2013-02-10 23:15:49 +08:00
|
|
|
import datetime
|
2011-09-10 03:22:28 +08:00
|
|
|
|
2013-02-10 23:15:49 +08:00
|
|
|
from django.conf import settings
|
2009-12-22 23:18:51 +08:00
|
|
|
from django.core.exceptions import FieldError
|
|
|
|
from django.db.backends.util import truncate_name
|
2012-09-09 07:51:36 +08:00
|
|
|
from django.db.models.constants import LOOKUP_SEP
|
2013-03-25 00:40:40 +08:00
|
|
|
from django.db.models.query_utils import select_related_descend, QueryWrapper
|
2012-09-07 22:58:17 +08:00
|
|
|
from django.db.models.sql.constants import (SINGLE, MULTI, ORDER_DIR,
|
2012-12-21 03:25:48 +08:00
|
|
|
GET_ITERATOR_CHUNK_SIZE, SelectInfo)
|
2009-12-22 23:18:51 +08:00
|
|
|
from django.db.models.sql.datastructures import EmptyResultSet
|
|
|
|
from django.db.models.sql.expressions import SQLEvaluator
|
2012-02-22 13:26:50 +08:00
|
|
|
from django.db.models.sql.query import get_order_dir, Query
|
2011-04-21 04:42:07 +08:00
|
|
|
from django.db.utils import DatabaseError
|
2012-07-21 03:14:27 +08:00
|
|
|
from django.utils import six
|
2013-02-10 23:15:49 +08:00
|
|
|
from django.utils.six.moves import zip
|
|
|
|
from django.utils import timezone
|
2009-12-22 23:18:51 +08:00
|
|
|
|
2011-09-10 03:22:28 +08:00
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
class SQLCompiler(object):
|
|
|
|
def __init__(self, query, connection, using):
|
|
|
|
self.query = query
|
|
|
|
self.connection = connection
|
|
|
|
self.using = using
|
|
|
|
self.quote_cache = {}
|
|
|
|
|
|
|
|
def pre_sql_setup(self):
|
|
|
|
"""
|
|
|
|
Does any necessary class setup immediately prior to producing SQL. This
|
|
|
|
is for things that can't necessarily be done in __init__ because we
|
|
|
|
might not have all the pieces in place at that time.
|
2011-12-23 04:42:40 +08:00
|
|
|
# TODO: after the query has been executed, the altered state should be
|
|
|
|
# cleaned. We are not using a clone() of the query here.
|
2009-12-22 23:18:51 +08:00
|
|
|
"""
|
|
|
|
if not self.query.tables:
|
2013-05-14 03:40:39 +08:00
|
|
|
self.query.join((None, self.query.get_meta().db_table, None))
|
2009-12-22 23:18:51 +08:00
|
|
|
if (not self.query.select and self.query.default_cols and not
|
|
|
|
self.query.included_inherited_models):
|
|
|
|
self.query.setup_inherited_models()
|
|
|
|
if self.query.select_related and not self.query.related_select_cols:
|
|
|
|
self.fill_related_selections()
|
|
|
|
|
|
|
|
def quote_name_unless_alias(self, name):
|
|
|
|
"""
|
|
|
|
A wrapper around connection.ops.quote_name that doesn't quote aliases
|
|
|
|
for table names. This avoids problems with some SQL dialects that treat
|
|
|
|
quoted strings specially (e.g. PostgreSQL).
|
|
|
|
"""
|
|
|
|
if name in self.quote_cache:
|
|
|
|
return self.quote_cache[name]
|
|
|
|
if ((name in self.query.alias_map and name not in self.query.table_map) or
|
|
|
|
name in self.query.extra_select):
|
|
|
|
self.quote_cache[name] = name
|
|
|
|
return name
|
|
|
|
r = self.connection.ops.quote_name(name)
|
|
|
|
self.quote_cache[name] = r
|
|
|
|
return r
|
|
|
|
|
|
|
|
def as_sql(self, with_limits=True, with_col_aliases=False):
|
|
|
|
"""
|
|
|
|
Creates the SQL for this query. Returns the SQL string and list of
|
|
|
|
parameters.
|
|
|
|
|
|
|
|
If 'with_limits' is False, any limit/offset information is not included
|
|
|
|
in the query.
|
|
|
|
"""
|
2010-10-14 07:36:16 +08:00
|
|
|
if with_limits and self.query.low_mark == self.query.high_mark:
|
|
|
|
return '', ()
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
self.pre_sql_setup()
|
2011-12-23 04:42:40 +08:00
|
|
|
# After executing the query, we must get rid of any joins the query
|
|
|
|
# setup created. So, take note of alias counts before the query ran.
|
|
|
|
# However we do not want to get rid of stuff done in pre_sql_setup(),
|
|
|
|
# as the pre_sql_setup will modify query state in a way that forbids
|
|
|
|
# another run of it.
|
|
|
|
self.refcounts_before = self.query.alias_refcount.copy()
|
2013-02-10 23:15:49 +08:00
|
|
|
out_cols, s_params = self.get_columns(with_col_aliases)
|
2009-12-22 23:18:51 +08:00
|
|
|
ordering, ordering_group_by = self.get_ordering()
|
|
|
|
|
2011-12-23 04:42:40 +08:00
|
|
|
distinct_fields = self.get_distinct()
|
|
|
|
|
|
|
|
# This must come after 'select', 'ordering' and 'distinct' -- see
|
|
|
|
# docstring of get_from_clause() for details.
|
2009-12-22 23:18:51 +08:00
|
|
|
from_, f_params = self.get_from_clause()
|
|
|
|
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
|
|
|
|
where, w_params = self.query.where.as_sql(qn=qn, connection=self.connection)
|
|
|
|
having, h_params = self.query.having.as_sql(qn=qn, connection=self.connection)
|
Refactored qs.add_q() and utils/tree.py
The sql/query.py add_q method did a lot of where/having tree hacking to
get complex queries to work correctly. The logic was refactored so that
it should be simpler to understand. The new logic should also produce
leaner WHERE conditions.
The changes cascade somewhat, as some other parts of Django (like
add_filter() and WhereNode) expect boolean trees in certain format or
they fail to work. So to fix the add_q() one must fix utils/tree.py,
some things in add_filter(), WhereNode and so on.
This commit also fixed add_filter to see negate clauses up the path.
A query like .exclude(Q(reversefk__in=a_list)) didn't work similarly to
.filter(~Q(reversefk__in=a_list)). The reason for this is that only
the immediate parent negate clauses were seen by add_filter, and thus a
tree like AND: (NOT AND: (AND: condition)) will not be handled
correctly, as there is one intermediary AND node in the tree. The
example tree is generated by .exclude(~Q(reversefk__in=a_list)).
Still, aggregation lost connectors in OR cases, and F() objects and
aggregates in same filter clause caused GROUP BY problems on some
databases.
Fixed #17600, fixed #13198, fixed #17025, fixed #17000, fixed #11293.
2012-05-25 05:27:24 +08:00
|
|
|
having_group_by = self.query.having.get_cols()
|
2009-12-22 23:18:51 +08:00
|
|
|
params = []
|
2012-07-21 03:14:27 +08:00
|
|
|
for val in six.itervalues(self.query.extra_select):
|
2009-12-22 23:18:51 +08:00
|
|
|
params.extend(val[1])
|
|
|
|
|
|
|
|
result = ['SELECT']
|
2011-12-23 04:42:40 +08:00
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
if self.query.distinct:
|
2011-12-23 04:42:40 +08:00
|
|
|
result.append(self.connection.ops.distinct_sql(distinct_fields))
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
result.append(', '.join(out_cols + self.query.ordering_aliases))
|
2013-02-10 23:15:49 +08:00
|
|
|
params.extend(s_params)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
result.append('FROM')
|
|
|
|
result.extend(from_)
|
|
|
|
params.extend(f_params)
|
|
|
|
|
|
|
|
if where:
|
|
|
|
result.append('WHERE %s' % where)
|
|
|
|
params.extend(w_params)
|
|
|
|
|
Refactored qs.add_q() and utils/tree.py
The sql/query.py add_q method did a lot of where/having tree hacking to
get complex queries to work correctly. The logic was refactored so that
it should be simpler to understand. The new logic should also produce
leaner WHERE conditions.
The changes cascade somewhat, as some other parts of Django (like
add_filter() and WhereNode) expect boolean trees in certain format or
they fail to work. So to fix the add_q() one must fix utils/tree.py,
some things in add_filter(), WhereNode and so on.
This commit also fixed add_filter to see negate clauses up the path.
A query like .exclude(Q(reversefk__in=a_list)) didn't work similarly to
.filter(~Q(reversefk__in=a_list)). The reason for this is that only
the immediate parent negate clauses were seen by add_filter, and thus a
tree like AND: (NOT AND: (AND: condition)) will not be handled
correctly, as there is one intermediary AND node in the tree. The
example tree is generated by .exclude(~Q(reversefk__in=a_list)).
Still, aggregation lost connectors in OR cases, and F() objects and
aggregates in same filter clause caused GROUP BY problems on some
databases.
