• Language: en

Source code for django.db.models.query

"""
The main QuerySet implementation. This provides the public API for the ORM.
"""

import copy
import operator
import sys
import warnings
from collections import OrderedDict, namedtuple
from functools import lru_cache
from itertools import chain

from django.conf import settings
from django.core import exceptions
from django.db import (
    DJANGO_VERSION_PICKLE_KEY, IntegrityError, connections, router,
    transaction,
)
from django.db.models import DateField, DateTimeField, sql
from django.db.models.constants import LOOKUP_SEP
from django.db.models.deletion import Collector
from django.db.models.expressions import F
from django.db.models.fields import AutoField
from django.db.models.functions import Trunc
from django.db.models.query_utils import FilteredRelation, InvalidQuery, Q
from django.db.models.sql.constants import CURSOR, GET_ITERATOR_CHUNK_SIZE
from django.utils import timezone
from django.utils.deprecation import RemovedInDjango30Warning
from django.utils.functional import cached_property, partition
from django.utils.version import get_version

# The maximum number of items to display in a QuerySet.__repr__
REPR_OUTPUT_SIZE = 20

# Pull into this namespace for backwards compatibility.
EmptyResultSet = sql.EmptyResultSet


class BaseIterable:
    def __init__(self, queryset, chunked_fetch=False, chunk_size=GET_ITERATOR_CHUNK_SIZE):
        self.queryset = queryset
        self.chunked_fetch = chunked_fetch
        self.chunk_size = chunk_size


class ModelIterable(BaseIterable):
    """Iterable that yields a model instance for each row."""

    def __iter__(self):
        queryset = self.queryset
        db = queryset.db
        compiler = queryset.query.get_compiler(using=db)
        # Execute the query. This will also fill compiler.select, klass_info,
        # and annotations.
        results = compiler.execute_sql(chunked_fetch=self.chunked_fetch, chunk_size=self.chunk_size)
        select, klass_info, annotation_col_map = (compiler.select, compiler.klass_info,
                                                  compiler.annotation_col_map)
        model_cls = klass_info['model']
        select_fields = klass_info['select_fields']
        model_fields_start, model_fields_end = select_fields[0], select_fields[-1] + 1
        init_list = [f[0].target.attname
                     for f in select[model_fields_start:model_fields_end]]
        related_populators = get_related_populators(klass_info, select, db)
        for row in compiler.results_iter(results):
            obj = model_cls.from_db(db, init_list, row[model_fields_start:model_fields_end])
            if related_populators:
                for rel_populator in related_populators:
                    rel_populator.populate(row, obj)
            if annotation_col_map:
                for attr_name, col_pos in annotation_col_map.items():
                    setattr(obj, attr_name, row[col_pos])

            # Add the known related objects to the model, if there are any
            if queryset._known_related_objects:
                for field, rel_objs in queryset._known_related_objects.items():
                    # Avoid overwriting objects loaded e.g. by select_related
                    if field.is_cached(obj):
                        continue
                    pk = getattr(obj, field.get_attname())
                    try:
                        rel_obj = rel_objs[pk]
                    except KeyError:
                        pass  # may happen in qs1 | qs2 scenarios
                    else:
                        setattr(obj, field.name, rel_obj)

            yield obj


class ValuesIterable(BaseIterable):
    """
    Iterable returned by QuerySet.values() that yields a dict for each row.
    """

    def __iter__(self):
        queryset = self.queryset
        query = queryset.query
        compiler = query.get_compiler(queryset.db)

        field_names = list(query.values_select)
        extra_names = list(query.extra_select)
        annotation_names = list(query.annotation_select)

        # extra(select=...) cols are always at the start of the row.
        names = extra_names + field_names + annotation_names

        indexes = range(len(names))
        for row in compiler.results_iter():
            yield {names[i]: row[i] for i in indexes}


class ValuesListIterable(BaseIterable):
    """
    Iterable returned by QuerySet.values_list(flat=False) that yields a tuple
    for each row.
    """

    def __iter__(self):
        queryset = self.queryset
        query = queryset.query
        compiler = query.get_compiler(queryset.db)

        if queryset._fields:
            field_names = list(query.values_select)
            extra_names = list(query.extra_select)
            annotation_names = list(query.annotation_select)

            # extra(select=...) cols are always at the start of the row.
            names = extra_names + field_names + annotation_names

            fields = list(queryset._fields) + [f for f in annotation_names if f not in queryset._fields]
            if fields != names:
                # Reorder according to fields.
                index_map = {name: idx for idx, name in enumerate(names)}
                rowfactory = operator.itemgetter(*[index_map[f] for f in fields])
                return map(rowfactory, compiler.results_iter())
        return compiler.results_iter(tuple_expected=True)


class NamedValuesListIterable(ValuesListIterable):
    """
    Iterable returned by QuerySet.values_list(named=True) that yields a
    namedtuple for each row.
    """

    @staticmethod
    @lru_cache()
    def create_namedtuple_class(*names):
        # Cache namedtuple() with @lru_cache() since it's too slow to be
        # called for every QuerySet evaluation.
        return namedtuple('Row', names)

    def __iter__(self):
        queryset = self.queryset
        if queryset._fields:
            names = queryset._fields
        else:
            query = queryset.query
            names = list(query.extra_select)
            names.extend(query.values_select)
            names.extend(query.annotation_select)
        tuple_class = self.create_namedtuple_class(*names)
        new = tuple.__new__
        for row in super().__iter__():
            yield new(tuple_class, row)


class FlatValuesListIterable(BaseIterable):
    """
    Iterable returned by QuerySet.values_list(flat=True) that yields single
    values.
    """

    def __iter__(self):
        queryset = self.queryset
        compiler = queryset.query.get_compiler(queryset.db)
        return chain.from_iterable(compiler.results_iter())


