=None, **update):
return ModelUpdate(cls, cls._normalize_data(__data, update))
@classmethod
def insert(cls, __data=None, **insert):
return ModelInsert(cls, cls._normalize_data(__data, insert))
@classmethod
def insert_many(cls, rows, fields=None):
return ModelInsert(cls, insert=rows, columns=fields)
@classmethod
def insert_from(cls, query, fields):
columns = [getattr(cls, field) if isinstance(field, basestring)
else field for field in fields]
return ModelInsert(cls, insert=query, columns=columns)
@classmethod
def replace(cls, __data=None, **insert):
return cls.insert(__data, **insert).on_conflict('REPLACE')
@classmethod
def replace_many(cls, rows, fields=None):
return (cls
.insert_many(rows=rows, fields=fields)
.on_conflict('REPLACE'))
@classmethod
def raw(cls, sql, *params):
return ModelRaw(cls, sql, params)
@classmethod
def delete(cls):
return ModelDelete(cls)
@classmethod
def create(cls, **query):
inst = cls(**query)
inst.save(force_insert=True)
return inst
@classmethod
def bulk_create(cls, model_list, batch_size=None):
if batch_size is not None:
batches = chunked(model_list, batch_size)
else:
batches = [model_list]
field_names = list(cls._meta.sorted_field_names)
if cls._meta.auto_increment:
pk_name = cls._meta.primary_key.name
field_names.remove(pk_name)
if cls._meta.database.returning_clause and \
cls._meta.primary_key is not False:
pk_fields = cls._meta.get_primary_keys()
else:
pk_fields = None
fields = [cls._meta.fields[field_name] for field_name in field_names]
attrs = []
for field in fields:
if isinstance(field, ForeignKeyField):
attrs.append(field.object_id_name)
else:
attrs.append(field.name)
for batch in batches:
accum = ([getattr(model, f) for f in attrs]
for model in batch)
res = cls.insert_many(accum, fields=fields).execute()
if pk_fields and res is not None:
for row, model in zip(res, batch):
for (pk_field, obj_id) in zip(pk_fields, row):
setattr(model, pk_field.name, obj_id)
@classmethod
def bulk_update(cls, model_list, fields, batch_size=None):
if isinstance(cls._meta.primary_key, CompositeKey):
raise ValueError('bulk_update() is not supported for models with '
'a composite primary key.')
# First normalize list of fields so all are field instances.
fields = [cls._meta.fields[f] if isinstance(f, basestring) else f
for f in fields]
# Now collect list of attribute names to use for values.
attrs = [field.object_id_name if isinstance(field, ForeignKeyField)
else field.name for field in fields]
if batch_size is not None:
batches = chunked(model_list, batch_size)
else:
batches = [model_list]
n = 0
pk = cls._meta.primary_key
for batch in batches:
id_list = [model._pk for model in batch]
update = {}
for field, attr in zip(fields, attrs):
accum = []
for model in batch:
value = getattr(model, attr)
if not isinstance(value, Node):
value = field.to_value(value)
accum.append((pk.to_value(model._pk), value))
case = Case(pk, accum)
update[field] = case
n += (cls.update(update)
.where(cls._meta.primary_key.in_(id_list))
.execute())
return n
@classmethod
def noop(cls):
return NoopModelSelect(cls, ())
@classmethod
def get(cls, *query, **filters):
sq = cls.select()
if query:
# Handle simple lookup using just the primary key.
if len(query) == 1 and isinstance(query[0], int):
sq = sq.where(cls._meta.primary_key == query[0])
else:
sq = sq.where(*query)
if filters:
sq = sq.filter(**filters)
return sq.get()
@classmethod
def get_or_none(cls, *query, **filters):
try:
return cls.get(*query, **filters)
except DoesNotExist:
pass
@classmethod
def get_by_id(cls, pk):
return cls.get(cls._meta.primary_key == pk)
@classmethod
def set_by_id(cls, key, value):
if key is None:
return cls.insert(value).execute()
else:
return (cls.update(value)
.where(cls._meta.primary_key == key).execute())
@classmethod
def delete_by_id(cls, pk):
return cls.delete().where(cls._meta.primary_key == pk).execute()
@classmethod
def get_or_create(cls, **kwargs):
defaults = kwargs.pop('defaults', {})
query = cls.select()
for field, value in kwargs.items():
query = query.where(getattr(cls, field) == value)
try:
return query.get(), False
except cls.DoesNotExist:
try:
if defaults:
kwargs.update(defaults)
with cls._meta.database.atomic():
return cls.create(**kwargs), True
except IntegrityError as exc:
try:
return query.get(), False
except cls.DoesNotExist:
raise exc
@classmethod
def filter(cls, *dq_nodes, **filters):
return cls.select().filter(*dq_nodes, **filters)
def get_id(self):