Fixed #17600, fixed #13198, fixed #17025, fixed #17000, fixed #11293.
2012-05-25 05:27:24 +08:00
|
|
|
grouping, gb_params = self.get_grouping(having_group_by, ordering_group_by)
|
2009-12-22 23:18:51 +08:00
|
|
|
if grouping:
|
2011-12-23 04:42:40 +08:00
|
|
|
if distinct_fields:
|
|
|
|
raise NotImplementedError(
|
|
|
|
"annotate() + distinct(fields) not implemented.")
|
2012-11-08 06:04:57 +08:00
|
|
|
if not ordering:
|
2009-12-22 23:18:51 +08:00
|
|
|
ordering = self.connection.ops.force_no_ordering()
|
|
|
|
result.append('GROUP BY %s' % ', '.join(grouping))
|
|
|
|
params.extend(gb_params)
|
|
|
|
|
|
|
|
if having:
|
|
|
|
result.append('HAVING %s' % having)
|
|
|
|
params.extend(h_params)
|
|
|
|
|
|
|
|
if ordering:
|
|
|
|
result.append('ORDER BY %s' % ', '.join(ordering))
|
|
|
|
|
|
|
|
if with_limits:
|
|
|
|
if self.query.high_mark is not None:
|
|
|
|
result.append('LIMIT %d' % (self.query.high_mark - self.query.low_mark))
|
|
|
|
if self.query.low_mark:
|
|
|
|
if self.query.high_mark is None:
|
|
|
|
val = self.connection.ops.no_limit_value()
|
|
|
|
if val:
|
|
|
|
result.append('LIMIT %d' % val)
|
|
|
|
result.append('OFFSET %d' % self.query.low_mark)
|
|
|
|
|
2011-04-21 04:42:07 +08:00
|
|
|
if self.query.select_for_update and self.connection.features.has_select_for_update:
|
|
|
|
# If we've been asked for a NOWAIT query but the backend does not support it,
|
|
|
|
# raise a DatabaseError otherwise we could get an unexpected deadlock.
|
|
|
|
nowait = self.query.select_for_update_nowait
|
|
|
|
if nowait and not self.connection.features.has_select_for_update_nowait:
|
|
|
|
raise DatabaseError('NOWAIT is not supported on this database backend.')
|
|
|
|
result.append(self.connection.ops.for_update_sql(nowait=nowait))
|
|
|
|
|
2011-12-23 04:42:40 +08:00
|
|
|
# Finally do cleanup - get rid of the joins we created above.
|
|
|
|
self.query.reset_refcounts(self.refcounts_before)
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
return ' '.join(result), tuple(params)
|
|
|
|
|
|
|
|
def as_nested_sql(self):
|
|
|
|
"""
|
|
|
|
Perform the same functionality as the as_sql() method, returning an
|
|
|
|
SQL string and parameters. However, the alias prefixes are bumped
|
2010-04-05 01:05:43 +08:00
|
|
|
beforehand (in a copy -- the current query isn't changed), and any
|
|
|
|
ordering is removed if the query is unsliced.
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
Used when nesting this query inside another.
|
|
|
|
"""
|
|
|
|
obj = self.query.clone()
|
2010-04-05 01:05:43 +08:00
|
|
|
if obj.low_mark == 0 and obj.high_mark is None:
|
|
|
|
# If there is no slicing in use, then we can safely drop all ordering
|
|
|
|
obj.clear_ordering(True)
|
2009-12-22 23:18:51 +08:00
|
|
|
obj.bump_prefix()
|
|
|
|
return obj.get_compiler(connection=self.connection).as_sql()
|
|
|
|
|
|
|
|
def get_columns(self, with_aliases=False):
|
|
|
|
"""
|
2013-02-10 23:15:49 +08:00
|
|
|
Returns the list of columns to use in the select statement, as well as
|
|
|
|
a list any extra parameters that need to be included. If no columns
|
|
|
|
have been specified, returns all columns relating to fields in the
|
|
|
|
model.
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
If 'with_aliases' is true, any column names that are duplicated
|
|
|
|
(without the table names) are given unique aliases. This is needed in
|
|
|
|
some cases to avoid ambiguity with nested queries.
|
|
|
|
"""
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
2012-07-21 03:14:27 +08:00
|
|
|
result = ['(%s) AS %s' % (col[0], qn2(alias)) for alias, col in six.iteritems(self.query.extra_select)]
|
2013-02-10 23:15:49 +08:00
|
|
|
params = []
|
2009-12-22 23:18:51 +08:00
|
|
|
aliases = set(self.query.extra_select.keys())
|
|
|
|
if with_aliases:
|
|
|
|
col_aliases = aliases.copy()
|
|
|
|
else:
|
|
|
|
col_aliases = set()
|
|
|
|
if self.query.select:
|
|
|
|
only_load = self.deferred_to_columns()
|
2012-10-26 01:57:32 +08:00
|
|
|
for col, _ in self.query.select:
|
2009-12-22 23:18:51 +08:00
|
|
|
if isinstance(col, (list, tuple)):
|
|
|
|
alias, column = col
|
2012-04-30 06:50:48 +08:00
|
|
|
table = self.query.alias_map[alias].table_name
|
2011-07-07 09:12:45 +08:00
|
|
|
if table in only_load and column not in only_load[table]:
|
2009-12-22 23:18:51 +08:00
|
|
|
continue
|
|
|
|
r = '%s.%s' % (qn(alias), qn(column))
|
|
|
|
if with_aliases:
|
|
|
|
if col[1] in col_aliases:
|
|
|
|
c_alias = 'Col%d' % len(col_aliases)
|
|
|
|
result.append('%s AS %s' % (r, c_alias))
|
|
|
|
aliases.add(c_alias)
|
|
|
|
col_aliases.add(c_alias)
|
|
|
|
else:
|
|
|
|
result.append('%s AS %s' % (r, qn2(col[1])))
|
|
|
|
aliases.add(r)
|
|
|
|
col_aliases.add(col[1])
|
|
|
|
else:
|
|
|
|
result.append(r)
|
|
|
|
aliases.add(r)
|
|
|
|
col_aliases.add(col[1])
|
|
|
|
else:
|
2013-02-10 23:15:49 +08:00
|
|
|
col_sql, col_params = col.as_sql(qn, self.connection)
|
|
|
|
result.append(col_sql)
|
|
|
|
params.extend(col_params)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
if hasattr(col, 'alias'):
|
|
|
|
aliases.add(col.alias)
|
|
|
|
col_aliases.add(col.alias)
|
|
|
|
|
|
|
|
elif self.query.default_cols:
|
|
|
|
cols, new_aliases = self.get_default_columns(with_aliases,
|
|
|
|
col_aliases)
|
|
|
|
result.extend(cols)
|
|
|
|
aliases.update(new_aliases)
|
|
|
|
|
|
|
|
max_name_length = self.connection.ops.max_name_length()
|
2013-02-10 23:15:49 +08:00
|
|
|
for alias, aggregate in self.query.aggregate_select.items():
|
|
|
|
agg_sql, agg_params = aggregate.as_sql(qn, self.connection)
|
|
|
|
if alias is None:
|
|
|
|
result.append(agg_sql)
|
|
|
|
else:
|
|
|
|
result.append('%s AS %s' % (agg_sql, qn(truncate_name(alias, max_name_length))))
|
|
|
|
params.extend(agg_params)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
2012-10-26 01:57:32 +08:00
|
|
|
for (table, col), _ in self.query.related_select_cols:
|
2009-12-22 23:18:51 +08:00
|
|
|
r = '%s.%s' % (qn(table), qn(col))
|
|
|
|
if with_aliases and col in col_aliases:
|
|
|
|
c_alias = 'Col%d' % len(col_aliases)
|
|
|
|
result.append('%s AS %s' % (r, c_alias))
|
|
|
|
aliases.add(c_alias)
|
|
|
|
col_aliases.add(c_alias)
|
|
|
|
else:
|
|
|
|
result.append(r)
|
|
|
|
aliases.add(r)
|
|
|
|
col_aliases.add(col)
|
|
|
|
|
|
|
|
self._select_aliases = aliases
|
2013-02-10 23:15:49 +08:00
|
|
|
return result, params
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
def get_default_columns(self, with_aliases=False, col_aliases=None,
|
2012-11-10 02:21:46 +08:00
|
|
|
start_alias=None, opts=None, as_pairs=False, from_parent=None):
|
2009-12-22 23:18:51 +08:00
|
|
|
"""
|
|
|
|
Computes the default columns for selecting every field in the base
|
|
|
|
model. Will sometimes be called to pull in related models (e.g. via
|
|
|
|
select_related), in which case "opts" and "start_alias" will be given
|
|
|
|
to provide a starting point for the traversal.
|
|
|
|
|
|
|
|
Returns a list of strings, quoted appropriately for use in SQL
|
|
|
|
directly, as well as a set of aliases used in the select statement (if
|
|
|
|
'as_pairs' is True, returns a list of (alias, col_name) pairs instead
|
|
|
|
of strings as the first component and None as the second component).
