Fixed #16187 -- refactored ORM lookup system

Allowed users to specify which lookups or transforms ("nested lookus")
are available for fields. The implementation is now class based.

Squashed commit of the following:

commit fa7a7195f1952a9c8dea7f6e89ee13f81757eda7
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 18 10:53:24 2014 +0200

    Added lookup registration API docs

commit eb1c8ce164325e0d8641f14202e12486c70efdb6
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Tue Jan 14 18:59:36 2014 +0200

    Release notes and other minor docs changes

commit 11501c29c9352d17f22f3a0f59d3b805913dedcc
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Jan 12 20:53:03 2014 +0200

    Forgot to add custom_lookups tests in prev commit

commit 83173b960ea7eb2b24d573f326be59948df33536
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Jan 12 19:59:12 2014 +0200

    Renamed Extract -> Transform

commit 3b18d9f3a1bcdd93280f79654eba0efa209377bd
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Jan 12 19:51:53 2014 +0200

    Removed suggestion of temporary lookup registration from docs

commit 21d0c7631c161fc0c67911480be5d3f13f1afa68
Merge: 2509006 f2dc442
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Jan 12 09:38:23 2014 -0800

    Merge pull request #2 from mjtamlyn/lookups_3

    Reworked custom lookups docs.

commit f2dc4429a1da04c858364972eea57a35a868dab4
Author: Marc Tamlyn <marc.tamlyn@gmail.com>
Date:   Sun Jan 12 13:15:05 2014 +0000

    Reworked custom lookups docs.

    Mostly just formatting and rewording, but also replaced the example
    using ``YearExtract`` to  use an example which is unlikely to ever be
    possible directly in the ORM.

commit 250900650628d1f11beadb22814abd666029fb81
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Jan 12 13:19:13 2014 +0200

    Removed unused import

commit 4fba5dfaa022653ffa72497258ffd8f8b7476f92
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 22:34:41 2014 +0200

    Added docs to index

commit 6d53963f375c77a1f287833b19b976d23f36c30b
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 22:10:24 2014 +0200

    Dead code removal

commit f9cc0390078e21f1ea5a7bc1f15b09f8f6b0904d
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 19:00:43 2014 +0200

    A new try for docs

commit 33aa18a6e3c831930bda0028222a26f9c1d96e66
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 14:57:12 2014 +0200

    Renamed get_cols to get_group_by_cols

commit c7d5f8661b7d364962bed2e6f81161c1b4f1bcc3
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 14:45:53 2014 +0200

    Altered query string customization for backends vendors

    The new way is trying to call first method 'as_' + connection.vendor.
    If that doesn't exist, then call as_sql().

    Also altered how lookup registration is done. There is now
    RegisterLookupMixin class that is used by Field, Extract and
    sql.Aggregate. This allows one to register lookups for extracts and
    aggregates in the same way lookup registration is done for fields.

commit 90e7004ec14e15503f828cc9bde2a7dab593814d
Merge: 66649ff f7c2c0a
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 13:21:01 2014 +0200

    Merge branch 'master' into lookups_3

commit 66649ff891c7c73c7eecf6038c9a6802611b5d8a
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Jan 11 13:16:01 2014 +0200

    Some rewording in docs

commit 31b8faa62714b4b6b6057a9f5cc106c4dd73caab
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 29 15:52:29 2013 +0200

    Cleanup based on review comments

commit 1016159f34674c0df871ed891cde72be8340bb5d
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 28 18:37:04 2013 +0200

    Proof-of-concept fix for #16731

    Implemented only for SQLite and PostgreSQL, and only for startswith
    and istartswith lookups.

commit 193cd097ca8f2cc6a911e57b8e3fb726f96ee6a6
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 28 17:57:58 2013 +0200

    Fixed #11722 -- iexact=F() produced invalid SQL

commit 08ed3c3b49e100ed9019831e770c25c8f61b70f9
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 21 23:59:52 2013 +0200

    Made Lookup and Extract available from django.db.models

commit b99c8d83c972786c6fcd0e84c9e5cb08c1368300
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 21 23:06:29 2013 +0200