# Using getattr(self, pk-name) could accidentally trigger a query if
# the primary-key is a foreign-key. So we use the safe_name attribute,
# which defaults to the field-name, but will be the object_id_name for
# foreign-key fields.
if self._meta.primary_key is not False:
return getattr(self, self._meta.primary_key.safe_name)
_pk = property(get_id)
@_pk.setter
def _pk(self, value):
setattr(self, self._meta.primary_key.name, value)
def _pk_expr(self):
return self._meta.primary_key == self._pk
def _prune_fields(self, field_dict, only):
new_data = {}
for field in only:
if isinstance(field, basestring):
field = self._meta.combined[field]
if field.name in field_dict:
new_data[field.name] = field_dict[field.name]
return new_data
def _populate_unsaved_relations(self, field_dict):
for foreign_key_field in self._meta.refs:
foreign_key = foreign_key_field.name
conditions = (
foreign_key in field_dict and
field_dict[foreign_key] is None and
self.__rel__.get(foreign_key) is not None)
if conditions:
setattr(self, foreign_key, getattr(self, foreign_key))
field_dict[foreign_key] = self.__data__[foreign_key]
def save(self, force_insert=False, only=None):
field_dict = self.__data__.copy()
if self._meta.primary_key is not False:
pk_field = self._meta.primary_key
pk_value = self._pk
else:
pk_field = pk_value = None
if only is not None:
field_dict = self._prune_fields(field_dict, only)
elif self._meta.only_save_dirty and not force_insert:
field_dict = self._prune_fields(field_dict, self.dirty_fields)
if not field_dict:
self._dirty.clear()
return False
self._populate_unsaved_relations(field_dict)
rows = 1
if self._meta.auto_increment and pk_value is None:
field_dict.pop(pk_field.name, None)
if pk_value is not None and not force_insert:
if self._meta.composite_key:
for pk_part_name in pk_field.field_names:
field_dict.pop(pk_part_name, None)
else:
field_dict.pop(pk_field.name, None)
if not field_dict:
raise ValueError('no data to save!')
rows = self.update(**field_dict).where(self._pk_expr()).execute()
elif pk_field is not None:
pk = self.insert(**field_dict).execute()
if pk is not None and (self._meta.auto_increment or
pk_value is None):
self._pk = pk
# Although we set the primary-key, do not mark it as dirty.
self._dirty.discard(pk_field.name)
else:
self.insert(**field_dict).execute()
self._dirty -= set(field_dict) # Remove any fields we saved.
return rows
def is_dirty(self):
return bool(self._dirty)
@property
def dirty_fields(self):
return [f for f in self._meta.sorted_fields if f.name in self._dirty]
def dependencies(self, search_nullable=False):
model_class = type(self)
stack = [(type(self), None)]
seen = set()
while stack:
klass, query = stack.pop()
if klass in seen:
continue
seen.add(klass)
for fk, rel_model in klass._meta.backrefs.items():
if rel_model is model_class or query is None:
node = (fk == self.__data__[fk.rel_field.name])
else:
node = fk << query
subquery = (rel_model.select(rel_model._meta.primary_key)
.where(node))
if not fk.null or search_nullable:
stack.append((rel_model, subquery))
yield (node, fk)
def delete_instance(self, recursive=False, delete_nullable=False):
if recursive:
dependencies = self.dependencies(delete_nullable)
for query, fk in reversed(list(dependencies)):
model = fk.model
if fk.null and not delete_nullable:
model.update(**{fk.name: None}).where(query).execute()
else:
model.delete().where(query).execute()
return type(self).delete().where(self._pk_expr()).execute()
def __hash__(self):
return hash((self.__class__, self._pk))
def __eq__(self, other):
return (
other.__class__ == self.__class__ and
self._pk is not None and
self._pk == other._pk)
def __ne__(self, other):
return not self == other
def __sql__(self, ctx):
# NOTE: when comparing a foreign-key field whose related-field is not a
# primary-key, then doing an equality test for the foreign-key with a
# model instance will return the wrong value; since we would return
# the primary key for a given model instance.
#
# This checks to see if we have a converter in the scope, and that we
# are converting a foreign-key expression. If so, we hand the model
# instance to the converter rather than blindly grabbing the primary-
# key. In the event the provided converter fails to handle the model
# instance, then we will return the primary-key.
if ctx.state.converter is not None and ctx.state.is_fk_expr:
try:
return ctx.sql(Value(self, converter=ctx.state.converter))
except (TypeError, ValueError):
pass
return ctx.sql(Value(getattr(self, self._meta.primary_key.name),
converter=self._meta.primary_key.db_value))
@classmethod
def bind(cls, database, bind_refs=True, bind_backrefs=True, _exclude=None):
is_different = cls._meta.database is not database
cls._meta.set_database(database)
if bind_refs or bind_backrefs:
if _exclude is None:
_exclude = set()
G = cls._meta.model_graph(refs=bind_refs, backrefs=bind_backrefs)
for _, model, is_backref in G:
if model not in _exclude:
model._meta.set_database(database)
_exclude.add(model)
return is_different
@classmethod
def bind_ctx(cls, database, bind_refs=True, bind_backrefs=True):
return _BoundModelsContext((cls,), database, bind_refs, bind_backrefs)
@classmethod
def table_exists(cls):
M = cls._meta
return cls._schema.database.table_exists(M.table.__name__, M.schema)
@classmethod
def create_table(cls, safe=True, **options):
if 'fail_silently' in options:
__deprecated__('"fail_silently" has been deprecated in favor of '
'"safe" for the create_table() method.')