|
|
|
|
"""
|
|
|
|
result = []
|
|
|
|
if opts is None:
|
2013-05-14 03:40:39 +08:00
|
|
|
opts = self.query.get_meta()
|
2009-12-22 23:18:51 +08:00
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
|
|
|
aliases = set()
|
|
|
|
only_load = self.deferred_to_columns()
|
2013-02-21 17:12:08 +08:00
|
|
|
if not start_alias:
|
|
|
|
start_alias = self.query.get_initial_alias()
|
|
|
|
# The 'seen_models' is used to optimize checking the needed parent
|
|
|
|
# alias for a given field. This also includes None -> start_alias to
|
|
|
|
# be used by local fields.
|
|
|
|
seen_models = {None: start_alias}
|
|
|
|
|
2013-03-25 00:40:40 +08:00
|
|
|
for field, model in opts.get_concrete_fields_with_model():
|
2012-11-10 02:21:46 +08:00
|
|
|
if from_parent and model is not None and issubclass(from_parent, model):
|
|
|
|
# Avoid loading data for already loaded parents.
|
2010-03-20 23:02:59 +08:00
|
|
|
continue
|
2013-02-21 17:12:08 +08:00
|
|
|
alias = self.query.join_parent_model(opts, model, start_alias,
|
|
|
|
seen_models)
|
2012-04-30 06:50:48 +08:00
|
|
|
table = self.query.alias_map[alias].table_name
|
2009-12-22 23:18:51 +08:00
|
|
|
if table in only_load and field.column not in only_load[table]:
|
|
|
|
continue
|
|
|
|
if as_pairs:
|
|
|
|
result.append((alias, field.column))
|
|
|
|
aliases.add(alias)
|
|
|
|
continue
|
|
|
|
if with_aliases and field.column in col_aliases:
|
|
|
|
c_alias = 'Col%d' % len(col_aliases)
|
|
|
|
result.append('%s.%s AS %s' % (qn(alias),
|
|
|
|
qn2(field.column), c_alias))
|
|
|
|
col_aliases.add(c_alias)
|
|
|
|
aliases.add(c_alias)
|
|
|
|
else:
|
|
|
|
r = '%s.%s' % (qn(alias), qn2(field.column))
|
|
|
|
result.append(r)
|
|
|
|
aliases.add(r)
|
|
|
|
if with_aliases:
|
|
|
|
col_aliases.add(field.column)
|
|
|
|
return result, aliases
|
|
|
|
|
2011-12-23 04:42:40 +08:00
|
|
|
def get_distinct(self):
|
|
|
|
"""
|
|
|
|
Returns a quoted list of fields to use in DISTINCT ON part of the query.
|
|
|
|
|
|
|
|
Note that this method can alter the tables in the query, and thus it
|
|
|
|
must be called before get_from_clause().
|
|
|
|
"""
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
|
|
|
result = []
|
2013-05-14 03:40:39 +08:00
|
|
|
opts = self.query.get_meta()
|
2011-12-23 04:42:40 +08:00
|
|
|
|
|
|
|
for name in self.query.distinct_fields:
|
|
|
|
parts = name.split(LOOKUP_SEP)
|
2013-03-25 00:40:40 +08:00
|
|
|
field, cols, alias, _, _ = self._setup_joins(parts, opts, None)
|
|
|
|
cols, alias = self._final_join_removal(cols, alias)
|
|
|
|
for col in cols:
|
|
|
|
result.append("%s.%s" % (qn(alias), qn2(col)))
|
2011-12-23 04:42:40 +08:00
|
|
|
return result
|
|
|
|
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
def get_ordering(self):
|
|
|
|
"""
|
|
|
|
Returns a tuple containing a list representing the SQL elements in the
|
|
|
|
"order by" clause, and the list of SQL elements that need to be added
|
|
|
|
to the GROUP BY clause as a result of the ordering.
|
|
|
|
|
|
|
|
Also sets the ordering_aliases attribute on this instance to a list of
|
|
|
|
extra aliases needed in the select.
|
|
|
|
|
|
|
|
Determining the ordering SQL can change the tables we need to include,
|
|
|
|
so this should be run *before* get_from_clause().
|
|
|
|
"""
|
|
|
|
if self.query.extra_order_by:
|
|
|
|
ordering = self.query.extra_order_by
|
|
|
|
elif not self.query.default_ordering:
|
|
|
|
ordering = self.query.order_by
|
|
|
|
else:
|
2012-01-03 17:06:19 +08:00
|
|
|
ordering = (self.query.order_by
|
2013-05-14 03:40:39 +08:00
|
|
|
or self.query.get_meta().ordering
|
2012-01-03 17:06:19 +08:00
|
|
|
or [])
|
2009-12-22 23:18:51 +08:00
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
|
|
|
distinct = self.query.distinct
|
|
|
|
select_aliases = self._select_aliases
|
|
|
|
result = []
|
|
|
|
group_by = []
|
|
|
|
ordering_aliases = []
|
|
|
|
if self.query.standard_ordering:
|
|
|
|
asc, desc = ORDER_DIR['ASC']
|
|
|
|
else:
|
|
|
|
asc, desc = ORDER_DIR['DESC']
|
|
|
|
|
|
|
|
# It's possible, due to model inheritance, that normal usage might try
|
|
|
|
# to include the same field more than once in the ordering. We track
|
|
|
|
# the table/column pairs we use and discard any after the first use.
|
|
|
|
processed_pairs = set()
|
|
|
|
|
|
|
|
for field in ordering:
|
|
|
|
if field == '?':
|
|
|
|
result.append(self.connection.ops.random_function_sql())
|
|
|
|
continue
|
|
|
|
if isinstance(field, int):
|
|
|
|
if field < 0:
|
|
|
|
order = desc
|
|
|
|
field = -field
|
|
|
|
else:
|
|
|
|
order = asc
|
|
|
|
result.append('%s %s' % (field, order))
|
2012-11-08 06:04:57 +08:00
|
|
|
group_by.append((str(field), []))
|
2009-12-22 23:18:51 +08:00
|
|
|
continue
|
|
|
|
col, order = get_order_dir(field, asc)
|
|
|
|
if col in self.query.aggregate_select:
|
2011-01-26 11:48:15 +08:00
|
|
|
result.append('%s %s' % (qn(col), order))
|
2009-12-22 23:18:51 +08:00
|
|
|
continue
|
|
|
|
if '.' in field:
|
|
|
|
# This came in through an extra(order_by=...) addition. Pass it
|
|
|
|
# on verbatim.
|
|
|
|
table, col = col.split('.', 1)
|
|
|
|
if (table, col) not in processed_pairs:
|
|
|
|
elt = '%s.%s' % (qn(table), col)
|
|
|
|
processed_pairs.add((table, col))
|
|
|
|
if not distinct or elt in select_aliases:
|
|
|
|
result.append('%s %s' % (elt, order))
|
|
|
|
group_by.append((elt, []))
|
|
|
|
elif get_order_dir(field)[0] not in self.query.extra_select:
|
|
|
|
# 'col' is of the form 'field' or 'field1__field2' or
|
|
|
|
# '-field1__field2__field', etc.
|
2013-03-25 00:40:40 +08:00
|
|
|
for table, cols, order in self.find_ordering_name(field,
|
2013-05-14 03:40:39 +08:00
|
|
|
self.query.get_meta(), default_order=asc):
|
2013-03-25 00:40:40 +08:00
|
|
|
for col in cols:
|
|
|
|
if (table, col) not in processed_pairs:
|
|
|
|
elt = '%s.%s' % (qn(table), qn2(col))
|
|
|
|
processed_pairs.add((table, col))
|
|
|
|
if distinct and elt not in select_aliases:
|
|
|
|
ordering_aliases.append(elt)
|
|
|
|
result.append('%s %s' % (elt, order))
|
|
|
|
group_by.append((elt, []))
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
|
|
|
elt = qn2(col)
|
|
|
|
if distinct and col not in select_aliases:
|
|
|
|
ordering_aliases.append(elt)
|
|
|
|
result.append('%s %s' % (elt, order))
|
|
|
|
group_by.append(self.query.extra_select[col])
|
|
|
|
self.query.ordering_aliases = ordering_aliases
|
|
|
|
return result, group_by
|
|
|
|
|
|
|
|
def find_ordering_name(self, name, opts, alias=None, default_order='ASC',
|
|
|
|
already_seen=None):
|
|
|
|
"""
|
|
|
|
Returns the table alias (the name might be ambiguous, the alias will
|
|
|
|
not be) and column name for ordering by the given 'name' parameter.
|
|
|
|
The 'name' is of the form 'field1__field2__...__fieldN'.