    Fixed review notes by Loic

commit 049eebc0703c151127f4f0265beceea7b8b39e72
Merge: ed8fab7 b80a835
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 21 22:53:10 2013 +0200

    Merge branch 'master' into lookups_3

    Conflicts:
    	django/db/models/fields/__init__.py
    	django/db/models/sql/compiler.py
    	django/db/models/sql/query.py
    	tests/null_queries/tests.py

commit ed8fab7fe8867ff3eb801c3697a426478387bb2f
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Dec 21 22:47:23 2013 +0200

    Made Extracts aware of full lookup path

commit 27a57b7aed91b2f346abc4a77da838bffa17c727
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 1 21:10:11 2013 +0200

    Removed debugger import

commit 074e0f5aca0572e368c11e6d2c73c9026e7d63d7
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 1 21:02:16 2013 +0200

    GIS lookup support added

commit 760e28e72bae475b442b026650969b0d182dbe53
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 1 20:04:31 2013 +0200

    Removed usage of Constraint, used Lookup instead

commit eac47766844b90e7d3269e7a8c012eee34ec0093
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 1 02:22:30 2013 +0200

    Minor cleanup of Lookup API

commit 2adf50428d59a783078b0da3d5d035106640c899
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sun Dec 1 02:14:19 2013 +0200

    Added documentation, polished implementation

commit 32c04357a87e3727a34f8c5e6ec0114d1fbbb303
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Nov 30 23:10:15 2013 +0200

    Avoid OrderedDict creation on lookup aggregate check

commit 7c8b3a32cc17b4dbca160921d48125f1631e0df4
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Sat Nov 30 23:04:34 2013 +0200

    Implemented nested lookups

    But there is no support of using lookups outside filtering yet.

commit 4d219d4cdef21d9c14e5d6b9299d583d1975fcba
Author: Anssi Kääriäinen <akaariai@gmail.com>
Date:   Wed Nov 27 22:07:30 2013 +0200

    Initial implementation of custom lookups

7 files changed, 203 insertions, 125 deletions

diff --git a/django/db/models/sql/aggregates.py b/django/db/models/sql/aggregates.py
index 8542a330c6..aef8b493bb 100644
--- a/django/db/models/sql/aggregates.py
+++ b/django/db/models/sql/aggregates.py
@@ -4,6 +4,7 @@ Classes to represent the default SQL aggregate functions
 import copy
 
 from django.db.models.fields import IntegerField, FloatField
+from django.db.models.lookups import RegisterLookupMixin
 
 
 __all__ = ['Aggregate', 'Avg', 'Count', 'Max', 'Min', 'StdDev', 'Sum', 'Variance']
@@ -14,7 +15,7 @@ ordinal_aggregate_field = IntegerField()
 computed_aggregate_field = FloatField()
 
 
-class Aggregate(object):
+class Aggregate(RegisterLookupMixin):
     """
     Default SQL Aggregate.
     """
@@ -93,6 +94,13 @@ class Aggregate(object):
 
         return self.sql_template % substitutions, params
 
+    def get_group_by_cols(self):
+        return []
+
+    @property
+    def output_type(self):
+        return self.field
+
 
 class Avg(Aggregate):
     is_computed = True
diff --git a/django/db/models/sql/compiler.py b/django/db/models/sql/compiler.py
index 41bba93206..123427cf8b 100644
--- a/django/db/models/sql/compiler.py
+++ b/django/db/models/sql/compiler.py
@@ -45,7 +45,7 @@ class SQLCompiler(object):
         if self.query.select_related and not self.query.related_select_cols:
             self.fill_related_selections()
 
-    def quote_name_unless_alias(self, name):
+    def __call__(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
@@ -61,6 +61,22 @@ class SQLCompiler(object):
         self.quote_cache[name] = r
         return r
 