safe = options.pop('fail_silently')
if safe and not cls._schema.database.safe_create_index \
and cls.table_exists():
return
if cls._meta.temporary:
options.setdefault('temporary', cls._meta.temporary)
cls._schema.create_all(safe, **options)
@classmethod
def drop_table(cls, safe=True, drop_sequences=True, **options):
if safe and not cls._schema.database.safe_drop_index \
and not cls.table_exists():
return
if cls._meta.temporary:
options.setdefault('temporary', cls._meta.temporary)
cls._schema.drop_all(safe, drop_sequences, **options)
@classmethod
def truncate_table(cls, **options):
cls._schema.truncate_table(**options)
@classmethod
def index(cls, *fields, **kwargs):
return ModelIndex(cls, fields, **kwargs)
@classmethod
def add_index(cls, *fields, **kwargs):
if len(fields) == 1 and isinstance(fields[0], (SQL, Index)):
cls._meta.indexes.append(fields[0])
else:
cls._meta.indexes.append(ModelIndex(cls, fields, **kwargs))
class ModelAlias(Node):
"""Provide a separate reference to a model in a query."""
def __init__(self, model, alias=None):
self.__dict__['model'] = model
self.__dict__['alias'] = alias
def __getattr__(self, attr):
# Hack to work-around the fact that properties or other objects
# implementing the descriptor protocol (on the model being aliased),
# will not work correctly when we use getattr(). So we explicitly pass
# the model alias to the descriptor's getter.
try:
obj = self.model.__dict__[attr]
except KeyError:
pass
else:
if isinstance(obj, ModelDescriptor):
return obj.__get__(None, self)
model_attr = getattr(self.model, attr)
if isinstance(model_attr, Field):
self.__dict__[attr] = FieldAlias.create(self, model_attr)
return self.__dict__[attr]
return model_attr
def __setattr__(self, attr, value):
raise AttributeError('Cannot set attributes on model aliases.')
def get_field_aliases(self):
return [getattr(self, n) for n in self.model._meta.sorted_field_names]
def select(self, *selection):
if not selection:
selection = self.get_field_aliases()
return ModelSelect(self, selection)
def __call__(self, **kwargs):
return self.model(**kwargs)
def __sql__(self, ctx):
if ctx.scope == SCOPE_VALUES:
# Return the quoted table name.
return ctx.sql(self.model)
if self.alias:
ctx.alias_manager[self] = self.alias
if ctx.scope == SCOPE_SOURCE:
# Define the table and its alias.
return (ctx
.sql(self.model._meta.entity)
.literal(' AS ')
.sql(Entity(ctx.alias_manager[self])))
else:
# Refer to the table using the alias.
return ctx.sql(Entity(ctx.alias_manager[self]))
class FieldAlias(Field):
def __init__(self, source, field):
self.source = source
self.model = source.model
self.field = field
@classmethod
def create(cls, source, field):
class _FieldAlias(cls, type(field)):
pass
return _FieldAlias(source, field)
def clone(self):
return FieldAlias(self.source, self.field)
def adapt(self, value): return self.field.adapt(value)
def python_value(self, value): return self.field.python_value(value)
def db_value(self, value): return self.field.db_value(value)
def __getattr__(self, attr):
return self.source if attr == 'model' else getattr(self.field, attr)
def __sql__(self, ctx):
return ctx.sql(Column(self.source, self.field.column_name))
def sort_models(models):
models = set(models)
seen = set()
ordering = []
def dfs(model):
if model in models and model not in seen:
seen.add(model)
for foreign_key, rel_model in model._meta.refs.items():
# Do not depth-first search deferred foreign-keys as this can
# cause tables to be created in the incorrect order.
if not foreign_key.deferred:
dfs(rel_model)
if model._meta.depends_on:
for dependency in model._meta.depends_on:
dfs(dependency)
ordering.append(model)
names = lambda m: (m._meta.name, m._meta.table_name)
for m in sorted(models, key=names):
dfs(m)
return ordering
class _ModelQueryHelper(object):
default_row_type = ROW.MODEL
def __init__(self, *args, **kwargs):
super(_ModelQueryHelper, self).__init__(*args, **kwargs)
if not self._database:
self._database = self.model._meta.database
@Node.copy
def objects(self, constructor=None):
self._row_type = ROW.CONSTRUCTOR
self._constructor = self.model if constructor is None else constructor
def _get_cursor_wrapper(self, cursor):
row_type = self._row_type or self.default_row_type
if row_type == ROW.MODEL:
return self._get_model_cursor_wrapper(cursor)
elif row_type == ROW.DICT:
return ModelDictCursorWrapper(cursor, self.model, self._returning)
elif row_type == ROW.TUPLE:
return ModelTupleCursorWrapper(cursor, self.model, self._returning)
elif row_type == ROW.NAMED_TUPLE:
return ModelNamedTupleCursorWrapper(cursor, self.model,
self._returning)
elif row_type == ROW.CONSTRUCTOR:
return ModelObjectCursorWrapper(cursor, self.model,
self._returning, self._constructor)
else:
raise ValueError('Unrecognized row type: "%s".' % row_type)
def _get_model_cursor_wrapper(self, cursor):
return ModelObjectCursorWrapper(cursor, self.model, [], self.model)
class ModelRaw(_ModelQueryHelper, RawQuery):
def __init__(self, model, sql, params, **kwargs):
self.model = model
self._returning = ()
super(ModelRaw, self).__init__(sql=sql, params=params, **kwargs)
def get(self):
try:
return self.execute()[0]
except IndexError:
sql, params = self.sql()
raise self.model.DoesNotExist('%s instance matching query does '
'not exist:\nSQL: %s\nParams: %s' %
(self.model, sql, params))
class BaseModelSelect(_ModelQueryHelper):
def union_all(self, rhs):
return ModelCompoundSelectQuery(self.model, self, 'UNION ALL', rhs)
__add__ = union_all
def union(self, rhs):
return ModelCompoundSelectQuery(self.model, self, 'UNION', rhs)
__or__ = union
def intersect(self, rhs):
return ModelCompoundSelectQuery(self.model, self, 'INTERSECT', rhs)
__and__ = intersect
def except_(self, rhs):
return ModelCompoundSelectQuery(self.model, self, 'EXCEPT', rhs)
__sub__ = except_
def __iter__(self):
if not self._cursor_wrapper:
self.execute()
return iter(self._cursor_wrapper)
def prefetch(self, *subqueries):
return prefetch(self, *subqueries)
def get(self, database=None):
clone = self.paginate(1, 1)
clone._cursor_wrapper = None
try:
return clone.execute(database)[0]
except IndexError:
sql, params = clone.sql()
raise self.model.DoesNotExist('%s instance matching query does '
'not exist:\nSQL: %s\nParams: %s' %
(clone.model, sql, params))
def get_or_none(self, database=None):
try:
return self.get(database=database)
except self.model.DoesNotExist:
pass
@Node.copy
def group_by(self, *columns):
grouping = []
for column in columns:
if is_model(column):
grouping.extend(column._meta.sorted_fields)
elif isinstance(column, Table):
if not column._columns:
raise ValueError('Cannot pass a table to group_by() that '
'does not have columns explicitly '
'declared.')