|
|
|
|
"""
|
|
|
|
name, order = get_order_dir(name, default_order)
|
|
|
|
pieces = name.split(LOOKUP_SEP)
|
2013-03-25 00:40:40 +08:00
|
|
|
field, cols, alias, joins, opts = self._setup_joins(pieces, opts, alias)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
# If we get to this point and the field is a relation to another model,
|
|
|
|
# append the default ordering for that model.
|
|
|
|
if field.rel and len(joins) > 1 and opts.ordering:
|
|
|
|
# Firstly, avoid infinite loops.
|
|
|
|
if not already_seen:
|
|
|
|
already_seen = set()
|
2012-04-30 06:50:48 +08:00
|
|
|
join_tuple = tuple([self.query.alias_map[j].table_name for j in joins])
|
2009-12-22 23:18:51 +08:00
|
|
|
if join_tuple in already_seen:
|
|
|
|
raise FieldError('Infinite loop caused by ordering.')
|
|
|
|
already_seen.add(join_tuple)
|
|
|
|
|
|
|
|
results = []
|
|
|
|
for item in opts.ordering:
|
|
|
|
results.extend(self.find_ordering_name(item, opts, alias,
|
|
|
|
order, already_seen))
|
|
|
|
return results
|
2013-03-25 00:40:40 +08:00
|
|
|
cols, alias = self._final_join_removal(cols, alias)
|
|
|
|
return [(alias, cols, order)]
|
2011-12-23 04:42:40 +08:00
|
|
|
|
|
|
|
def _setup_joins(self, pieces, opts, alias):
|
|
|
|
"""
|
|
|
|
A helper method for get_ordering and get_distinct. This method will
|
|
|
|
call query.setup_joins, handle refcounts and then promote the joins.
|
|
|
|
|
|
|
|
Note that get_ordering and get_distinct must produce same target
|
|
|
|
columns on same input, as the prefixes of get_ordering and get_distinct
|
|
|
|
must match. Executing SQL where this is not true is an error.
|
|
|
|
"""
|
|
|
|
if not alias:
|
|
|
|
alias = self.query.get_initial_alias()
|
2013-03-25 00:40:40 +08:00
|
|
|
field, targets, opts, joins, _ = self.query.setup_joins(
|
2012-12-21 03:25:48 +08:00
|
|
|
pieces, opts, alias)
|
2012-05-24 04:25:23 +08:00
|
|
|
# We will later on need to promote those joins that were added to the
|
|
|
|
# query afresh above.
|
|
|
|
joins_to_promote = [j for j in joins if self.query.alias_refcount[j] < 2]
|
2011-12-23 04:42:40 +08:00
|
|
|
alias = joins[-1]
|
2013-03-25 00:40:40 +08:00
|
|
|
cols = [target.column for target in targets]
|
2011-12-23 04:42:40 +08:00
|
|
|
if not field.rel:
|
|
|
|
# To avoid inadvertent trimming of a necessary alias, use the
|
|
|
|
# refcount to show that we are referencing a non-relation field on
|
|
|
|
# the model.
|
|
|
|
self.query.ref_alias(alias)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
2011-12-23 04:42:40 +08:00
|
|
|
# Must use left outer joins for nullable fields and their relations.
|
|
|
|
# Ordering or distinct must not affect the returned set, and INNER
|
|
|
|
# JOINS for nullable fields could do this.
|
2012-08-22 02:54:14 +08:00
|
|
|
self.query.promote_joins(joins_to_promote)
|
2013-03-25 00:40:40 +08:00
|
|
|
return field, cols, alias, joins, opts
|
2011-12-23 04:42:40 +08:00
|
|
|
|
2013-03-25 00:40:40 +08:00
|
|
|
def _final_join_removal(self, cols, alias):
|
2011-12-23 04:42:40 +08:00
|
|
|
"""
|
|
|
|
A helper method for get_distinct and get_ordering. This method will
|
|
|
|
trim extra not-needed joins from the tail of the join chain.
|
|
|
|
|
|
|
|
This is very similar to what is done in trim_joins, but we will
|
|
|
|
trim LEFT JOINS here. It would be a good idea to consolidate this
|
|
|
|
method and query.trim_joins().
|
|
|
|
"""
|
2009-12-22 23:18:51 +08:00
|
|
|
if alias:
|
|
|
|
while 1:
|
|
|
|
join = self.query.alias_map[alias]
|
2013-03-25 00:40:40 +08:00
|
|
|
lhs_cols, rhs_cols = zip(*[(lhs_col, rhs_col) for lhs_col, rhs_col in join.join_cols])
|
|
|
|
if set(cols) != set(rhs_cols):
|
2009-12-22 23:18:51 +08:00
|
|
|
break
|
2013-03-25 00:40:40 +08:00
|
|
|
|
|
|
|
cols = [lhs_cols[rhs_cols.index(col)] for col in cols]
|
2009-12-22 23:18:51 +08:00
|
|
|
self.query.unref_alias(alias)
|
2012-04-30 06:50:48 +08:00
|
|
|
alias = join.lhs_alias
|
2013-03-25 00:40:40 +08:00
|
|
|
return cols, alias
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
def get_from_clause(self):
|
|
|
|
"""
|
|
|
|
Returns a list of strings that are joined together to go after the
|
|
|
|
"FROM" part of the query, as well as a list any extra parameters that
|
|
|
|
need to be included. Sub-classes, can override this to create a
|
|
|
|
from-clause via a "select".
|
|
|
|
|
|
|
|
This should only be called after any SQL construction methods that
|
2011-12-23 04:42:40 +08:00
|
|
|
might change the tables we need. This means the select columns,
|
|
|
|
ordering and distinct must be done first.
|
2009-12-22 23:18:51 +08:00
|
|
|
"""
|
|
|
|
result = []
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
|
|
|
first = True
|
2012-08-25 21:33:07 +08:00
|
|
|
from_params = []
|
2009-12-22 23:18:51 +08:00
|
|
|
for alias in self.query.tables:
|
|
|
|
if not self.query.alias_refcount[alias]:
|
|
|
|
continue
|
|
|
|
try:
|
2013-03-25 00:40:40 +08:00
|
|
|
name, alias, join_type, lhs, join_cols, _, join_field = self.query.alias_map[alias]
|
2009-12-22 23:18:51 +08:00
|
|
|
except KeyError:
|
|
|
|
# Extra tables can end up in self.tables, but not in the
|
|
|
|
# alias_map if they aren't in a join. That's OK. We skip them.
|
|
|
|
continue
|
2013-05-17 22:33:36 +08:00
|
|
|
alias_str = '' if alias == name else (' %s' % alias)
|
2009-12-22 23:18:51 +08:00
|
|
|
if join_type and not first:
|
2013-03-25 00:40:40 +08:00
|
|
|
extra_cond = join_field.get_extra_restriction(
|
|
|
|
self.query.where_class, alias, lhs)
|
|
|
|
if extra_cond:
|
|
|
|
extra_sql, extra_params = extra_cond.as_sql(
|
|
|
|
qn, self.connection)
|
|
|
|
extra_sql = 'AND (%s)' % extra_sql
|
2012-08-25 21:33:07 +08:00
|
|
|
from_params.extend(extra_params)
|
|
|
|
else:
|
2013-03-25 00:40:40 +08:00
|
|
|
extra_sql = ""
|
|
|
|
result.append('%s %s%s ON ('
|
|
|
|
% (join_type, qn(name), alias_str))
|
|
|
|
for index, (lhs_col, rhs_col) in enumerate(join_cols):
|
|
|
|
if index != 0:
|
|
|
|
result.append(' AND ')
|
|
|
|
result.append('%s.%s = %s.%s' %
|
|
|
|
(qn(lhs), qn2(lhs_col), qn(alias), qn2(rhs_col)))
|
|
|
|
result.append('%s)' % extra_sql)
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
2013-05-17 22:33:36 +08:00
|
|
|
connector = '' if first else ', '
|
2009-12-22 23:18:51 +08:00
|
|
|
result.append('%s%s%s' % (connector, qn(name), alias_str))
|
|
|
|
first = False
|
|
|
|
for t in self.query.extra_tables:
|
|
|
|
alias, unused = self.query.table_alias(t)
|
|
|
|
# Only add the alias if it's not already present (the table_alias()
|
|
|
|
# calls increments the refcount, so an alias refcount of one means
|
|
|
|
# this is the only reference.
|
|
|
|
if alias not in self.query.alias_map or self.query.alias_refcount[alias] == 1:
|
2013-05-17 22:33:36 +08:00
|
|
|
connector = '' if first else ', '
|
2009-12-22 23:18:51 +08:00
|
|
|
result.append('%s%s' % (connector, qn(alias)))
|
|
|
|
first = False
|
2012-08-25 21:33:07 +08:00
|
|
|
return result, from_params
|
2009-12-22 23:18:51 +08:00
|
|
|
|
Refactored qs.add_q() and utils/tree.py
The sql/query.py add_q method did a lot of where/having tree hacking to
get complex queries to work correctly. The logic was refactored so that
it should be simpler to understand. The new logic should also produce
leaner WHERE conditions.
The changes cascade somewhat, as some other parts of Django (like
add_filter() and WhereNode) expect boolean trees in certain format or
they fail to work. So to fix the add_q() one must fix utils/tree.py,
some things in add_filter(), WhereNode and so on.
This commit also fixed add_filter to see negate clauses up the path.