[docs]class QuerySet: """Represent a lazy database lookup for a set of objects.""" def __init__(self, model=None, query=None, using=None, hints=None): self.model = model self._db = using self._hints = hints or {} self.query = query or sql.Query(self.model) self._result_cache = None self._sticky_filter = False self._for_write = False self._prefetch_related_lookups = () self._prefetch_done = False self._known_related_objects = {} # {rel_field: {pk: rel_obj}} self._iterable_class = ModelIterable self._fields = None def as_manager(cls): # Address the circular dependency between `Queryset` and `Manager`. from django.db.models.manager import Manager manager = Manager.from_queryset(cls)() manager._built_with_as_manager = True return manager as_manager.queryset_only = True as_manager = classmethod(as_manager) ######################## # PYTHON MAGIC METHODS # ######################## def __deepcopy__(self, memo): """Don't populate the QuerySet's cache.""" obj = self.__class__() for k, v in self.__dict__.items(): if k == '_result_cache': obj.__dict__[k] = None else: obj.__dict__[k] = copy.deepcopy(v, memo) return obj def __getstate__(self): # Force the cache to be fully populated. self._fetch_all() obj_dict = self.__dict__.copy() obj_dict[DJANGO_VERSION_PICKLE_KEY] = get_version() return obj_dict def __setstate__(self, state): msg = None pickled_version = state.get(DJANGO_VERSION_PICKLE_KEY) if pickled_version: current_version = get_version() if current_version != pickled_version: msg = ( "Pickled queryset instance's Django version %s does not " "match the current version %s." % (pickled_version, current_version) ) else: msg = "Pickled queryset instance's Django version is not specified." if msg: warnings.warn(msg, RuntimeWarning, stacklevel=2) self.__dict__.update(state) def __repr__(self): data = list(self[:REPR_OUTPUT_SIZE + 1]) if len(data) > REPR_OUTPUT_SIZE: data[-1] = "...(remaining elements truncated)..." return '<%s %r>' % (self.__class__.__name__, data) def __len__(self): self._fetch_all() return len(self._result_cache) def __iter__(self): """ The queryset iterator protocol uses three nested iterators in the default case: 1. sql.compiler.execute_sql() - Returns 100 rows at time (constants.GET_ITERATOR_CHUNK_SIZE) using cursor.fetchmany(). This part is responsible for doing some column masking, and returning the rows in chunks. 2. sql.compiler.results_iter() - Returns one row at time. At this point the rows are still just tuples. In some cases the return values are converted to Python values at this location. 3. self.iterator() - Responsible for turning the rows into model objects. """ self._fetch_all() return iter(self._result_cache) def __bool__(self): self._fetch_all() return bool(self._result_cache) def __getitem__(self, k): """Retrieve an item or slice from the set of results.""" if not isinstance(k, (int, slice)): raise TypeError assert ((not isinstance(k, slice) and (k >= 0)) or (isinstance(k, slice) and (k.start is None or k.start >= 0) and (k.stop is None or k.stop >= 0))), \ "Negative indexing is not supported." if self._result_cache is not None: return self._result_cache[k] if isinstance(k, slice): qs = self._chain() if k.start is not None: start = int(k.start) else: start = None if k.stop is not None: stop = int(k.stop) else: stop = None qs.query.set_limits(start, stop) return list(qs)[::k.step] if k.step else qs qs = self._chain() qs.query.set_limits(k, k + 1) qs._fetch_all() return qs._result_cache[0] def __and__(self, other): self._merge_sanity_check(other) if isinstance(other, EmptyQuerySet): return other if isinstance(self, EmptyQuerySet): return self combined = self._chain() combined._merge_known_related_objects(other) combined.query.combine(other.query, sql.AND) return combined def __or__(self, other): self._merge_sanity_check(other) if isinstance(self, EmptyQuerySet): return other if isinstance(other, EmptyQuerySet): return self combined = self._chain() combined._merge_known_related_objects(other) combined.query.combine(other.query, sql.OR) return combined #################################### # METHODS THAT DO DATABASE QUERIES # #################################### def _iterator(self, use_chunked_fetch, chunk_size): yield from self._iterable_class(self, chunked_fetch=use_chunked_fetch, chunk_size=chunk_size) def iterator(self, chunk_size=2000): """ An iterator over the results from applying this QuerySet to the database. """ if chunk_size <= 0: raise ValueError('Chunk size must be strictly positive.') use_chunked_fetch = not connections[self.db].settings_dict.get('DISABLE_SERVER_SIDE_CURSORS') return self._iterator(use_chunked_fetch, chunk_size) def aggregate(self, *args, **kwargs): """ Return a dictionary containing the calculations (aggregation) over the current queryset. If args is present the expression is passed as a kwarg using the Aggregate object's default alias. """ if self.query.distinct_fields: raise NotImplementedError("aggregate() + distinct(fields) not implemented.") self._validate_values_are_expressions(args + tuple(kwargs.values()), method_name='aggregate') for arg in args: # The default_alias property raises TypeError if default_alias # can't be set automatically or AttributeError if it isn't an # attribute. try: arg.default_alias except (AttributeError, TypeError): raise TypeError("Complex aggregates require an alias") kwargs[arg.default_alias] = arg query = self.query.chain() for (alias, aggregate_expr) in kwargs.items(): query.add_annotation(aggregate_expr, alias, is_summary=True) if not query.annotations[alias].contains_aggregate: raise TypeError("%s is not an aggregate expression" % alias) return query.get_aggregation(self.db, kwargs) def count(self): """ Perform a SELECT COUNT() and return the number of records as an integer. If the QuerySet is already fully cached, return the length of the cached results set to avoid multiple SELECT COUNT(*) calls. """ if self._result_cache is not None: return len(self._result_cache) return self.query.get_count(using=self.db) def get(self, *args, **kwargs): """ Perform the query and return a single object matching the given keyword arguments. """ clone = self.filter(*args, **kwargs) if self.query.can_filter() and not self.query.distinct_fields: clone = clone.order_by() num = len(clone) if num == 1: return clone._result_cache[0] if not num: raise self.model.DoesNotExist( "%s matching query does not exist." % self.model._meta.object_name ) raise self.model.MultipleObjectsReturned( "get() returned more than one %s -- it returned %s!" % (self.model._meta.object_name, num) ) def create(self, **kwargs): """ Create a new object with the given kwargs, saving it to the database and returning the created object. """ obj = self.model(**kwargs) self._for_write = True obj.save(force_insert=True, using=self.db) return obj def _populate_pk_values(self, objs): for obj in objs: if obj.pk is None: obj.pk = obj._meta.pk.get_pk_value_on_save(obj) def bulk_create(self, objs, batch_size=None): """ Insert each of the instances into the database. Do *not* call save() on each of the instances, do not send any pre/post_save signals, and do not set the primary key attribute if it is an autoincrement field (except if features.can_return_ids_from_bulk_insert=True). Multi-table models are not supported. """ # When you bulk insert you don't get the primary keys back (if it's an # autoincrement, except if can_return_ids_from_bulk_insert=True), so # you can't insert into the child tables which references this. There # are two workarounds: # 1) This could be implemented if you didn't have an autoincrement pk # 2) You could do it by doing O(n) normal inserts into the parent # tables to get the primary keys back and then doing a single bulk # insert into the childmost table. # We currently set the primary keys on the objects when using # PostgreSQL via the RETURNING ID clause. It should be possible for # Oracle as well, but the semantics for extracting the primary keys is # trickier so it's not done yet. assert batch_size is None or batch_size > 0 # Check that the parents share the same concrete model with the our # model to detect the inheritance pattern ConcreteGrandParent -> # MultiTableParent -> ProxyChild. Simply checking self.model._meta.proxy # would not identify that case as involving multiple tables. for parent in self.model._meta.get_parent_list(): if parent._meta.concrete_model is not self.model._meta.concrete_model: raise ValueError("Can't bulk create a multi-table inherited model") if not objs: return objs self._for_write = True connection = connections[self.db] fields = self.model._meta.concrete_fields objs = list(objs) self._populate_pk_values(objs) with transaction.atomic(using=self.db, savepoint=False): objs_with_pk, objs_without_pk = partition(lambda o: o.pk is None, objs) if objs_with_pk: self._batched_insert(objs_with_pk, fields, batch_size) if objs_without_pk: fields = [f for f in fields if not isinstance(f, AutoField)] ids = self._batched_insert(objs_without_pk, fields, batch_size) if connection.features.can_return_ids_from_bulk_insert: assert len(ids) == len(objs_without_pk) for obj_without_pk, pk in zip(objs_without_pk, ids): obj_without_pk.pk = pk obj_without_pk._state.adding = False obj_without_pk._state.db = self.db return objs def get_or_create(self, defaults=None, **kwargs): """ Look up an object with the given kwargs, creating one if necessary. Return a tuple of (object, created), where created is a boolean specifying whether an object was created. """ lookup, params = self._extract_model_params(defaults, **kwargs) # The get() needs to be targeted at the write database in order # to avoid potential transaction consistency problems. self._for_write = True try: return self.get(**lookup), False except self.model.DoesNotExist: return self._create_object_from_params(lookup, params) def update_or_create(self, defaults=None, **kwargs): """ Look up an object with the given kwargs, updating one with defaults if it exists, otherwise create a new one. Return a tuple (object, created), where created is a boolean specifying whether an object was created. """ defaults = defaults or {} lookup, params = self._extract_model_params(defaults, **kwargs) self._for_write = True with transaction.atomic(using=self.db): try: obj = self.select_for_update().get(**lookup) except self.model.DoesNotExist: obj, created = self._create_object_from_params(lookup, params) if created: return obj, created for k, v in defaults.items(): setattr(obj, k, v() if callable(v) else v) obj.save(using=self.db) return obj, False def _create_object_from_params(self, lookup, params): """ Try to create an object using passed params. Used by get_or_create() and update_or_create(). """ try: with transaction.atomic(using=self.db): params = {k: v() if callable(v) else v for k, v in params.items()} obj = self.create(**params) return obj, True except IntegrityError: exc_info = sys.exc_info() try: return self.get(**lookup), False except self.model.DoesNotExist: pass raise exc_info[0](exc_info[1]).with_traceback(exc_info[2]) def _extract_model_params(self, defaults, **kwargs): """ Prepare `lookup` (kwargs that are valid model attributes), `params` (for creating a model instance) based on given kwargs; for use by get_or_create() and update_or_create(). """ defaults = defaults or {} lookup = kwargs.copy() for f in self.model._meta.fields: if f.attname in lookup: lookup[f.name] = lookup.pop(f.attname) params = {k: v for k, v in kwargs.items() if LOOKUP_SEP not in k} params.update(defaults) property_names = self.model._meta._property_names invalid_params = [] for param in params: try: self.model._meta.get_field(param) except exceptions.FieldDoesNotExist: # It's okay to use a model's property if it has a setter. if not (param in property_names and getattr(self.model, param).fset): invalid_params.append(param) if invalid_params: raise exceptions.FieldError( "Invalid field name(s) for model %s: '%s'." % ( self.model._meta.object_name, "', '".join(sorted(invalid_params)), )) return lookup, params def _earliest_or_latest(self, *fields, field_name=None): """ Return the latest object, according to the model's 'get_latest_by' option or optional given field_name. """ if fields and field_name is not None: raise ValueError('Cannot use both positional arguments and the field_name keyword argument.') order_by = None if field_name is not None: warnings.warn( 'The field_name keyword argument to earliest() and latest() ' 'is deprecated in favor of passing positional arguments.', RemovedInDjango30Warning, ) order_by = (field_name,) elif fields: order_by = fields else: order_by = getattr(self.model._meta, 'get_latest_by') if order_by and not isinstance(order_by, (tuple, list)): order_by = (order_by,) if order_by is None: raise ValueError( "earliest() and latest() require either fields as positional " "arguments or 'get_latest_by' in the model's Meta." ) assert self.query.can_filter(), \ "Cannot change a query once a slice has been taken." obj = self._chain() obj.query.set_limits(high=1) obj.query.clear_ordering(force_empty=True) obj.query.add_ordering(*order_by) return obj.get() def earliest(self, *fields, field_name=None): return self._earliest_or_latest(*fields, field_name=field_name) def latest(self, *fields, field_name=None): return self.reverse()._earliest_or_latest(*fields, field_name=field_name) def first(self): """Return the first object of a query or None if no match is found.""" for obj in (self if self.ordered else self.order_by('pk'))[:1]: return obj def last(self): """Return the last object of a query or None if no match is found.""" for obj in (self.reverse() if self.ordered else self.order_by('-pk'))[:1]: return obj def in_bulk(self, id_list=None, *, field_name='pk'): """ Return a dictionary mapping each of the given IDs to the object with that ID. If `id_list` isn't provided, evaluate the entire QuerySet. """ assert self.query.can_filter(), \ "Cannot use 'limit' or 'offset' with in_bulk" if