django1/django/db/models/sql/subqueries.py

193 lines
7.1 KiB
Python

"""
Query subclasses which provide extra functionality beyond simple data retrieval.
"""
from django.core.exceptions import FieldError
from django.db import connections
from django.db.models.query_utils import Q
from django.db.models.sql.constants import (
CURSOR, GET_ITERATOR_CHUNK_SIZE, NO_RESULTS,
)
from django.db.models.sql.query import Query
__all__ = ['DeleteQuery', 'UpdateQuery', 'InsertQuery', 'AggregateQuery']
class DeleteQuery(Query):
"""A DELETE SQL query."""
compiler = 'SQLDeleteCompiler'
def do_query(self, table, where, using):
self.alias_map = {table: self.alias_map[table]}
self.where = where
cursor = self.get_compiler(using).execute_sql(CURSOR)
return cursor.rowcount if cursor else 0
def delete_batch(self, pk_list, using):
"""
Set up and execute delete queries for all the objects in pk_list.
More than one physical query may be executed if there are a
lot of values in pk_list.
"""
# number of objects deleted
num_deleted = 0
field = self.get_meta().pk
for offset in range(0, len(pk_list), GET_ITERATOR_CHUNK_SIZE):
self.where = self.where_class()
self.add_q(Q(
**{field.attname + '__in': pk_list[offset:offset + GET_ITERATOR_CHUNK_SIZE]}))
num_deleted += self.do_query(self.get_meta().db_table, self.where, using=using)
return num_deleted
def delete_qs(self, query, using):
"""
Delete the queryset in one SQL query (if possible). For simple queries
this is done by copying the query.query.where to self.query, for
complex queries by using subquery.
"""
innerq = query.query
# Make sure the inner query has at least one table in use.
innerq.get_initial_alias()
# The same for our new query.
self.get_initial_alias()
innerq_used_tables = tuple([t for t in innerq.alias_map if innerq.alias_refcount[t]])
if not innerq_used_tables or innerq_used_tables == tuple(self.alias_map):
# There is only the base table in use in the query.
self.where = innerq.where
else:
pk = query.model._meta.pk
if not connections[using].features.update_can_self_select:
# We can't do the delete using subquery.
values = list(query.values_list('pk', flat=True))
if not values:
return 0
return self.delete_batch(values, using)
else:
innerq.clear_select_clause()
innerq.select = [
pk.get_col(self.get_initial_alias())
]
values = innerq
self.where = self.where_class()
self.add_q(Q(pk__in=values))
cursor = self.get_compiler(using).execute_sql(CURSOR)
return cursor.rowcount if cursor else 0
class UpdateQuery(Query):
"""An UPDATE SQL query."""
compiler = 'SQLUpdateCompiler'
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self._setup_query()
def _setup_query(self):
"""
Run on initialization and after cloning. Any attributes that would
normally be set in __init__ should go in here, instead, so that they
are also set up after a clone() call.
"""
self.values = []
self.related_ids = None
self.related_updates = {}
def clone(self, klass=None, **kwargs):
return super().clone(klass, related_updates=self.related_updates.copy(), **kwargs)
def update_batch(self, pk_list, values, using):
self.add_update_values(values)
for offset in range(0, len(pk_list), GET_ITERATOR_CHUNK_SIZE):
self.where = self.where_class()
self.add_q(Q(pk__in=pk_list[offset: offset + GET_ITERATOR_CHUNK_SIZE]))
self.get_compiler(using).execute_sql(NO_RESULTS)
def add_update_values(self, values):
"""
Convert a dictionary of field name to value mappings into an update
query. This is the entry point for the public update() method on
querysets.
"""
values_seq = []
for name, val in values.items():
field = self.get_meta().get_field(name)
direct = not (field.auto_created and not field.concrete) or not field.concrete
model = field.model._meta.concrete_model
if not direct or (field.is_relation and field.many_to_many):
raise FieldError(
'Cannot update model field %r (only non-relations and '
'foreign keys permitted).' % field
)
if model is not self.get_meta().model:
self.add_related_update(model, field, val)
continue
values_seq.append((field, model, val))
return self.add_update_fields(values_seq)
def add_update_fields(self, values_seq):
"""
Append a sequence of (field, model, value) triples to the internal list
that will be used to generate the UPDATE query. Might be more usefully
called add_update_targets() to hint at the extra information here.
"""
for field, model, val in values_seq:
if hasattr(val, 'resolve_expression'):
# Resolve expressions here so that annotations are no longer needed
val = val.resolve_expression(self, allow_joins=False, for_save=True)
self.values.append((field, model, val))
def add_related_update(self, model, field, value):
"""
Add (name, value) to an update query for an ancestor model.
Update are coalesced so that only one update query per ancestor is run.
"""
self.related_updates.setdefault(model, []).append((field, None, value))
def get_related_updates(self):
"""
Return a list of query objects: one for each update required to an
ancestor model. Each query will have the same filtering conditions as
the current query but will only update a single table.
"""
if not self.related_updates:
return []
result = []
for model, values in self.related_updates.items():
query = UpdateQuery(model)
query.values = values
if self.related_ids is not None:
query.add_filter(('pk__in', self.related_ids))
result.append(query)
return result
class InsertQuery(Query):
compiler = 'SQLInsertCompiler'
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.fields = []
self.objs = []
def insert_values(self, fields, objs, raw=False):
self.fields = fields
self.objs = objs
self.raw = raw
class AggregateQuery(Query):
"""
Take another query as a parameter to the FROM clause and only select the
elements in the provided list.
"""
compiler = 'SQLAggregateCompiler'
def add_subquery(self, query, using):
query.subquery = True
self.subquery, self.sub_params = query.get_compiler(using).as_sql(with_col_aliases=True)