A query like .exclude(Q(reversefk__in=a_list)) didn't work similarly to
.filter(~Q(reversefk__in=a_list)). The reason for this is that only
the immediate parent negate clauses were seen by add_filter, and thus a
tree like AND: (NOT AND: (AND: condition)) will not be handled
correctly, as there is one intermediary AND node in the tree. The
example tree is generated by .exclude(~Q(reversefk__in=a_list)).
Still, aggregation lost connectors in OR cases, and F() objects and
aggregates in same filter clause caused GROUP BY problems on some
databases.
Fixed #17600, fixed #13198, fixed #17025, fixed #17000, fixed #11293.
2012-05-25 05:27:24 +08:00
|
|
|
def get_grouping(self, having_group_by, ordering_group_by):
|
2009-12-22 23:18:51 +08:00
|
|
|
"""
|
|
|
|
Returns a tuple representing the SQL elements in the "group by" clause.
|
|
|
|
"""
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
result, params = [], []
|
|
|
|
if self.query.group_by is not None:
|
2012-11-08 06:04:57 +08:00
|
|
|
select_cols = self.query.select + self.query.related_select_cols
|
|
|
|
# Just the column, not the fields.
|
|
|
|
select_cols = [s[0] for s in select_cols]
|
2013-05-14 03:40:39 +08:00
|
|
|
if (len(self.query.get_meta().concrete_fields) == len(self.query.select)
|
2012-11-08 06:04:57 +08:00
|
|
|
and self.connection.features.allows_group_by_pk):
|
2010-11-26 22:24:08 +08:00
|
|
|
self.query.group_by = [
|
2013-05-14 03:40:39 +08:00
|
|
|
(self.query.get_meta().db_table, self.query.get_meta().pk.column)
|
2010-11-26 22:24:08 +08:00
|
|
|
]
|
2012-11-08 06:04:57 +08:00
|
|
|
select_cols = []
|
2011-05-08 03:02:51 +08:00
|
|
|
seen = set()
|
Refactored qs.add_q() and utils/tree.py
The sql/query.py add_q method did a lot of where/having tree hacking to
get complex queries to work correctly. The logic was refactored so that
it should be simpler to understand. The new logic should also produce
leaner WHERE conditions.
The changes cascade somewhat, as some other parts of Django (like
add_filter() and WhereNode) expect boolean trees in certain format or
they fail to work. So to fix the add_q() one must fix utils/tree.py,
some things in add_filter(), WhereNode and so on.
This commit also fixed add_filter to see negate clauses up the path.
A query like .exclude(Q(reversefk__in=a_list)) didn't work similarly to
.filter(~Q(reversefk__in=a_list)). The reason for this is that only
the immediate parent negate clauses were seen by add_filter, and thus a
tree like AND: (NOT AND: (AND: condition)) will not be handled
correctly, as there is one intermediary AND node in the tree. The
example tree is generated by .exclude(~Q(reversefk__in=a_list)).
Still, aggregation lost connectors in OR cases, and F() objects and
aggregates in same filter clause caused GROUP BY problems on some
databases.
Fixed #17600, fixed #13198, fixed #17025, fixed #17000, fixed #11293.
2012-05-25 05:27:24 +08:00
|
|
|
cols = self.query.group_by + having_group_by + select_cols
|
2010-11-26 22:24:08 +08:00
|
|
|
for col in cols:
|
2013-02-10 23:15:49 +08:00
|
|
|
col_params = ()
|
2009-12-22 23:18:51 +08:00
|
|
|
if isinstance(col, (list, tuple)):
|
2012-11-08 06:04:57 +08:00
|
|
|
sql = '%s.%s' % (qn(col[0]), qn(col[1]))
|
2009-12-22 23:18:51 +08:00
|
|
|
elif hasattr(col, 'as_sql'):
|
2013-02-10 23:15:49 +08:00
|
|
|
sql, col_params = col.as_sql(qn, self.connection)
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
2012-11-08 06:04:57 +08:00
|
|
|
sql = '(%s)' % str(col)
|
|
|
|
if sql not in seen:
|
|
|
|
result.append(sql)
|
2013-02-10 23:15:49 +08:00
|
|
|
params.extend(col_params)
|
2012-11-08 06:04:57 +08:00
|
|
|
seen.add(sql)
|
|
|
|
|
|
|
|
# Still, we need to add all stuff in ordering (except if the backend can
|
|
|
|
# group by just by PK).
|
|
|
|
if ordering_group_by and not self.connection.features.allows_group_by_pk:
|
|
|
|
for order, order_params in ordering_group_by:
|
|
|
|
# Even if we have seen the same SQL string, it might have
|
|
|
|
# different params, so, we add same SQL in "has params" case.
|
|
|
|
if order not in seen or params:
|
|
|
|
result.append(order)
|
|
|
|
params.extend(order_params)
|
|
|
|
seen.add(order)
|
|
|
|
|
|
|
|
# Unconditionally add the extra_select items.
|
|
|
|
for extra_select, extra_params in self.query.extra_select.values():
|
|
|
|
sql = '(%s)' % str(extra_select)
|
|
|
|
result.append(sql)
|
|
|
|
params.extend(extra_params)
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
return result, params
|
|
|
|
|
|
|
|
def fill_related_selections(self, opts=None, root_alias=None, cur_depth=1,
|
2012-08-11 03:00:21 +08:00
|
|
|
requested=None, restricted=None, nullable=None):
|
2009-12-22 23:18:51 +08:00
|
|
|
"""
|
|
|
|
Fill in the information needed for a select_related query. The current
|
|
|
|
depth is measured as the number of connections away from the root model
|
|
|
|
(for example, cur_depth=1 means we are looking at models with direct
|
|
|
|
connections to the root model).
|
|
|
|
"""
|
|
|
|
if not restricted and self.query.max_depth and cur_depth > self.query.max_depth:
|
|
|
|
# We've recursed far enough; bail out.
|
|
|
|
return
|
|
|
|
|
|
|
|
if not opts:
|
|
|
|
opts = self.query.get_meta()
|
|
|
|
root_alias = self.query.get_initial_alias()
|
|
|
|
self.query.related_select_cols = []
|
2012-06-26 23:08:42 +08:00
|
|
|
only_load = self.query.get_loaded_field_names()
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
# Setup for the case when only particular related fields should be
|
|
|
|
# included in the related selection.
|
2010-01-27 21:30:29 +08:00
|
|
|
if requested is None:
|
2009-12-22 23:18:51 +08:00
|
|
|
if isinstance(self.query.select_related, dict):
|
|
|
|
requested = self.query.select_related
|
|
|
|
restricted = True
|
|
|
|
else:
|
|
|
|
restricted = False
|
|
|
|
|
|
|
|
for f, model in opts.get_fields_with_model():
|
2012-09-16 02:11:14 +08:00
|
|
|
# The get_fields_with_model() returns None for fields that live
|
|
|
|
# in the field's local model. So, for those fields we want to use
|
|
|
|
# the f.model - that is the field's local model.
|
|
|
|
field_model = model or f.model
|
2012-06-26 23:08:42 +08:00
|
|
|
if not select_related_descend(f, restricted, requested,
|
2012-09-16 02:11:14 +08:00
|
|
|
only_load.get(field_model)):
|
2009-12-22 23:18:51 +08:00
|
|
|
continue
|
|
|
|
table = f.rel.to._meta.db_table
|
2010-03-08 20:47:16 +08:00
|
|
|
promote = nullable or f.null
|
2012-12-17 20:02:41 +08:00
|
|
|
alias = self.query.join_parent_model(opts, model, root_alias, {})
|
2013-03-25 00:40:40 +08:00
|
|
|
join_cols = f.get_joining_columns()
|
|
|
|
alias = self.query.join((alias, table, join_cols),
|
2013-02-19 10:47:47 +08:00
|
|
|
outer_if_first=promote, join_field=f)
|
2009-12-22 23:18:51 +08:00
|
|
|
columns, aliases = self.get_default_columns(start_alias=alias,
|
|
|
|
opts=f.rel.to._meta, as_pairs=True)
|
2012-10-26 01:57:32 +08:00
|
|
|
self.query.related_select_cols.extend(
|
2013-03-25 00:40:40 +08:00
|
|
|
SelectInfo(col, field) for col, field in zip(columns, f.rel.to._meta.concrete_fields))
|
2009-12-22 23:18:51 +08:00
|
|
|
if restricted:
|
|
|
|
next = requested.get(f.name, {})
|
|
|
|
else:
|
|
|
|
next = False
|
2010-03-08 20:47:16 +08:00
|
|
|
new_nullable = f.null or promote
|
2009-12-22 23:18:51 +08:00
|
|
|
self.fill_related_selections(f.rel.to._meta, alias, cur_depth + 1,
|
2012-08-11 03:00:21 +08:00
|
|
|
next, restricted, new_nullable)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
2010-01-27 21:30:29 +08:00
|
|
|
if restricted:
|
|
|
|
related_fields = [
|
|
|
|
(o.field, o.model)
|
|
|
|
for o in opts.get_all_related_objects()
|
|
|
|
if o.field.unique
|
|
|
|
]
|
|
|
|
for f, model in related_fields:
|
2012-06-26 23:08:42 +08:00
|
|
|
if not select_related_descend(f, restricted, requested,
|
|
|
|
only_load.get(model), reverse=True):
|
2010-01-27 21:30:29 +08:00
|
|
|
continue
|
|
|
|
|
2012-12-17 20:02:41 +08:00
|
|
|
alias = self.query.join_parent_model(opts, f.rel.to, root_alias, {})
|
2012-08-11 03:00:21 +08:00
|
|
|
table = model._meta.db_table
|
2010-01-27 21:30:29 +08:00
|
|
|
alias = self.query.join(
|
2013-03-25 00:40:40 +08:00
|
|
|
(alias, table, f.get_joining_columns(reverse_join=True)),
|
2013-02-19 10:47:47 +08:00
|
|
|
outer_if_first=True, join_field=f
|
2010-01-27 21:30:29 +08:00
|
|
|
)
|
2012-11-10 02:21:46 +08:00
|
|
|
from_parent = (opts.model if issubclass(model, opts.model)
|
|
|
|
else None)
|
2010-01-27 21:30:29 +08:00
|
|
|
columns, aliases = self.get_default_columns(start_alias=alias,
|
2012-11-10 02:21:46 +08:00
|
|
|
opts=model._meta, as_pairs=True, from_parent=from_parent)
|
2012-10-26 01:57:32 +08:00
|
|
|
self.query.related_select_cols.extend(
|
2012-11-10 02:21:46 +08:00
|
|
|
SelectInfo(col, field) for col, field
|
2013-03-25 00:40:40 +08:00
|
|
|
in zip(columns, model._meta.concrete_fields))
|
2010-01-27 21:30:29 +08:00
|
|
|
next = requested.get(f.related_query_name(), {})