grouping.extend([getattr(column, col_name)
for col_name in column._columns])
else:
grouping.append(column)
self._group_by = grouping
class ModelCompoundSelectQuery(BaseModelSelect, CompoundSelectQuery):
def __init__(self, model, *args, **kwargs):
self.model = model
super(ModelCompoundSelectQuery, self).__init__(*args, **kwargs)
def _get_model_cursor_wrapper(self, cursor):
return self.lhs._get_model_cursor_wrapper(cursor)
def _normalize_model_select(fields_or_models):
fields = []
for fm in fields_or_models:
if is_model(fm):
fields.extend(fm._meta.sorted_fields)
elif isinstance(fm, ModelAlias):
fields.extend(fm.get_field_aliases())
elif isinstance(fm, Table) and fm._columns:
fields.extend([getattr(fm, col) for col in fm._columns])
else:
fields.append(fm)
return fields
class ModelSelect(BaseModelSelect, Select):
def __init__(self, model, fields_or_models, is_default=False):
self.model = self._join_ctx = model
self._joins = {}
self._is_default = is_default
fields = _normalize_model_select(fields_or_models)
super(ModelSelect, self).__init__([model], fields)
def clone(self):
clone = super(ModelSelect, self).clone()
if clone._joins:
clone._joins = dict(clone._joins)
return clone
def select(self, *fields_or_models):
if fields_or_models or not self._is_default:
self._is_default = False
fields = _normalize_model_select(fields_or_models)
return super(ModelSelect, self).select(*fields)
return self
def switch(self, ctx=None):
self._join_ctx = self.model if ctx is None else ctx
return self
def _get_model(self, src):
if is_model(src):
return src, True
elif isinstance(src, Table) and src._model:
return src._model, False
elif isinstance(src, ModelAlias):
return src.model, False
elif isinstance(src, ModelSelect):
return src.model, False
return None, False
def _normalize_join(self, src, dest, on, attr):
# Allow "on" expression to have an alias that determines the
# destination attribute for the joined data.
on_alias = isinstance(on, Alias)
if on_alias:
attr = attr or on._alias
on = on.alias()
# Obtain references to the source and destination models being joined.
src_model, src_is_model = self._get_model(src)
dest_model, dest_is_model = self._get_model(dest)
if src_model and dest_model:
self._join_ctx = dest
constructor = dest_model
# In the case where the "on" clause is a Column or Field, we will
# convert that field into the appropriate predicate expression.
if not (src_is_model and dest_is_model) and isinstance(on, Column):
if on.source is src:
to_field = src_model._meta.columns[on.name]
elif on.source is dest:
to_field = dest_model._meta.columns[on.name]
else:
raise AttributeError('"on" clause Column %s does not '
'belong to %s or %s.' %
(on, src_model, dest_model))
on = None
elif isinstance(on, Field):
to_field = on
on = None
else:
to_field = None
fk_field, is_backref = self._generate_on_clause(
src_model, dest_model, to_field, on)
if on is None:
src_attr = 'name' if src_is_model else 'column_name'
dest_attr = 'name' if dest_is_model else 'column_name'
if is_backref:
lhs = getattr(dest, getattr(fk_field, dest_attr))
rhs = getattr(src, getattr(fk_field.rel_field, src_attr))
else:
lhs = getattr(src, getattr(fk_field, src_attr))
rhs = getattr(dest, getattr(fk_field.rel_field, dest_attr))
on = (lhs == rhs)
if not attr:
if fk_field is not None and not is_backref:
attr = fk_field.name
else:
attr = dest_model._meta.name
elif on_alias and fk_field is not None and \
attr == fk_field.object_id_name and not is_backref:
raise ValueError('Cannot assign join alias to "%s", as this '
'attribute is the object_id_name for the '
'foreign-key field "%s"' % (attr, fk_field))
elif isinstance(dest, Source):
constructor = dict
attr = attr or dest._alias
if not attr and isinstance(dest, Table):
attr = attr or dest.__name__
return (on, attr, constructor)
def _generate_on_clause(self, src, dest, to_field=None, on=None):