+    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).
+        """
+        return self(name)
+
+    def compile(self, node):
+        vendor_impl = getattr(
+            node, 'as_' + self.connection.vendor, None)
+        if vendor_impl:
+            return vendor_impl(self, self.connection)
+        else:
+            return node.as_sql(self, self.connection)
+
     def as_sql(self, with_limits=True, with_col_aliases=False):
         """
         Creates the SQL for this query. Returns the SQL string and list of
@@ -88,11 +104,9 @@ class SQLCompiler(object):
         # docstring of get_from_clause() for details.
         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)
-        having_group_by = self.query.having.get_cols()
+        where, w_params = self.compile(self.query.where)
+        having, h_params = self.compile(self.query.having)
+        having_group_by = self.query.having.get_group_by_cols()
         params = []
         for val in six.itervalues(self.query.extra_select):
             params.extend(val[1])
@@ -180,7 +194,7 @@ class SQLCompiler(object):
         (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
+        qn = self
         qn2 = self.connection.ops.quote_name
         result = ['(%s) AS %s' % (col[0], qn2(alias)) for alias, col in six.iteritems(self.query.extra_select)]
         params = []
@@ -213,7 +227,7 @@ class SQLCompiler(object):
                         aliases.add(r)
                         col_aliases.add(col[1])
                 else:
-                    col_sql, col_params = col.as_sql(qn, self.connection)
+                    col_sql, col_params = self.compile(col)
                     result.append(col_sql)
                     params.extend(col_params)
 
@@ -229,7 +243,7 @@ class SQLCompiler(object):
 
         max_name_length = self.connection.ops.max_name_length()
         for alias, aggregate in self.query.aggregate_select.items():
-            agg_sql, agg_params = aggregate.as_sql(qn, self.connection)
+            agg_sql, agg_params = self.compile(aggregate)
             if alias is None:
                 result.append(agg_sql)
             else:
@@ -267,7 +281,7 @@ class SQLCompiler(object):
         result = []
         if opts is None:
             opts = self.query.get_meta()
-        qn = self.quote_name_unless_alias
+        qn = self
         qn2 = self.connection.ops.quote_name
         aliases = set()
         only_load = self.deferred_to_columns()
@@ -319,7 +333,7 @@ class SQLCompiler(object):
         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
+        qn = self
         qn2 = self.connection.ops.quote_name
         result = []
         opts = self.query.get_meta()
@@ -352,7 +366,7 @@ class SQLCompiler(object):
             ordering = (self.query.order_by
                         or self.query.get_meta().ordering
                         or [])
-        qn = self.quote_name_unless_alias
+        qn = self
         qn2 = self.connection.ops.quote_name
         distinct = self.query.distinct
         select_aliases = self._select_aliases
@@ -490,7 +504,7 @@ class SQLCompiler(object):
         ordering and distinct must be done first.
         """
         result = []
-        qn = self.quote_name_unless_alias
+        qn = self
         qn2 = self.connection.ops.quote_name
         first = True
         from_params = []
@@ -508,8 +522,7 @@ class SQLCompiler(object):
                 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, extra_params = self.compile(extra_cond)
                     extra_sql = 'AND (%s)' % extra_sql
                     from_params.extend(extra_params)
                 else:
@@ -541,7 +554,7 @@ class SQLCompiler(object):
         """
         Returns a tuple representing the SQL elements in the "group by" clause.
         """
-        qn = self.quote_name_unless_alias
+        qn = self
         result, params = [], []
         if self.query.group_by is not None:
             select_cols = self.query.select + self.query.related_select_cols
@@ -560,7 +573,7 @@ class SQLCompiler(object):
                 if isinstance(col, (list, tuple)):
                     sql = '%s.%s' % (qn(col[0]), qn(col[1]))
                 elif hasattr(col, 'as_sql'):
-                    sql, col_params = col.as_sql(qn, self.connection)
+                    self.compile(col)
                 else:
                     sql = '(%s)' % str(col)
                 if sql not in seen:
@@ -784,7 +797,7 @@ class SQLCompiler(object):
         return result
 