|
2012-08-22 02:54:14 +08:00
|
|
|
# Use True here because we are looking at the _reverse_ side of
|
|
|
|
# the relation, which is always nullable.
|
|
|
|
new_nullable = True
|
2010-01-27 21:30:29 +08:00
|
|
|
|
|
|
|
self.fill_related_selections(model._meta, table, cur_depth+1,
|
2012-08-11 03:00:21 +08:00
|
|
|
next, restricted, new_nullable)
|
2010-01-27 21:30:29 +08:00
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
def deferred_to_columns(self):
|
|
|
|
"""
|
|
|
|
Converts the self.deferred_loading data structure to mapping of table
|
|
|
|
names to sets of column names which are to be loaded. Returns the
|
|
|
|
dictionary.
|
|
|
|
"""
|
|
|
|
columns = {}
|
|
|
|
self.query.deferred_to_data(columns, self.query.deferred_to_columns_cb)
|
|
|
|
return columns
|
|
|
|
|
|
|
|
def results_iter(self):
|
|
|
|
"""
|
|
|
|
Returns an iterator over the results from executing this query.
|
|
|
|
"""
|
|
|
|
resolve_columns = hasattr(self, 'resolve_columns')
|
|
|
|
fields = None
|
2010-11-23 21:54:58 +08:00
|
|
|
has_aggregate_select = bool(self.query.aggregate_select)
|
2009-12-22 23:18:51 +08:00
|
|
|
for rows in self.execute_sql(MULTI):
|
|
|
|
for row in rows:
|
|
|
|
if resolve_columns:
|
|
|
|
if fields is None:
|
|
|
|
# We only set this up here because
|
2012-10-26 01:57:32 +08:00
|
|
|
# related_select_cols isn't populated until
|
2009-12-22 23:18:51 +08:00
|
|
|
# execute_sql() has been called.
|
2012-10-26 07:25:59 +08:00
|
|
|
|
|
|
|
# We also include types of fields of related models that
|
|
|
|
# will be included via select_related() for the benefit
|
|
|
|
# of MySQL/MySQLdb when boolean fields are involved
|
|
|
|
# (#15040).
|
|
|
|
|
|
|
|
# This code duplicates the logic for the order of fields
|
|
|
|
# found in get_columns(). It would be nice to clean this up.
|
2012-10-26 01:57:32 +08:00
|
|
|
if self.query.select:
|
|
|
|
fields = [f.field for f in self.query.select]
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
2013-05-14 03:40:39 +08:00
|
|
|
fields = self.query.get_meta().concrete_fields
|
2012-10-26 01:57:32 +08:00
|
|
|
fields = fields + [f.field for f in self.query.related_select_cols]
|
2012-10-26 07:25:59 +08:00
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
# If the field was deferred, exclude it from being passed
|
|
|
|
# into `resolve_columns` because it wasn't selected.
|
|
|
|
only_load = self.deferred_to_columns()
|
|
|
|
if only_load:
|
2013-05-14 03:40:39 +08:00
|
|
|
db_table = self.query.get_meta().db_table
|
2009-12-22 23:18:51 +08:00
|
|
|
fields = [f for f in fields if db_table in only_load and
|
|
|
|
f.column in only_load[db_table]]
|
|
|
|
row = self.resolve_columns(row, fields)
|
|
|
|
|
2010-11-23 21:54:58 +08:00
|
|
|
if has_aggregate_select:
|
2012-08-14 20:09:23 +08:00
|
|
|
aggregate_start = len(self.query.extra_select) + len(self.query.select)
|
2009-12-22 23:18:51 +08:00
|
|
|
aggregate_end = aggregate_start + len(self.query.aggregate_select)
|
|
|
|
row = tuple(row[:aggregate_start]) + tuple([
|
|
|
|
self.query.resolve_aggregate(value, aggregate, self.connection)
|
|
|
|
for (alias, aggregate), value
|
|
|
|
in zip(self.query.aggregate_select.items(), row[aggregate_start:aggregate_end])
|
|
|
|
]) + tuple(row[aggregate_end:])
|
|
|
|
|
|
|
|
yield row
|
|
|
|
|
|
|
|
def execute_sql(self, result_type=MULTI):
|
|
|
|
"""
|
|
|
|
Run the query against the database and returns the result(s). The
|
|
|
|
return value is a single data item if result_type is SINGLE, or an
|
|
|
|
iterator over the results if the result_type is MULTI.
|
|
|
|
|
|
|
|
result_type is either MULTI (use fetchmany() to retrieve all rows),
|
|
|
|
SINGLE (only retrieve a single row), or None. In this last case, the
|
|
|
|
cursor is returned if any query is executed, since it's used by
|
|
|
|
subclasses such as InsertQuery). It's possible, however, that no query
|
|
|
|
is needed, as the filters describe an empty set. In that case, None is
|
|
|
|
returned, to avoid any unnecessary database interaction.
|
|
|
|
"""
|
|
|
|
try:
|
|
|
|
sql, params = self.as_sql()
|
|
|
|
if not sql:
|
|
|
|
raise EmptyResultSet
|
|
|
|
except EmptyResultSet:
|
|
|
|
if result_type == MULTI:
|
2012-09-07 22:58:17 +08:00
|
|
|
return iter([])
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
|
|
|
return
|
|
|
|
|
|
|
|
cursor = self.connection.cursor()
|
|
|
|
cursor.execute(sql, params)
|
|
|
|
|
|
|
|
if not result_type:
|
|
|
|
return cursor
|
|
|
|
if result_type == SINGLE:
|
|
|
|
if self.query.ordering_aliases:
|
|
|
|
return cursor.fetchone()[:-len(self.query.ordering_aliases)]
|
|
|
|
return cursor.fetchone()
|
|
|
|
|
|
|
|
# The MULTI case.
|
|
|
|
if self.query.ordering_aliases:
|
|
|
|
result = order_modified_iter(cursor, len(self.query.ordering_aliases),
|
|
|
|
self.connection.features.empty_fetchmany_value)
|
|
|
|
else:
|
|
|
|
result = iter((lambda: cursor.fetchmany(GET_ITERATOR_CHUNK_SIZE)),
|
|
|
|
self.connection.features.empty_fetchmany_value)
|
|
|
|
if not self.connection.features.can_use_chunked_reads:
|
|
|
|
# If we are using non-chunked reads, we return the same data
|
|
|
|
# structure as normally, but ensure it is all read into memory
|
|
|
|
# before going any further.
|
|
|
|
return list(result)
|
|
|
|
return result
|
|
|
|
|
2013-03-25 00:40:40 +08:00
|
|
|
def as_subquery_condition(self, alias, columns):
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
qn2 = self.connection.ops.quote_name
|
|
|
|
if len(columns) == 1:
|
|
|
|
sql, params = self.as_sql()
|
|
|
|
return '%s.%s IN (%s)' % (qn(alias), qn2(columns[0]), sql), params
|
|
|
|
|
|
|
|
for index, select_col in enumerate(self.query.select):
|
|
|
|
lhs = '%s.%s' % (qn(select_col.col[0]), qn2(select_col.col[1]))
|
|
|
|
rhs = '%s.%s' % (qn(alias), qn2(columns[index]))
|
|
|
|
self.query.where.add(
|
|
|
|
QueryWrapper('%s = %s' % (lhs, rhs), []), 'AND')
|
|
|
|
|
|
|
|
sql, params = self.as_sql()
|
|
|
|
return 'EXISTS (%s)' % sql, params
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
class SQLInsertCompiler(SQLCompiler):
|
2013-04-09 03:28:15 +08:00
|
|
|
|
|
|
|
def __init__(self, *args, **kwargs):
|
|
|
|
self.return_id = False
|
|
|
|
super(SQLInsertCompiler, self).__init__(*args, **kwargs)
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
def placeholder(self, field, val):
|
|
|
|
if field is None:
|
|
|
|
# A field value of None means the value is raw.
|
|
|
|
return val
|
|
|
|
elif hasattr(field, 'get_placeholder'):
|
|
|
|
# Some fields (e.g. geo fields) need special munging before
|
|
|
|
# they can be inserted.
|
|
|
|
return field.get_placeholder(val, self.connection)
|
|
|
|
else:
|
|
|
|
# Return the common case for the placeholder
|
|
|
|
return '%s'
|
|
|
|
|
|
|
|
def as_sql(self):
|
|
|
|
# We don't need quote_name_unless_alias() here, since these are all
|
|
|
|
# going to be column names (so we can avoid the extra overhead).
|
|
|
|
qn = self.connection.ops.quote_name
|
2013-05-14 03:40:39 +08:00
|
|
|
opts = self.query.get_meta()
|
2009-12-22 23:18:51 +08:00
|
|
|
result = ['INSERT INTO %s' % qn(opts.db_table)]
|
2011-09-10 03:22:28 +08:00
|
|
|
|
|
|
|
has_fields = bool(self.query.fields)
|
|
|
|
fields = self.query.fields if has_fields else [opts.pk]
|
|
|
|
result.append('(%s)' % ', '.join([qn(f.column) for f in fields]))
|
|
|
|
|
|
|
|
if has_fields:
|
|
|
|
params = values = [
|
|
|
|
[
|
|
|
|
f.get_db_prep_save(getattr(obj, f.attname) if self.query.raw else f.pre_save(obj, True), connection=self.connection)
|
|
|
|
for f in fields
|
|
|
|
]
|
|
|
|
for obj in self.query.objs
|
|
|
|
]
|
|
|
|
else:
|
|
|
|
values = [[self.connection.ops.pk_default_value()] for obj in self.query.objs]
|
|
|
|
params = [[]]
|
|
|
|
fields = [None]
|
2011-09-11 06:02:13 +08:00
|
|
|
can_bulk = (not any(hasattr(field, "get_placeholder") for field in fields) and
|
|
|
|
not self.return_id and self.connection.features.has_bulk_insert)
|
2011-09-10 03:22:28 +08:00
|
|
|
|
|
|
|
if can_bulk:
|
|
|
|
placeholders = [["%s"] * len(fields)]
|
|
|
|
else:
|
|
|
|
placeholders = [
|
2012-05-07 23:25:12 +08:00
|
|
|
[self.placeholder(field, v) for field, v in zip(fields, val)]
|
2011-09-10 03:22:28 +08:00
|
|
|
for val in values
|
|
|
|
]
|
2012-11-15 20:23:02 +08:00
|
|
|
# Oracle Spatial needs to remove some values due to #10888
|
|
|
|
params = self.connection.ops.modify_insert_params(placeholders, params)
|
2009-12-22 23:18:51 +08:00
|
|
|
if self.return_id and self.connection.features.can_return_id_from_insert:
|
2011-10-17 04:06:59 +08:00
|
|
|
params = params[0]
|
2009-12-22 23:18:51 +08:00
|
|
|
col = "%s.%s" % (qn(opts.db_table), qn(opts.pk.column))
|
2011-09-10 03:22:28 +08:00
|
|
|
result.append("VALUES (%s)" % ", ".join(placeholders[0]))
|
2009-12-22 23:18:51 +08:00
|
|
|
r_fmt, r_params = self.connection.ops.return_insert_id()
|
2012-10-10 05:06:37 +08:00
|
|
|
# Skip empty r_fmt to allow subclasses to customize behaviour for
|
|
|
|
# 3rd party backends. Refs #19096.
|
|
|
|
if r_fmt:
|
|
|
|
result.append(r_fmt % col)
|
|
|
|
params += r_params
|
2011-09-10 03:22:28 +08:00
|
|
|
return [(" ".join(result), tuple(params))]
|
2011-09-11 06:02:13 +08:00
|
|
|
if can_bulk:
|
2011-09-10 03:22:28 +08:00
|
|
|
result.append(self.connection.ops.bulk_insert_sql(fields, len(values)))
|
|
|
|
return [(" ".join(result), tuple([v for val in values for v in val]))]
|
|
|
|
else:
|
|
|
|
return [
|
|
|
|
(" ".join(result + ["VALUES (%s)" % ", ".join(p)]), vals)
|
2012-05-07 23:25:12 +08:00
|
|
|
for p, vals in zip(placeholders, params)
|
2011-09-10 03:22:28 +08:00
|
|
|
]
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
def execute_sql(self, return_id=False):
|
2011-09-10 03:22:28 +08:00
|
|
|
assert not (return_id and len(self.query.objs) != 1)
|
2009-12-22 23:18:51 +08:00
|
|
|
self.return_id = return_id
|
2011-09-10 03:22:28 +08:00
|
|
|
cursor = self.connection.cursor()
|
|
|
|
for sql, params in self.as_sql():
|
|
|
|
cursor.execute(sql, params)
|
2009-12-22 23:18:51 +08:00
|
|
|
if not (return_id and cursor):
|
|
|
|
return
|
|
|
|
if self.connection.features.can_return_id_from_insert:
|
|
|
|
return self.connection.ops.fetch_returned_insert_id(cursor)
|
|
|
|
return self.connection.ops.last_insert_id(cursor,
|
2013-05-14 03:40:39 +08:00
|
|
|
self.query.get_meta().db_table, self.query.get_meta().pk.column)
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
|
|
|
|
class SQLDeleteCompiler(SQLCompiler):
|
|
|
|
def as_sql(self):
|
|
|
|
"""
|
|
|
|
Creates the SQL for this query. Returns the SQL string and list of
|
|
|
|
parameters.
|
|
|
|
"""
|
|
|
|
assert len(self.query.tables) == 1, \
|
|
|
|
"Can only delete from one table at a time."
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
result = ['DELETE FROM %s' % qn(self.query.tables[0])]
|
|
|
|
where, params = self.query.where.as_sql(qn=qn, connection=self.connection)
|
2012-09-20 23:51:30 +08:00
|
|
|
if where:
|
|
|
|
result.append('WHERE %s' % where)
|
2009-12-22 23:18:51 +08:00
|
|
|
return ' '.join(result), tuple(params)
|
|
|
|
|
|
|
|
class SQLUpdateCompiler(SQLCompiler):
|
|
|
|
def as_sql(self):
|
|
|
|
"""
|
|
|
|
Creates the SQL for this query. Returns the SQL string and list of
|
|
|
|
parameters.
|
|
|
|
"""
|
|
|
|
self.pre_sql_setup()
|
|
|
|
if not self.query.values:
|
|
|
|
return '', ()
|
|
|
|
table = self.query.tables[0]
|
|
|
|
qn = self.quote_name_unless_alias
|
|
|
|
result = ['UPDATE %s' % qn(table)]
|
|
|
|
result.append('SET')
|
|
|
|
values, update_params = [], []
|
|
|
|
for field, model, val in self.query.values:
|
|
|
|
if hasattr(val, 'prepare_database_save'):
|
|
|
|
val = val.prepare_database_save(field)
|
|
|
|
else:
|
|
|
|
val = field.get_db_prep_save(val, connection=self.connection)
|
|
|
|
|
|
|
|
# Getting the placeholder for the field.
|
|
|
|
if hasattr(field, 'get_placeholder'):
|
|
|
|
placeholder = field.get_placeholder(val, self.connection)
|
|
|
|
else:
|
|
|
|
placeholder = '%s'
|
|
|
|
|
|
|
|
if hasattr(val, 'evaluate'):
|
|
|
|
val = SQLEvaluator(val, self.query, allow_joins=False)
|
|
|
|
name = field.column
|
|
|
|
if hasattr(val, 'as_sql'):
|
|
|
|
sql, params = val.as_sql(qn, self.connection)
|
|
|
|
values.append('%s = %s' % (qn(name), sql))
|
|
|
|
update_params.extend(params)
|
|
|
|
elif val is not None:
|
|
|
|
values.append('%s = %s' % (qn(name), placeholder))
|
|
|
|
update_params.append(val)
|
|
|
|
else:
|
|
|
|
values.append('%s = NULL' % qn(name))
|
|
|
|
if not values:
|
|
|
|
return '', ()
|
|
|
|
result.append(', '.join(values))
|
|
|
|
where, params = self.query.where.as_sql(qn=qn, connection=self.connection)
|
|
|
|
if where:
|
|
|
|
result.append('WHERE %s' % where)
|
|
|
|
return ' '.join(result), tuple(update_params + params)
|
|
|
|
|
|
|
|
def execute_sql(self, result_type):
|
|
|
|
"""
|
|
|
|
Execute the specified update. Returns the number of rows affected by
|
|
|
|
the primary update query. The "primary update query" is the first
|
|
|
|
non-empty query that is executed. Row counts for any subsequent,
|
|
|
|
related queries are not available.