meta = src._meta
is_backref = fk_fields = False
# Get all the foreign keys between source and dest, and determine if
# the join is via a back-reference.
if dest in meta.model_refs:
fk_fields = meta.model_refs[dest]
elif dest in meta.model_backrefs:
fk_fields = meta.model_backrefs[dest]
is_backref = True
if not fk_fields:
if on is not None:
return None, False
raise ValueError('Unable to find foreign key between %s and %s. '
'Please specify an explicit join condition.' %
(src, dest))
elif to_field is not None:
# If the foreign-key field was specified explicitly, remove all
# other foreign-key fields from the list.
target = (to_field.field if isinstance(to_field, FieldAlias)
else to_field)
fk_fields = [f for f in fk_fields if (
(f is target) or
(is_backref and f.rel_field is to_field))]
if len(fk_fields) == 1:
return fk_fields[0], is_backref
if on is None:
# If multiple foreign-keys exist, try using the FK whose name
# matches that of the related model. If not, raise an error as this
# is ambiguous.
for fk in fk_fields:
if fk.name == dest._meta.name:
return fk, is_backref
raise ValueError('More than one foreign key between %s and %s.'
' Please specify which you are joining on.' %
(src, dest))
# If there are multiple foreign-keys to choose from and the join
# predicate is an expression, we'll try to figure out which
# foreign-key field we're joining on so that we can assign to the
# correct attribute when resolving the model graph.
to_field = None
if isinstance(on, Expression):
lhs, rhs = on.lhs, on.rhs
# Coerce to set() so that we force Python to compare using the
# object's hash rather than equality test, which returns a
# false-positive due to overriding __eq__.
fk_set = set(fk_fields)
if isinstance(lhs, Field):
lhs_f = lhs.field if isinstance(lhs, FieldAlias) else lhs
if lhs_f in fk_set:
to_field = lhs_f
elif isinstance(rhs, Field):
rhs_f = rhs.field if isinstance(rhs, FieldAlias) else rhs
if rhs_f in fk_set:
to_field = rhs_f
return to_field, False
@Node.copy
def join(self, dest, join_type=JOIN.INNER, on=None, src=None, attr=None):
src = self._join_ctx if src is None else src
if join_type == JOIN.LATERAL or join_type == JOIN.LEFT_LATERAL:
on = True
elif join_type != JOIN.CROSS:
on, attr, constructor = self._normalize_join(src, dest, on, attr)
if attr:
self._joins.setdefault(src, [])
self._joins[src].append((dest, attr, constructor, join_type))
elif on is not None:
raise ValueError('Cannot specify on clause with cross join.')
if not self._from_list:
raise ValueError('No sources to join on.')
item = self._from_list.pop()
self._from_list.append(Join(item, dest, join_type, on))
def join_from(self, src, dest, join_type=JOIN.INNER, on=None, attr=None):
return self.join(dest, join_type, on, src, attr)
def _get_model_cursor_wrapper(self, cursor):
if len(self._from_list) == 1 and not self._joins:
return ModelObjectCursorWrapper(cursor, self.model,
self._returning, self.model)
return ModelCursorWrapper(cursor, self.model, self._returning,
self._from_list, self._joins)
def ensure_join(self, lm, rm, on=None, **join_kwargs):
join_ctx = self._join_ctx
for dest, _, constructor, _ in self._joins.get(lm, []):
if dest == rm:
return self
return self.switch(lm).join(rm, on=on, **join_kwargs).switch(join_ctx)
def convert_dict_to_node(self, qdict):
accum = []
joins = []
fks = (ForeignKeyField, BackrefAccessor)
for key, value in sorted(qdict.items()):
curr = self.model
if '__' in key and key.rsplit('__', 1)[1] in DJANGO_MAP:
key, op = key.rsplit('__', 1)
op = DJANGO_MAP[op]
elif value is None:
op = DJANGO_MAP['is']
else:
op = DJANGO_MAP['eq']
if '__' not in key:
# Handle simplest case. This avoids joining over-eagerly when a
# direct FK lookup is all that is required.
model_attr = getattr(curr, key)
else:
for piece in key.split('__'):
for dest, attr, _, _ in self._joins.get(curr, ()):
if attr == piece or (isinstance(dest, ModelAlias) and
dest.alias == piece):
curr = dest
break
else:
model_attr = getattr(curr, piece)
if value is not None and isinstance(model_attr, fks):
curr = model_attr.rel_model
joins.append(model_attr)
accum.append(op(model_attr, value))
return accum, joins
def filter(self, *args, **kwargs):
# normalize args and kwargs into a new expression
if args and kwargs:
dq_node = (reduce(operator.and_, [a.clone() for a in args]) &
DQ(**kwargs))
elif args:
dq_node = (reduce(operator.and_, [a.clone() for a in args]) &
ColumnBase())
elif kwargs:
dq_node = DQ(**kwargs) & ColumnBase()
else:
return self.clone()
# dq_node should now be an Expression, lhs = Node(), rhs = ...
q = collections.deque([dq_node])
dq_joins = []
seen_joins = set()
while q:
curr = q.popleft()
if not isinstance(curr, Expression):
continue
for side, piece in (('lhs', curr.lhs), ('rhs', curr.rhs)):
if isinstance(piece, DQ):
query, joins = self.convert_dict_to_node(piece.query)
for join in joins:
if join not in seen_joins:
dq_joins.append(join)
seen_joins.add(join)
expression = reduce(operator.and_, query)