     def as_subquery_condition(self, alias, columns, qn):
-        inner_qn = self.quote_name_unless_alias
+        inner_qn = self
         qn2 = self.connection.ops.quote_name
         if len(columns) == 1:
             sql, params = self.as_sql()
@@ -895,9 +908,9 @@ class SQLDeleteCompiler(SQLCompiler):
         """
         assert len(self.query.tables) == 1, \
             "Can only delete from one table at a time."
-        qn = self.quote_name_unless_alias
+        qn = self
         result = ['DELETE FROM %s' % qn(self.query.tables[0])]
-        where, params = self.query.where.as_sql(qn=qn, connection=self.connection)
+        where, params = self.compile(self.query.where)
         if where:
             result.append('WHERE %s' % where)
         return ' '.join(result), tuple(params)
@@ -913,7 +926,7 @@ class SQLUpdateCompiler(SQLCompiler):
         if not self.query.values:
             return '', ()
         table = self.query.tables[0]
-        qn = self.quote_name_unless_alias
+        qn = self
         result = ['UPDATE %s' % qn(table)]
         result.append('SET')
         values, update_params = [], []
@@ -933,7 +946,7 @@ class SQLUpdateCompiler(SQLCompiler):
                 val = SQLEvaluator(val, self.query, allow_joins=False)
             name = field.column
             if hasattr(val, 'as_sql'):
-                sql, params = val.as_sql(qn, self.connection)
+                sql, params = self.compile(val)
                 values.append('%s = %s' % (qn(name), sql))
                 update_params.extend(params)
             elif val is not None:
@@ -944,7 +957,7 @@ class SQLUpdateCompiler(SQLCompiler):
         if not values:
             return '', ()
         result.append(', '.join(values))
-        where, params = self.query.where.as_sql(qn=qn, connection=self.connection)
+        where, params = self.compile(self.query.where)
         if where:
             result.append('WHERE %s' % where)
         return ' '.join(result), tuple(update_params + params)
@@ -1024,11 +1037,11 @@ class SQLAggregateCompiler(SQLCompiler):
         parameters.
         """
         if qn is None:
-            qn = self.quote_name_unless_alias
+            qn = self
 
         sql, params = [], []
         for aggregate in self.query.aggregate_select.values():
-            agg_sql, agg_params = aggregate.as_sql(qn, self.connection)
+            agg_sql, agg_params = self.compile(aggregate)
             sql.append(agg_sql)
             params.extend(agg_params)
         sql = ', '.join(sql)
diff --git a/django/db/models/sql/datastructures.py b/django/db/models/sql/datastructures.py
index f45ecaf76d..421c3cd860 100644
--- a/django/db/models/sql/datastructures.py
+++ b/django/db/models/sql/datastructures.py
@@ -5,18 +5,27 @@ the SQL domain.
 
 
 class Col(object):
-    def __init__(self, alias, col):
-        self.alias = alias
-        self.col = col
+    def __init__(self, alias, target, source):
+        self.alias, self.target, self.source = alias, target, source
 
     def as_sql(self, qn, connection):
-        return '%s.%s' % (qn(self.alias), self.col), []
+        return "%s.%s" % (qn(self.alias), qn(self.target.column)), []
 
-    def prepare(self):
-        return self
+    @property
+    def output_type(self):
+        return self.source
 
     def relabeled_clone(self, relabels):
-        return self.__class__(relabels.get(self.alias, self.alias), self.col)
+        return self.__class__(relabels.get(self.alias, self.alias), self.target, self.source)
+
+    def get_group_by_cols(self):
+        return [(self.alias, self.target.column)]
+
+    def get_lookup(self, name):
+        return self.output_type.get_lookup(name)
+
+    def prepare(self):
+        return self
 
 
 class EmptyResultSet(Exception):
diff --git a/django/db/models/sql/expressions.py b/django/db/models/sql/expressions.py
index 9f29e2ace5..e31eaa8a2f 100644
--- a/django/db/models/sql/expressions.py
+++ b/django/db/models/sql/expressions.py
@@ -24,11 +24,11 @@ class SQLEvaluator(object):
                                    (change_map.get(col[0], col[0]), col[1])))
         return clone
 