field_name != 'pk' and not self.model._meta.get_field(field_name).unique: raise ValueError("in_bulk()'s field_name must be a unique field but %r isn't." % field_name) if id_list is not None: if not id_list: return {} filter_key = '{}__in'.format(field_name) batch_size = connections[self.db].features.max_query_params id_list = tuple(id_list) # If the database has a limit on the number of query parameters # (e.g. SQLite), retrieve objects in batches if necessary. if batch_size and batch_size < len(id_list): qs = () for offset in range(0, len(id_list), batch_size): batch = id_list[offset:offset + batch_size] qs += tuple(self.filter(**{filter_key: batch}).order_by()) else: qs = self.filter(**{filter_key: id_list}).order_by() else: qs = self._chain() return {getattr(obj, field_name): obj for obj in qs} def delete(self): """Delete the records in the current QuerySet.""" assert self.query.can_filter(), \ "Cannot use 'limit' or 'offset' with delete." if self._fields is not None: raise TypeError("Cannot call delete() after .values() or .values_list()") del_query = self._chain() # The delete is actually 2 queries - one to find related objects, # and one to delete. Make sure that the discovery of related # objects is performed on the same database as the deletion. del_query._for_write = True # Disable non-supported fields. del_query.query.select_for_update = False del_query.query.select_related = False del_query.query.clear_ordering(force_empty=True) collector = Collector(using=del_query.db) collector.collect(del_query) deleted, _rows_count = collector.delete() # Clear the result cache, in case this QuerySet gets reused. self._result_cache = None return deleted, _rows_count delete.alters_data = True delete.queryset_only = True def _raw_delete(self, using): """ Delete objects found from the given queryset in single direct SQL query. No signals are sent and there is no protection for cascades. """ return sql.DeleteQuery(self.model).delete_qs(self, using) _raw_delete.alters_data = True def update(self, **kwargs): """ Update all elements in the current QuerySet, setting all the given fields to the appropriate values. """ assert self.query.can_filter(), \ "Cannot update a query once a slice has been taken." self._for_write = True query = self.query.chain(sql.UpdateQuery) query.add_update_values(kwargs) # Clear any annotations so that they won't be present in subqueries. query._annotations = None with transaction.atomic(using=self.db, savepoint=False): rows = query.get_compiler(self.db).execute_sql(CURSOR) self._result_cache = None return rows update.alters_data = True def _update(self, values): """ A version of update() that accepts field objects instead of field names. Used primarily for model saving and not intended for use by general code (it requires too much poking around at model internals to be useful at that level). """ assert self.query.can_filter(), \ "Cannot update a query once a slice has been taken." query = self.query.chain(sql.UpdateQuery) query.add_update_fields(values) self._result_cache = None return query.get_compiler(self.db).execute_sql(CURSOR) _update.alters_data = True _update.queryset_only = False def exists(self): if self._result_cache is None: return self.query.has_results(using=self.db) return bool(self._result_cache) def _prefetch_related_objects(self): # This method can only be called once the result cache has been filled. prefetch_related_objects(self._result_cache, *self._prefetch_related_lookups) self._prefetch_done = True ################################################## # PUBLIC METHODS THAT RETURN A QUERYSET SUBCLASS # ################################################## def raw(self, raw_query, params=None, translations=None, using=None): if using is None: using = self.db return RawQuerySet(raw_query, model=self.model, params=params, translations=translations, using=using) def _values(self, *fields, **expressions): clone = self._chain() if expressions: clone = clone.annotate(**expressions) clone._fields = fields clone.query.set_values(fields) return clone def values(self, *fields, **expressions): fields += tuple(expressions) clone = self._values(*fields, **expressions) clone._iterable_class = ValuesIterable return clone def values_list(self, *fields, flat=False, named=False): if flat and named: raise TypeError("'flat' and 'named' can't be used together.") if flat and len(fields) > 1: raise TypeError("'flat' is not valid when values_list is called with more than one field.") field_names = {f for f in fields if not hasattr(f, 'resolve_expression')} _fields = [] expressions = {} counter = 1 for field in fields: if hasattr(field, 'resolve_expression'): field_id_prefix = getattr(field, 'default_alias', field.__class__.__name__.lower()) while True: field_id = field_id_prefix + str(counter) counter += 1 if field_id not in field_names: break expressions[field_id] = field _fields.append(field_id) else: _fields.append(field) clone = self._values(*_fields, **expressions) clone._iterable_class = ( NamedValuesListIterable if named else FlatValuesListIterable if flat else ValuesListIterable ) return clone def dates(self, field_name, kind, order='ASC'): """ Return a list of date objects representing all available dates for the given field_name, scoped to 'kind'. """ assert kind in ("year", "month", "day"), \ "'kind' must be one of 'year', 'month' or 'day'." assert order in ('ASC', 'DESC'), \ "'order' must be either 'ASC' or 'DESC'." return self.annotate( datefield=Trunc(field_name, kind, output_field=DateField()), plain_field=F(field_name) ).values_list( 'datefield', flat=True ).distinct().filter(plain_field__isnull=False).order_by(('-' if order == 'DESC' else '') + 'datefield') def datetimes(self, field_name, kind, order='ASC', tzinfo=None): """ Return a list of datetime objects representing all available datetimes for the given field_name, scoped to 'kind'. """ assert kind in ("year", "month", "day", "hour", "minute", "second"), \ "'kind' must be one of 'year', 'month', 'day', 'hour', 'minute' or 'second'." assert order in ('ASC', 'DESC'), \ "'order' must be either 'ASC' or 'DESC'." if settings.USE_TZ: if tzinfo is None: tzinfo = timezone.get_current_timezone() else: tzinfo = None return self.annotate( datetimefield=Trunc(field_name, kind, output_field=DateTimeField(), tzinfo=tzinfo), plain_field=F(field_name) ).values_list( 'datetimefield', flat=True ).distinct().filter(plain_field__isnull=False).order_by(('-' if order == 'DESC' else '') + 'datetimefield') def none(self): """Return an empty QuerySet.""" clone = self._chain() clone.query.set_empty() return clone ################################################################## # PUBLIC METHODS THAT ALTER ATTRIBUTES AND RETURN A NEW QUERYSET # ################################################################## def all(self): """ Return a new QuerySet that is a copy of the current one. This allows a QuerySet to proxy for a model manager in some cases. """ return self._chain() def filter(self, *args, **kwargs): """ Return a new QuerySet instance with the args ANDed to the existing set. """ return self._filter_or_exclude(False, *args, **kwargs) def exclude(self, *args, **kwargs): """ Return a new QuerySet instance with NOT (args) ANDed to the existing set. """ return self._filter_or_exclude(True, *args, **kwargs) def _filter_or_exclude(self, negate, *args, **kwargs): if args or kwargs: assert self.query.can_filter(), \ "Cannot filter a query once a slice has been taken." clone = self._chain() if negate: clone.query.add_q(~Q(*args, **kwargs)) else: clone.query.add_q(Q(*args, **kwargs)) return clone def complex_filter(self, filter_obj): """ Return a new QuerySet instance with filter_obj added to the filters. filter_obj can be a Q object or a dictionary of keyword lookup arguments. This exists to support framework features such as 'limit_choices_to', and usually it will be more natural to use other methods. """ if isinstance(filter_obj, Q): clone = self._chain() clone.query.add_q(filter_obj) return clone else: return self._filter_or_exclude(None, **filter_obj) def _combinator_query(self, combinator, *other_qs, all=False): # Clone the query to inherit the select list and everything clone = self._chain() # Clear limits and ordering so they can be reapplied clone.query.clear_ordering(True) clone.query.clear_limits() clone.query.combined_queries = (self.query,) + tuple(qs.query for qs in other_qs) clone.query.combinator = combinator clone.query.combinator_all = all return clone def union(self, *other_qs, all=False): # If the query is an EmptyQuerySet, combine all nonempty querysets. if isinstance(self, EmptyQuerySet): qs = [q for q in other_qs if not isinstance(q, EmptyQuerySet)] return qs[0]._combinator_query('union', *qs[1:], all=all) if qs else self return self._combinator_query('union', *other_qs, all=all) def intersection(self, *other_qs): # If any query is an EmptyQuerySet, return it. if isinstance(self, EmptyQuerySet): return self for other in other_qs: if isinstance(other, EmptyQuerySet): return other return self._combinator_query('intersection', *other_qs) def difference(self, *other_qs): # If the query is an EmptyQuerySet, return it. if isinstance(self, EmptyQuerySet): return self return self._combinator_query('difference', *other_qs) def select_for_update(self, nowait=False, skip_locked=False, of=()): """ Return a new QuerySet instance that will select objects with a FOR UPDATE lock. """ if nowait and skip_locked: raise ValueError('The nowait option cannot be used with skip_locked.') obj = self._chain() obj._for_write = True obj.query.select_for_update = True obj.query.select_for_update_nowait = nowait obj.query.select_for_update_skip_locked = skip_locked obj.query.select_for_update_of = of return obj def select_related(self, *fields): """ Return a new QuerySet instance that will select related objects. If fields are specified, they must be ForeignKey fields and only those related objects are included in the selection. If select_related(None) is called, clear the list. """ if self._fields is not None: raise TypeError("Cannot call select_related() after .values() or .values_list()") obj = self._chain() if fields == (None,): obj.query.select_related = False elif fields: obj.query.add_select_related(fields) else: obj.query.select_related = True return obj def prefetch_related(self, *lookups): """ Return a new QuerySet instance that will prefetch the specified Many-To-One and Many-To-Many related objects when the QuerySet is evaluated. When prefetch_related() is called more than once, append to the list of prefetch lookups. If prefetch_related(None) is called, clear the list. """ clone = self._chain() if lookups == (None,): clone._prefetch_related_lookups = () else: for lookup in lookups: if isinstance(lookup, Prefetch): lookup = lookup.prefetch_to lookup = lookup.split(LOOKUP_SEP, 1)[0] if lookup in self.query._filtered_relations: raise ValueError('prefetch_related() is not supported with FilteredRelation.') clone._prefetch_related_lookups = clone._prefetch_related_lookups + lookups return clone def annotate(self, *args, **kwargs): """ Return a query set in which the returned objects have been annotated with extra data or aggregations. """ self._validate_values_are_expressions(args + tuple(kwargs.values()), method_name='annotate') annotations = OrderedDict() # To preserve ordering of args for arg in args: # The default_alias property may raise a TypeError. try: if arg.default_alias in kwargs: raise ValueError("The named annotation '%s' conflicts with the " "default name for another annotation." % arg.default_alias) except TypeError: raise TypeError("Complex annotations require an alias") annotations[arg.default_alias] = arg annotations.update(kwargs) clone = self._chain() names = self._fields if names is None: names = {f.name for f in self.model._meta.get_fields()} for alias, annotation in annotations.items(): if alias in names: raise ValueError("The annotation '%s' conflicts with a field on " "the model." % alias) if isinstance(annotation, FilteredRelation): clone.query.add_filtered_relation(annotation, alias) else: clone.query.add_annotation(annotation, alias, is_summary=False) for alias, annotation in clone.query.annotations.items(): if alias in annotations and annotation.contains_aggregate: if clone._fields is None: clone.query.group_by = True else: clone.query.set_group_by() break return clone def order_by(self, *field_names): """Return a new QuerySet instance with the ordering changed.""" assert self.query.can_filter(), \ "Cannot reorder a query once a slice has been taken." obj = self._chain() obj.query.clear_ordering(force_empty=False) obj.query.add_ordering(*field_names) return obj def distinct(self, *field_names): """ Return a new QuerySet instance that will select only distinct results. """ assert self.query.can_filter(), \ "Cannot create distinct fields once a slice has been taken." obj = self._chain() obj.query.add_distinct_fields(*field_names) return obj def extra(self, select=None, where=None, params=None, tables=None, order_by=None, select_params=None): """Add extra SQL fragments to the query.""" assert self.query.can_filter(), \ "Cannot change a query once a slice has been taken" clone = self._chain() clone.query.add_extra(select, select_params, where, params, tables, order_by) return clone def reverse(self): """Reverse the ordering of the QuerySet.""" if not self.query.can_filter(): raise TypeError('Cannot reverse a query once a slice has been taken.') clone = self._chain() clone.query.standard_ordering = not clone.query.standard_ordering return clone def defer(self, *fields): """ Defer the loading of data for certain fields until they are accessed. Add the set of deferred fields to any existing set of deferred fields. The only exception to this is if None is passed in as the only parameter, in which case removal all deferrals. """ if self._fields is not None: raise TypeError("Cannot call defer() after .values() or .values_list()") clone = self._chain() if fields == (None,): clone.query.clear_deferred_loading() else: clone.query.add_deferred_loading(fields) return clone def only(self, *fields): """ Essentially, the opposite of defer(). Only the fields passed into this method and that are not already specified as deferred are loaded immediately when the queryset is evaluated. """ if self._fields is not None: raise TypeError("Cannot call only() after .values() or .values_list()") if fields == (None,): # Can only pass None to defer(), not only(), as the rest option. # That won't stop people trying to do this, so let's be explicit. raise TypeError("Cannot pass None as an argument to only().") for field in fields: field = field.split(LOOKUP_SEP, 1)[0] if field in self.query._filtered_relations: raise ValueError('only() is not supported with FilteredRelation.') clone = self._chain() clone.query.add_immediate_loading(fields) return clone def using(self, alias): """Select which database this QuerySet should execute against.""" clone = self._chain() clone._db = alias return clone ################################### # PUBLIC INTROSPECTION ATTRIBUTES # ################################### @property def ordered(self): """ Return True if the QuerySet is ordered -- i.e. has an order_by() clause or a default ordering on the model. """ if self.query.extra_order_by or self.query.order_by: return True elif self.query.default_ordering and self.query.get_meta().ordering: return True else: return False @property def db(self): """Return the database used if this query is executed now.""" if self._for_write: return self._db or router.db_for_write(self.model, **self._hints) return self._db or router.db_for_read(self.model, **self._hints) ################### # PRIVATE METHODS # ################### def _insert(self, objs, fields, return_id=False, raw=False, using=None): """ Insert a new record for the given model. This provides an interface to the InsertQuery class and is how Model.save() is implemented. """ self._for_write = True if using is None: using = self.db query = sql.InsertQuery(self.model) query.insert_values(fields, objs, raw=raw) return query.get_compiler(using=using).execute_sql(return_id) _insert.alters_data = True _insert.queryset_only = False def _batched_insert(self, objs, fields, batch_size): """ Helper method for bulk_create() to insert objs one batch at a time. """ ops = connections[self.db].ops batch_size = (batch_size or max(ops.bulk_batch_size(fields, objs), 1)) inserted_ids = [] for item in [objs[i:i + batch_size] for i in range(0, len(objs), batch_size)]: if connections[self.db].features.can_return_ids_from_bulk_insert: inserted_id = self._insert(item, fields=fields, using=self.db, return_id=True) if isinstance(inserted_id, list): inserted_ids.extend(inserted_id) else: inserted_ids.append(inserted_id) else: self._insert(item, fields=fields, using=self.db) return inserted_ids def _chain(self, **kwargs): """ Return a copy of the current QuerySet that's ready for another operation. """ obj = self._clone() if obj._sticky_filter: obj.query.filter_is_sticky = True obj._sticky_filter = False obj.__dict__.update(kwargs) return obj def _clone(self): """ Return a copy of the current QuerySet. A lightweight alternative to deepcopy(). """ c = self.__class__(model=self.model, query=self.query.chain(), using=self._db, hints=self._hints) c._sticky_filter = self._sticky_filter c._for_write = self._for_write c._prefetch_related_lookups = self._prefetch_related_lookups[:] c._known_related_objects = self._known_related_objects c._iterable_class = self._iterable_class c._fields = self._fields return c def _fetch_all(self): if self._result_cache is None: self._result_cache = list(self._iterable_class(self)) if self._prefetch_related_lookups and not self._prefetch_done: self._prefetch_related_objects() def _next_is_sticky(self): """ Indicate that the next filter call and the one following that should be treated as a single filter. This is only important when it comes to determining when to reuse tables for many-to-many filters. Required so that we can filter naturally on the results of related managers. This doesn't return a clone of the current QuerySet (it returns "self"). The method is only used internally and should be immediately followed by a filter() that does create a clone. """ self._sticky_filter = True return self def _merge_sanity_check(self, other): """Check that two QuerySet classes may be merged.""" if self._fields is not None and ( set(self.query.values_select) != set(other.query.values_select) or set(self.query.extra_select) != set(other.query.extra_select) or set(self.query.annotation_select) != set(other.query.annotation_select)): raise TypeError( "Merging '%s' classes must involve the same values in each case." % self.__class__.__name__ ) def _merge_known_related_objects(self, other): """ Keep track of all known related objects from either QuerySet instance. """ for field, objects in other._known_related_objects.items(): self._known_related_objects.setdefault(field, {}).update(objects) def resolve_expression(self, *args, **kwargs): if self._fields and len(self._fields) > 1: # values() queryset can only be used as nested queries # if they are set up to select only a single field. raise TypeError('Cannot use multi-field values as a filter value.') query = self.query.resolve_expression(*args, **kwargs) query._db = self._db return query resolve_expression.queryset_only = True def _add_hints(self, **hints): """ Update hinting information for use by routers. Add new key/values or overwrite existing key/values. """ self._hints.update(hints) def _has_filters(self): """ Check if this QuerySet has any filtering going on. This isn't equivalent with checking if all objects are present in results, for example, qs[1:]._has_filters() -> False. """ return self.query.has_filters() @staticmethod def _validate_values_are_expressions(values, method_name): invalid_args = sorted(str(arg) for arg in values if not hasattr(arg, 'resolve_expression')) if invalid_args: raise TypeError( 'QuerySet.%s() received non-expression(s): %s.' % ( method_name, ', '.join(invalid_args), ) )