# Apply values from the DQ object.
if piece._negated:
expression = Negated(expression)
#expression._alias = piece._alias
setattr(curr, side, expression)
else:
q.append(piece)
if not args or not kwargs:
dq_node = dq_node.lhs
query = self.clone()
for field in dq_joins:
if isinstance(field, ForeignKeyField):
lm, rm = field.model, field.rel_model
field_obj = field
elif isinstance(field, BackrefAccessor):
lm, rm = field.model, field.rel_model
field_obj = field.field
query = query.ensure_join(lm, rm, field_obj)
return query.where(dq_node)
def create_table(self, name, safe=True, **meta):
return self.model._schema.create_table_as(name, self, safe, **meta)
def __sql_selection__(self, ctx, is_subquery=False):
if self._is_default and is_subquery and len(self._returning) > 1 and \
self.model._meta.primary_key is not False:
return ctx.sql(self.model._meta.primary_key)
return ctx.sql(CommaNodeList(self._returning))
class NoopModelSelect(ModelSelect):
def __sql__(self, ctx):
return self.model._meta.database.get_noop_select(ctx)
def _get_cursor_wrapper(self, cursor):
return CursorWrapper(cursor)
class _ModelWriteQueryHelper(_ModelQueryHelper):
def __init__(self, model, *args, **kwargs):
self.model = model
super(_ModelWriteQueryHelper, self).__init__(model, *args, **kwargs)
def returning(self, *returning):
accum = []
for item in returning:
if is_model(item):
accum.extend(item._meta.sorted_fields)
else:
accum.append(item)
return super(_ModelWriteQueryHelper, self).returning(*accum)
def _set_table_alias(self, ctx):
table = self.model._meta.table
ctx.alias_manager[table] = table.__name__
class ModelUpdate(_ModelWriteQueryHelper, Update):
pass
class ModelInsert(_ModelWriteQueryHelper, Insert):
default_row_type = ROW.TUPLE
def __init__(self, *args, **kwargs):
super(ModelInsert, self).__init__(*args, **kwargs)
if self._returning is None and self.model._meta.database is not None:
if self.model._meta.database.returning_clause:
self._returning = self.model._meta.get_primary_keys()
def returning(self, *returning):
# By default ModelInsert will yield a `tuple` containing the
# primary-key of the newly inserted row. But if we are explicitly
# specifying a returning clause and have not set a row type, we will
# default to returning model instances instead.
if returning and self._row_type is None:
self._row_type = ROW.MODEL
return super(ModelInsert, self).returning(*returning)
def get_default_data(self):
return self.model._meta.defaults
def get_default_columns(self):
fields = self.model._meta.sorted_fields
return fields[1:] if self.model._meta.auto_increment else fields
class ModelDelete(_ModelWriteQueryHelper, Delete):
pass
class ManyToManyQuery(ModelSelect):
def __init__(self, instance, accessor, rel, *args, **kwargs):
self._instance = instance
self._accessor = accessor
self._src_attr = accessor.src_fk.rel_field.name
self._dest_attr = accessor.dest_fk.rel_field.name
super(ManyToManyQuery, self).__init__(rel, (rel,), *args, **kwargs)
def _id_list(self, model_or_id_list):
if isinstance(model_or_id_list[0], Model):
return [getattr(obj, self._dest_attr) for obj in model_or_id_list]
return model_or_id_list
def add(self, value, clear_existing=False):
if clear_existing:
self.clear()
accessor = self._accessor
src_id = getattr(self._instance, self._src_attr)
if isinstance(value, SelectQuery):
query = value.columns(
Value(src_id),
accessor.dest_fk.rel_field)
accessor.through_model.insert_from(
fields=[accessor.src_fk, accessor.dest_fk],
query=query).execute()
else:
value = ensure_tuple(value)
if not value: return
inserts = [{
accessor.src_fk.name: src_id,
accessor.dest_fk.name: rel_id}
for rel_id in self._id_list(value)]
accessor.through_model.insert_many(inserts).execute()
def remove(self, value):
src_id = getattr(self._instance, self._src_attr)
if isinstance(value, SelectQuery):
column = getattr(value.model, self._dest_attr)
subquery = value.columns(column)
return (self._accessor.through_model
.delete()
.where(
(self._accessor.dest_fk << subquery) &
(self._accessor.src_fk == src_id))
.execute())
else:
value = ensure_tuple(value)
if not value:
return
return (self._accessor.through_model
.delete()
.where(
(self._accessor.dest_fk << self._id_list(value)) &
(self._accessor.src_fk == src_id))
.execute())
def clear(self):
src_id = getattr(self._instance, self._src_attr)
return (self._accessor.through_model
.delete()
.where(self._accessor.src_fk == src_id)
.execute())
def safe_python_value(conv_func):
def validate(value):
try:
return conv_func(value)
except (TypeError, ValueError):
return value
return validate
class BaseModelCursorWrapper(DictCursorWrapper):
def __init__(self, cursor, model, columns):
super(BaseModelCursorWrapper, self).__init__(cursor)
self.model = model
self.select = columns or []
def _initialize_columns(self):
combined = self.model._meta.combined
table = self.model._meta.table
description = self.cursor.description
self.ncols = len(self.cursor.description)
self.columns = []
self.converters = converters = [None] * self.ncols
self.fields = fields = [None] * self.ncols
for idx, description_item in enumerate(description):
column = orig_column = description_item[0]
# Try to clean-up messy column descriptions when people do not
# provide an alias. The idea is that we take something like:
# SUM("t1"."price") -> "price") -> price
dot_index = column.rfind('.')