|
|
|
|
"""
|
|
|
|
cursor = super(SQLUpdateCompiler, self).execute_sql(result_type)
|
2013-05-17 22:33:36 +08:00
|
|
|
rows = cursor.rowcount if cursor else 0
|
2009-12-22 23:18:51 +08:00
|
|
|
is_empty = cursor is None
|
|
|
|
del cursor
|
|
|
|
for query in self.query.get_related_updates():
|
|
|
|
aux_rows = query.get_compiler(self.using).execute_sql(result_type)
|
|
|
|
if is_empty:
|
|
|
|
rows = aux_rows
|
|
|
|
is_empty = False
|
|
|
|
return rows
|
|
|
|
|
|
|
|
def pre_sql_setup(self):
|
|
|
|
"""
|
|
|
|
If the update depends on results from other tables, we need to do some
|
|
|
|
munging of the "where" conditions to match the format required for
|
|
|
|
(portable) SQL updates. That is done here.
|
|
|
|
|
|
|
|
Further, if we are going to be running multiple updates, we pull out
|
|
|
|
the id values to update at this point so that they don't change as a
|
|
|
|
result of the progressive updates.
|
|
|
|
"""
|
|
|
|
self.query.select_related = False
|
|
|
|
self.query.clear_ordering(True)
|
|
|
|
super(SQLUpdateCompiler, self).pre_sql_setup()
|
|
|
|
count = self.query.count_active_tables()
|
|
|
|
if not self.query.related_updates and count == 1:
|
|
|
|
return
|
|
|
|
|
|
|
|
# We need to use a sub-select in the where clause to filter on things
|
|
|
|
# from other tables.
|
|
|
|
query = self.query.clone(klass=Query)
|
|
|
|
query.bump_prefix()
|
|
|
|
query.extra = {}
|
|
|
|
query.select = []
|
2013-05-14 03:40:39 +08:00
|
|
|
query.add_fields([query.get_meta().pk.name])
|
2012-05-12 18:01:45 +08:00
|
|
|
# Recheck the count - it is possible that fiddling with the select
|
|
|
|
# fields above removes tables from the query. Refs #18304.
|
|
|
|
count = query.count_active_tables()
|
|
|
|
if not self.query.related_updates and count == 1:
|
|
|
|
return
|
|
|
|
|
2009-12-22 23:18:51 +08:00
|
|
|
must_pre_select = count > 1 and not self.connection.features.update_can_self_select
|
|
|
|
|
|
|
|
# Now we adjust the current query: reset the where clause and get rid
|
|
|
|
# of all the tables we don't need (since they're in the sub-select).
|
|
|
|
self.query.where = self.query.where_class()
|
|
|
|
if self.query.related_updates or must_pre_select:
|
|
|
|
# Either we're using the idents in multiple update queries (so
|
|
|
|
# don't want them to change), or the db backend doesn't support
|
|
|
|
# selecting from the updating table (e.g. MySQL).
|
|
|
|
idents = []
|
|
|
|
for rows in query.get_compiler(self.using).execute_sql(MULTI):
|
|
|
|
idents.extend([r[0] for r in rows])
|
|
|
|
self.query.add_filter(('pk__in', idents))
|
|
|
|
self.query.related_ids = idents
|
|
|
|
else:
|
|
|
|
# The fast path. Filters and updates in one query.
|
|
|
|
self.query.add_filter(('pk__in', query))
|
|
|
|
for alias in self.query.tables[1:]:
|
|
|
|
self.query.alias_refcount[alias] = 0
|
|
|
|
|
|
|
|
class SQLAggregateCompiler(SQLCompiler):
|
|
|
|
def as_sql(self, qn=None):
|
|
|
|
"""
|
|
|
|
Creates the SQL for this query. Returns the SQL string and list of
|
|
|
|
parameters.
|
|
|
|
"""
|
|
|
|
if qn is None:
|
|
|
|
qn = self.quote_name_unless_alias
|
2011-12-23 04:42:40 +08:00
|
|
|
|
2013-02-10 23:15:49 +08:00
|
|
|
sql, params = [], []
|
|
|
|
for aggregate in self.query.aggregate_select.values():
|
|
|
|
agg_sql, agg_params = aggregate.as_sql(qn, self.connection)
|
|
|
|
sql.append(agg_sql)
|
|
|
|
params.extend(agg_params)
|
|
|
|
sql = ', '.join(sql)
|
|
|
|
params = tuple(params)
|
|
|
|
|
|
|
|
sql = 'SELECT %s FROM (%s) subquery' % (sql, self.query.subquery)
|
|
|
|
params = params + self.query.sub_params
|
|
|
|
return sql, params
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
class SQLDateCompiler(SQLCompiler):
|
|
|
|
def results_iter(self):
|
|
|
|
"""
|
|
|
|
Returns an iterator over the results from executing this query.
|
|
|
|
"""
|
|
|
|
resolve_columns = hasattr(self, 'resolve_columns')
|
|
|
|
if resolve_columns:
|
2013-02-10 23:15:49 +08:00
|
|
|
from django.db.models.fields import DateField
|
|
|
|
fields = [DateField()]
|
2009-12-22 23:18:51 +08:00
|
|
|
else:
|
2013-02-10 23:15:49 +08:00
|
|
|
from django.db.backends.util import typecast_date
|
2009-12-22 23:18:51 +08:00
|
|
|
needs_string_cast = self.connection.features.needs_datetime_string_cast
|
|
|
|
|
|
|
|
offset = len(self.query.extra_select)
|
|
|
|
for rows in self.execute_sql(MULTI):
|
|
|
|
for row in rows:
|
|
|
|
date = row[offset]
|
|
|
|
if resolve_columns:
|
|
|
|
date = self.resolve_columns(row, fields)[offset]
|
|
|
|
elif needs_string_cast:
|
2013-02-10 23:15:49 +08:00
|
|
|
date = typecast_date(str(date))
|
|
|
|
if isinstance(date, datetime.datetime):
|
|
|
|
date = date.date()
|
2009-12-22 23:18:51 +08:00
|
|
|
yield date
|
|
|
|
|
2013-02-10 23:15:49 +08:00
|
|
|
class SQLDateTimeCompiler(SQLCompiler):
|
|
|
|
def results_iter(self):
|
|
|
|
"""
|
|
|
|
Returns an iterator over the results from executing this query.
|
|
|
|
"""
|
|
|
|
resolve_columns = hasattr(self, 'resolve_columns')
|
|
|
|
if resolve_columns:
|
|
|
|
from django.db.models.fields import DateTimeField
|
|
|
|
fields = [DateTimeField()]
|
|
|
|
else:
|
|
|
|
from django.db.backends.util import typecast_timestamp
|
|
|
|
needs_string_cast = self.connection.features.needs_datetime_string_cast
|
|
|
|
|
|
|
|
offset = len(self.query.extra_select)
|
|
|
|
for rows in self.execute_sql(MULTI):
|
|
|
|
for row in rows:
|
|
|
|
datetime = row[offset]
|
|
|
|
if resolve_columns:
|
|
|
|
datetime = self.resolve_columns(row, fields)[offset]
|
|
|
|
elif needs_string_cast:
|
|
|
|
datetime = typecast_timestamp(str(datetime))
|
|
|
|
# Datetimes are artifically returned in UTC on databases that
|
|
|
|
# don't support time zone. Restore the zone used in the query.
|
|
|
|
if settings.USE_TZ:
|
2013-03-01 00:35:13 +08:00
|
|
|
if datetime is None:
|
|
|
|
raise ValueError("Database returned an invalid value "
|
|
|
|
"in QuerySet.dates(). Are time zone "
|
2013-05-08 18:57:35 +08:00
|
|
|
"definitions and pytz installed?")
|
2013-02-10 23:15:49 +08:00
|
|
|
datetime = datetime.replace(tzinfo=None)
|
|
|
|
datetime = timezone.make_aware(datetime, self.query.tzinfo)
|
|
|
|
yield datetime
|
2009-12-22 23:18:51 +08:00
|
|
|
|
|
|
|
def order_modified_iter(cursor, trim, sentinel):
|
|
|
|
"""
|
|
|
|
Yields blocks of rows from a cursor. We use this iterator in the special
|
|
|
|
case when extra output columns have been added to support ordering
|
|
|
|
requirements. We must trim those extra columns before anything else can use
|
|
|
|
the results, since they're only needed to make the SQL valid.
|
|
|
|
"""
|
|
|
|
for rows in iter((lambda: cursor.fetchmany(GET_ITERATOR_CHUNK_SIZE)),
|
|
|
|
sentinel):
|
|
|
|
yield [r[:-trim] for r in rows]
|