class InstanceCheckMeta(type): def __instancecheck__(self, instance): return isinstance(instance, QuerySet) and instance.query.is_empty() class EmptyQuerySet(metaclass=InstanceCheckMeta): """ Marker class to checking if a queryset is empty by .none(): isinstance(qs.none(), EmptyQuerySet) -> True """ def __init__(self, *args, **kwargs): raise TypeError("EmptyQuerySet can't be instantiated") class RawQuerySet: """ Provide an iterator which converts the results of raw SQL queries into annotated model instances. """ def __init__(self, raw_query, model=None, query=None, params=None, translations=None, using=None, hints=None): self.raw_query = raw_query self.model = model self._db = using self._hints = hints or {} self.query = query or sql.RawQuery(sql=raw_query, using=self.db, params=params) self.params = params or () self.translations = translations or {} def resolve_model_init_order(self): """Resolve the init field names and value positions.""" converter = connections[self.db].introspection.column_name_converter model_init_fields = [f for f in self.model._meta.fields if converter(f.column) in self.columns] annotation_fields = [(column, pos) for pos, column in enumerate(self.columns) if column not in self.model_fields] model_init_order = [self.columns.index(converter(f.column)) for f in model_init_fields] model_init_names = [f.attname for f in model_init_fields] return model_init_names, model_init_order, annotation_fields def __iter__(self): # Cache some things for performance reasons outside the loop. db = self.db compiler = connections[db].ops.compiler('SQLCompiler')( self.query, connections[db], db ) query = iter(self.query) try: model_init_names, model_init_pos, annotation_fields = self.resolve_model_init_order() if self.model._meta.pk.attname not in model_init_names: raise InvalidQuery('Raw query must include the primary key') model_cls = self.model fields = [self.model_fields.get(c) for c in self.columns] converters = compiler.get_converters([ f.get_col(f.model._meta.db_table) if f else None for f in fields ]) if converters: query = compiler.apply_converters(query, converters) for values in query: # Associate fields to values model_init_values = [values[pos] for pos in model_init_pos] instance = model_cls.from_db(db, model_init_names, model_init_values) if annotation_fields: for column, pos in annotation_fields: setattr(instance, column, values[pos]) yield instance finally: # Done iterating the Query. If it has its own cursor, close it. if hasattr(self.query, 'cursor') and self.query.cursor: self.query.cursor.close() def __repr__(self): return "<%s: %s>" % (self.__class__.__name__, self.query) def __getitem__(self, k): return list(self)[k] @property def db(self): """Return the database used if this query is executed now.""" return self._db or router.db_for_read(self.model, **self._hints) def using(self, alias): """Select the database this RawQuerySet should execute against.""" return RawQuerySet( self.raw_query, model=self.model, query=self.query.chain(using=alias), params=self.params, translations=self.translations, using=alias, ) @cached_property def columns(self): """ A list of model field names in the order they'll appear in the query results. """ columns = self.query.get_columns() # Adjust any column names which don't match field names for (query_name, model_name) in self.translations.items(): # Ignore translations for nonexistent column names try: index = columns.index(query_name) except ValueError: pass else: columns[index] = model_name return columns @cached_property def model_fields(self): """A dict mapping column names to model field names.""" converter = connections[self.db].introspection.table_name_converter model_fields = {} for field in self.model._meta.fields: name, column = field.get_attname_column() model_fields[converter(column)] = field return model_fields
[docs]class Prefetch: def __init__(self, lookup, queryset=None, to_attr=None): # `prefetch_through` is the path we traverse to perform the prefetch. self.prefetch_through = lookup # `prefetch_to` is the path to the attribute that stores the result. self.prefetch_to = lookup if queryset is not None and not issubclass(queryset._iterable_class, ModelIterable): raise ValueError('Prefetch querysets cannot use values().') if to_attr: self.prefetch_to = LOOKUP_SEP.join(lookup.split(LOOKUP_SEP)[:-1] + [to_attr]) self.queryset = queryset self.to_attr = to_attr def __getstate__(self): obj_dict = self.__dict__.copy() if self.queryset is not None: # Prevent the QuerySet from being evaluated obj_dict['queryset'] = self.queryset._chain( _result_cache=[], _prefetch_done=True, ) return obj_dict def add_prefix(self, prefix): self.prefetch_through = prefix + LOOKUP_SEP + self.prefetch_through self.prefetch_to = prefix + LOOKUP_SEP + self.prefetch_to def get_current_prefetch_to(self, level): return LOOKUP_SEP.join(self.prefetch_to.split(LOOKUP_SEP)[:level + 1]) def get_current_to_attr(self, level): parts = self.prefetch_to.split(LOOKUP_SEP) to_attr = parts[level] as_attr = self.to_attr and level == len(parts) - 1 return to_attr, as_attr def get_current_queryset(self, level): if self.get_current_prefetch_to(level) == self.prefetch_to: return self.queryset return None def __eq__(self, other): return isinstance(other, Prefetch) and self.prefetch_to == other.prefetch_to def __hash__(self): return hash((self.__class__, self.prefetch_to))
def normalize_prefetch_lookups(lookups, prefix=None): """Normalize lookups into Prefetch objects.""" ret = [] for lookup in lookups: if not isinstance(lookup, Prefetch): lookup = Prefetch(lookup) if prefix: lookup.add_prefix(prefix) ret.append(lookup) return ret def get_prefetcher(instance, through_attr, to_attr): """ For the attribute 'through_attr' on the given instance, find an object that has a get_prefetch_queryset(). Return a 4 tuple containing: (the object with get_prefetch_queryset (or None), the descriptor object representing this relationship (or None), a boolean that is False if the attribute was not found at all, a boolean that is True if the attribute has already been fetched) """ prefetcher = None is_fetched = False # For singly related objects, we have to avoid getting the attribute # from the object, as this will trigger the query. So we first try # on the class, in order to get the descriptor object. rel_obj_descriptor = getattr(instance.__class__, through_attr, None) if rel_obj_descriptor is None: attr_found = hasattr(instance, through_attr) else: attr_found = True if rel_obj_descriptor: # singly related object, descriptor object has the # get_prefetch_queryset() method. if hasattr(rel_obj_descriptor, 'get_prefetch_queryset'): prefetcher = rel_obj_descriptor if rel_obj_descriptor.is_cached(instance): is_fetched = True else: # descriptor doesn't support prefetching, so we go ahead and get # the attribute on the instance rather than the class to # support many related managers rel_obj = getattr(instance, through_attr) if hasattr(rel_obj, 'get_prefetch_queryset'): prefetcher = rel_obj if through_attr != to_attr: # Special case cached_property instances because hasattr # triggers attribute computation and assignment. if isinstance(getattr(instance.__class__, to_attr, None), cached_property): is_fetched = to_attr in instance.