if dot_index != -1:
column = column[dot_index + 1:]
column = column.strip('()"`')
self.columns.append(column)
# Now we'll see what they selected and see if we can improve the
# column-name being returned - e.g. by mapping it to the selected
# field's name.
try:
raw_node = self.select[idx]
except IndexError:
if column in combined:
raw_node = node = combined[column]
else:
continue
else:
node = raw_node.unwrap()
# If this column was given an alias, then we will use whatever
# alias was returned by the cursor.
is_alias = raw_node.is_alias()
if is_alias:
self.columns[idx] = orig_column
# Heuristics used to attempt to get the field associated with a
# given SELECT column, so that we can accurately convert the value
# returned by the database-cursor into a Python object.
if isinstance(node, Field):
if raw_node._coerce:
converters[idx] = node.python_value
fields[idx] = node
if not is_alias:
self.columns[idx] = node.name
elif isinstance(node, ColumnBase) and raw_node._converter:
converters[idx] = raw_node._converter
elif isinstance(node, Function) and node._coerce:
if node._python_value is not None:
converters[idx] = node._python_value
elif node.arguments and isinstance(node.arguments[0], Node):
# If the first argument is a field or references a column
# on a Model, try using that field's conversion function.
# This usually works, but we use "safe_python_value()" so
# that if a TypeError or ValueError occurs during
# conversion we can just fall-back to the raw cursor value.
first = node.arguments[0].unwrap()
if isinstance(first, Entity):
path = first._path[-1] # Try to look-up by name.
first = combined.get(path)
if isinstance(first, Field):
converters[idx] = safe_python_value(first.python_value)
elif column in combined:
if node._coerce:
converters[idx] = combined[column].python_value
if isinstance(node, Column) and node.source == table:
fields[idx] = combined[column]
initialize = _initialize_columns
def process_row(self, row):
raise NotImplementedError
class ModelDictCursorWrapper(BaseModelCursorWrapper):
def process_row(self, row):
result = {}
columns, converters = self.columns, self.converters
fields = self.fields
for i in range(self.ncols):
attr = columns[i]
if attr in result: continue # Don't overwrite if we have dupes.
if converters[i] is not None:
result[attr] = converters[i](row[i])
else:
result[attr] = row[i]
return result
class ModelTupleCursorWrapper(ModelDictCursorWrapper):
constructor = tuple
def process_row(self, row):
columns, converters = self.columns, self.converters
return self.constructor([
(converters[i](row[i]) if converters[i] is not None else row[i])
for i in range(self.ncols)])
class ModelNamedTupleCursorWrapper(ModelTupleCursorWrapper):
def initialize(self):
self._initialize_columns()
attributes = []
for i in range(self.ncols):
attributes.append(self.columns[i])
self.tuple_class = collections.namedtuple('Row', attributes)
self.constructor = lambda row: self.tuple_class(*row)
class ModelObjectCursorWrapper(ModelDictCursorWrapper):
def __init__(self, cursor, model, select, constructor):
self.constructor = constructor
self.is_model = is_model(constructor)
super(ModelObjectCursorWrapper, self).__init__(cursor, model, select)
def process_row(self, row):
data = super(ModelObjectCursorWrapper, self).process_row(row)
if self.is_model:
# Clear out any dirty fields before returning to the user.
obj = self.constructor(__no_default__=1, **data)
obj._dirty.clear()
return obj
else:
return self.constructor(**data)
class ModelCursorWrapper(BaseModelCursorWrapper):
def __init__(self, cursor, model, select, from_list, joins):
super(ModelCursorWrapper, self).__init__(cursor, model, select)
self.from_list = from_list
self.joins = joins
def initialize(self):
self._initialize_columns()
selected_src = set([field.model for field in self.fields
if field is not None])
select, columns = self.select, self.columns
self.key_to_constructor = {self.model: self.model}
self.src_is_dest = {}
self.src_to_dest = []
accum = collections.deque(self.from_list)
dests = set()
while accum:
curr = accum.popleft()
if isinstance(curr, Join):
accum.append(curr.lhs)
accum.append(curr.rhs)
continue
if curr not in self.joins:
continue
is_dict = isinstance(curr, dict)
for key, attr, constructor, join_type in self.joins[curr]:
if key not in self.key_to_constructor:
self.key_to_constructor[key] = constructor
# (src, attr, dest, is_dict, join_type).
self.src_to_dest.append((curr, attr, key, is_dict,
join_type))
dests.add(key)
accum.append(key)
# Ensure that we accommodate everything selected.
for src in selected_src:
if src not in self.key_to_constructor:
if is_model(src):
self.key_to_constructor[src] = src
elif isinstance(src, ModelAlias):
self.key_to_constructor[src] = src.model
# Indicate which sources are also dests.
for src, _, dest, _, _ in self.src_to_dest:
self.src_is_dest[src] = src in dests and (dest in selected_src
or src in selected_src)
self.column_keys = []
for idx, node in enumerate(select):
key = self.model
field = self.fields[idx]
if field is not None:
if isinstance(field, FieldAlias):
key = field.source
else:
key = field.model
else:
if isinstance(node, Node):
node = node.unwrap()
if isinstance(node, Column):
key = node.source
self.column_keys.append(key)
def process_row(self, row):
objects = {}
object_list = []
for key, constructor in self.key_to_constructor.items():
objects[key] = constructor(__no_default__=True)
object_list.append(objects[key])
default_instance = objects[self.model]
set_keys = set()
for idx, key in enumerate(self.column_keys):
# Get the instance corresponding to the selected column/value,
# falling back to the "root" model instance.
instance = objects.get(key, default_instance)
column = self.columns[idx]
value = row[idx]
if value is not None:
set_keys.add(key)
if self.converters[idx]:
value = self.converters[idx](value)
if isinstance(instance, dict):
instance[column] = value
else:
setattr(instance, column, value)
# Need to do some analysis on the joins before this.
for (src, attr, dest, is_dict, join_type) in self.src_to_dest:
instance = objects[src]
try:
joined_instance = objects[dest]
except KeyError:
continue
# If no fields were set on the destination instance then do not
# assign an "empty" instance.
if instance is None or dest is None or \
(dest not in set_keys and not self.src_is_dest.get(dest)):
continue
# If no fields were set on either the source or the destination,
# then we have nothing to do here.
if instance not in set_keys and dest not in set_keys \
and join_type.endswith('OUTER JOIN'):
continue
if is_dict:
instance[attr] = joined_instance
else:
setattr(instance, attr, joined_instance)
# When instantiating models from a cursor, we clear the dirty fields.
for instance in object_list:
if isinstance(instance, Model):
instance._dirty.clear()
return objects[self.model]
class PrefetchQuery(collections.namedtuple('_PrefetchQuery', (
'query', 'fields', 'is_backref', 'rel_models', 'field_to_name', 'model'))):
def __new__(cls, query, fields=None, is_backref=None, rel_models=None,
field_to_name=None, model=None):
if fields:
if is_backref:
if rel_models is None:
rel_models = [field.model for field in fields]
foreign_key_attrs = [field.rel_field.name for field in fields]
else:
if rel_models is None:
rel_models = [field.rel_model for field in fields]
foreign_key_attrs = [field.name for field in fields]
field_to_name = list(zip(fields, foreign_key_attrs))
model = query.model
return super(PrefetchQuery, cls).__new__(
cls, query, fields, is_backref, rel_models, field_to_name, model)
def populate_instance(self, instance, id_map):
if self.is_backref:
for field in self.fields:
identifier = instance.__data__[field.name]
key = (field, identifier)
if key in id_map:
setattr(instance, field.name, id_map[key])
else:
for field, attname in self.field_to_name:
identifier = instance.__data__[field.rel_field.name]
key = (field, identifier)
rel_instances = id_map.get(key, [])
for inst in rel_instances:
setattr(inst, attname, instance)
inst._dirty.clear()
setattr(instance, field.backref, rel_instances)
def store_instance(self, instance, id_map):
for field, attname in self.field_to_name:
identity = field.rel_field.python_value(instance.__data__[attname])
key = (field, identity)
if self.is_backref:
id_map[key] = instance
else:
id_map.setdefault(key, [])
id_map[key].append(instance)
def prefetch_add_subquery(sq, subqueries):
fixed_queries = [PrefetchQuery(sq)]
for i, subquery in enumerate(subqueries):
if isinstance(subquery, tuple):
subquery, target_model = subquery
else:
target_model = None
if not isinstance(subquery, Query) and is_model(subquery) or \
isinstance(subquery, ModelAlias):
subquery = subquery.select()
subquery_model = subquery.model
fks = backrefs = None
for j in reversed(range(i + 1)):
fixed = fixed_queries[j]
last_query = fixed.query
last_model = last_obj = fixed.model
if isinstance(last_model, ModelAlias):
last_model = last_model.model
rels = subquery_model._meta.model_refs.get(last_model, [])
if rels:
fks = [getattr(subquery_model, fk.name) for fk in rels]
pks = [getattr(last_obj, fk.rel_field.name) for fk in rels]
else:
backrefs = subquery_model._meta.model_backrefs.get(last_model)
if (fks or backrefs) and ((target_model is last_obj) or
(target_model is None)):
break
if not fks and not backrefs:
tgt_err = ' using %s' % target_model if target_model else ''
raise AttributeError('Error: unable to find foreign key for '
'query: %s%s' % (subquery, tgt_err))
dest = (target_model,) if target_model else None
if fks:
expr = reduce(operator.or_, [
(fk << last_query.select(pk))
for (fk, pk) in zip(fks, pks)])
subquery = subquery.where(expr)
fixed_queries.append(PrefetchQuery(subquery, fks, False, dest))
elif backrefs:
expressions = []
for backref in backrefs:
rel_field = getattr(subquery_model, backref.rel_field.name)
fk_field = getattr(last_obj, backref.name)
expressions.append(rel_field << last_query.select(fk_field))
subquery = subquery.where(reduce(operator.or_, expressions))
fixed_queries.append(PrefetchQuery(subquery, backrefs, True, dest))
return fixed_queries
def prefetch(sq, *subqueries):
if not subqueries:
return sq
fixed_queries = prefetch_add_subquery(sq, subqueries)
deps = {}
rel_map = {}
for pq in reversed(fixed_queries):
query_model = pq.model
if pq.fields:
for rel_model in pq.rel_models:
rel_map.setdefault(rel_model, [])
rel_map[rel_model].append(pq)
deps.setdefault(query_model, {})
id_map = deps[query_model]
has_relations = bool(rel_map.get(query_model))
for instance in pq.query:
if pq.fields:
pq.store_instance(instance, id_map)
if has_relations:
for rel in rel_map[query_model]:
rel.populate_instance(instance, deps[rel.model])
return list(pq.query)