__dict__ else: is_fetched = hasattr(instance, to_attr) else: is_fetched = through_attr in instance._prefetched_objects_cache return prefetcher, rel_obj_descriptor, attr_found, is_fetched def prefetch_one_level(instances, prefetcher, lookup, level): """ Helper function for prefetch_related_objects(). Run prefetches on all instances using the prefetcher object, assigning results to relevant caches in instance. Return the prefetched objects along with any additional prefetches that must be done due to prefetch_related lookups found from default managers. """ # prefetcher must have a method get_prefetch_queryset() which takes a list # of instances, and returns a tuple: # (queryset of instances of self.model that are related to passed in instances, # callable that gets value to be matched for returned instances, # callable that gets value to be matched for passed in instances, # boolean that is True for singly related objects, # cache or field name to assign to, # boolean that is True when the previous argument is a cache name vs a field name). # The 'values to be matched' must be hashable as they will be used # in a dictionary. rel_qs, rel_obj_attr, instance_attr, single, cache_name, is_descriptor = ( prefetcher.get_prefetch_queryset(instances, lookup.get_current_queryset(level))) # We have to handle the possibility that the QuerySet we just got back # contains some prefetch_related lookups. We don't want to trigger the # prefetch_related functionality by evaluating the query. Rather, we need # to merge in the prefetch_related lookups. # Copy the lookups in case it is a Prefetch object which could be reused # later (happens in nested prefetch_related). additional_lookups = [ copy.copy(additional_lookup) for additional_lookup in getattr(rel_qs, '_prefetch_related_lookups', ()) ] if additional_lookups: # Don't need to clone because the manager should have given us a fresh # instance, so we access an internal instead of using public interface # for performance reasons. rel_qs._prefetch_related_lookups = () all_related_objects = list(rel_qs) rel_obj_cache = {} for rel_obj in all_related_objects: rel_attr_val = rel_obj_attr(rel_obj) rel_obj_cache.setdefault(rel_attr_val, []).append(rel_obj) to_attr, as_attr = lookup.get_current_to_attr(level) # Make sure `to_attr` does not conflict with a field. if as_attr and instances: # We assume that objects retrieved are homogeneous (which is the premise # of prefetch_related), so what applies to first object applies to all. model = instances[0].__class__ try: model._meta.get_field(to_attr) except exceptions.FieldDoesNotExist: pass else: msg = 'to_attr={} conflicts with a field on the {} model.' raise ValueError(msg.format(to_attr, model.__name__)) # Whether or not we're prefetching the last part of the lookup. leaf = len(lookup.prefetch_through.split(LOOKUP_SEP)) - 1 == level for obj in instances: instance_attr_val = instance_attr(obj) vals = rel_obj_cache.get(instance_attr_val, []) if single: val = vals[0] if vals else None if as_attr: # A to_attr has been given for the prefetch. setattr(obj, to_attr, val) elif is_descriptor: # cache_name points to a field name in obj. # This field is a descriptor for a related object. setattr(obj, cache_name, val) else: # No to_attr has been given for this prefetch operation and the # cache_name does not point to a descriptor. Store the value of # the field in the object's field cache. obj._state.fields_cache[cache_name] = val else: if as_attr: setattr(obj, to_attr, vals) else: manager = getattr(obj, to_attr) if leaf and lookup.queryset is not None: qs = manager._apply_rel_filters(lookup.queryset) else: qs = manager.get_queryset() qs._result_cache = vals # We don't want the individual qs doing prefetch_related now, # since we have merged this into the current work. qs._prefetch_done = True obj._prefetched_objects_cache[cache_name] = qs return all_related_objects, additional_lookups class RelatedPopulator: """ RelatedPopulator is used for select_related() object instantiation. The idea is that each select_related() model will be populated by a different RelatedPopulator instance. The RelatedPopulator instances get klass_info and select (computed in SQLCompiler) plus the used db as input for initialization. That data is used to compute which columns to use, how to instantiate the model, and how to populate the links between the objects. The actual creation of the objects is done in populate() method. This method gets row and from_obj as input and populates the select_related() model instance. """ def __init__(self, klass_info, select, db): self.db = db # Pre-compute needed attributes. The attributes are: # - model_cls: the possibly deferred model class to instantiate # - either: # - cols_start, cols_end: usually the columns in the row are # in the same order model_cls.__init__ expects them, so we # can instantiate by model_cls(*row[cols_start:cols_end]) # - reorder_for_init: When select_related descends to a child # class, then we want to reuse the already selected parent # data. However, in this case the parent data isn't necessarily # in the same order that Model.__init__ expects it to be, so # we have to reorder the parent data. The reorder_for_init # attribute contains a function used to reorder the field data # in the order __init__ expects it. # - pk_idx: the index of the primary key field in the reordered # model data. Used to check if a related object exists at all. # - init_list: the field attnames fetched from the database. For # deferred models this isn't the same as all attnames of the # model's fields. # - related_populators: a list of RelatedPopulator instances if # select_related() descends to related models from this model. # - local_setter, remote_setter: Methods to set cached values on # the object being populated and on the remote object. Usually # these are Field.set_cached_value() methods. select_fields = klass_info['select_fields'] from_parent = klass_info['from_parent'] if not from_parent: self.cols_start = select_fields[0] self.cols_end = select_fields[-1] + 1 self.init_list = [ f[0].target.attname for f in select[self.cols_start:self.cols_end] ] self.reorder_for_init = None else: attname_indexes = {select[idx][0].target.attname: idx for idx in select_fields} model_init_attnames = (f.attname for f in klass_info['model']._meta.concrete_fields) self.init_list = [attname for attname in model_init_attnames if attname in attname_indexes] self.reorder_for_init = operator.itemgetter(*[attname_indexes[attname] for attname in self.init_list]) self.model_cls = klass_info['model'] self.pk_idx = self.init_list.index(self.model_cls._meta.pk.attname) self.related_populators = get_related_populators(klass_info, select, self.db) self.local_setter = klass_info['local_setter'] self.remote_setter = klass_info['remote_setter'] def populate(self, row, from_obj): if self.reorder_for_init: obj_data = self.reorder_for_init(row) else: obj_data = row[self.cols_start:self.cols_end] if obj_data[self.pk_idx] is None: obj = None else: obj = self.model_cls.from_db(self.db, self.init_list, obj_data) if self.related_populators: for rel_iter in self.related_populators: rel_iter.populate(row, obj) self.local_setter(from_obj, obj) if obj is not None: self.remote_setter(obj, from_obj) def get_related_populators(klass_info, select, db): iterators = [] related_klass_infos = klass_info.get('related_klass_infos', []) for rel_klass_info in related_klass_infos: rel_cls = RelatedPopulator(rel_klass_info, select, db) iterators.append(rel_cls) return iterators
Back to Top