-    def get_cols(self):
+    def get_group_by_cols(self):
         cols = []
         for node, col in self.cols:
-            if hasattr(node, 'get_cols'):
-                cols.extend(node.get_cols())
+            if hasattr(node, 'get_group_by_cols'):
+                cols.extend(node.get_group_by_cols())
             elif isinstance(col, tuple):
                 cols.append(col)
         return cols
diff --git a/django/db/models/sql/query.py b/django/db/models/sql/query.py
index c3c8e55793..db4e6744bf 100644
--- a/django/db/models/sql/query.py
+++ b/django/db/models/sql/query.py
@@ -19,6 +19,7 @@ from django.db.models.constants import LOOKUP_SEP
 from django.db.models.aggregates import refs_aggregate
 from django.db.models.expressions import ExpressionNode
 from django.db.models.fields import FieldDoesNotExist
+from django.db.models.lookups import Transform
 from django.db.models.query_utils import Q
 from django.db.models.related import PathInfo
 from django.db.models.sql import aggregates as base_aggregates_module
@@ -1028,13 +1029,16 @@ class Query(object):
         # Add the aggregate to the query
         aggregate.add_to_query(self, alias, col=col, source=source, is_summary=is_summary)
 
-    def prepare_lookup_value(self, value, lookup_type, can_reuse):
+    def prepare_lookup_value(self, value, lookups, can_reuse):
+        # Default lookup if none given is exact.
+        if len(lookups) == 0:
+            lookups = ['exact']
         # Interpret '__exact=None' as the sql 'is NULL'; otherwise, reject all
         # uses of None as a query value.
         if value is None:
-            if lookup_type not in ('exact', 'iexact'):
+            if lookups[-1] not in ('exact', 'iexact'):
                 raise ValueError("Cannot use None as a query value")
-            lookup_type = 'isnull'
+            lookups[-1] = 'isnull'
             value = True
         elif callable(value):
             warnings.warn(
@@ -1055,40 +1059,54 @@ class Query(object):
         # stage. Using DEFAULT_DB_ALIAS isn't nice, but it is the best we
         # can do here. Similar thing is done in is_nullable(), too.
         if (connections[DEFAULT_DB_ALIAS].features.interprets_empty_strings_as_nulls and
-                lookup_type == 'exact' and value == ''):
+                lookups[-1] == 'exact' and value == ''):
             value = True
-            lookup_type = 'isnull'
-        return value, lookup_type
+            lookups[-1] = ['isnull']
+        return value, lookups
 
     def solve_lookup_type(self, lookup):
         """
         Solve the lookup type from the lookup (eg: 'foobar__id__icontains')
         """
-        lookup_type = 'exact'  # Default lookup type
-        lookup_parts = lookup.split(LOOKUP_SEP)
-        num_parts = len(lookup_parts)
-        if (len(lookup_parts) > 1 and lookup_parts[-1] in self.query_terms
-                and (not self._aggregates or lookup not in self._aggregates)):
-            # Traverse the lookup query to distinguish related fields from
-            # lookup types.
-            lookup_model = self.model
-            for counter, field_name in enumerate(lookup_parts):
-                try:
-                    lookup_field = lookup_model._meta.get_field(field_name)
-                except FieldDoesNotExist:
-                    # Not a field. Bail out.
-                    lookup_type = lookup_parts.pop()
-                    break
-                # Unless we're at the end of the list of lookups, let's attempt
-                # to continue traversing relations.
-                if (counter + 1) < num_parts:
-                    try:
-                        lookup_model = lookup_field.rel.to
-                    except AttributeError:
-                        # Not a related field. Bail out.
-                        lookup_type = lookup_parts.pop()
-                        break
-        return lookup_type, lookup_parts
+        lookup_splitted = lookup.split(LOOKUP_SEP)
+        if self._aggregates:
+            aggregate, aggregate_lookups = refs_aggregate(lookup_splitted, self.aggregates)
+            if aggregate:
+                return aggregate_lookups, (), aggregate
+        _, field, _, lookup_parts = self.names_to_path(lookup_splitted, self.get_meta())
+        field_parts = lookup_splitted[0:len(lookup_splitted) - len(lookup_parts)]
+        if len(lookup_parts) == 0:
+            lookup_parts = ['exact']
+        elif len(lookup_parts) > 1:
+            if not field_parts:
+                raise FieldError(
+                    'Invalid lookup "%s" for model %s".' %
+                    (lookup, self.get_meta().model.__name__))
+        return lookup_parts, field_parts, False
+
+    def build_lookup(self, lookups, lhs, rhs):
+        lookups = lookups[:]
+        while lookups:
+            lookup = lookups[0]
+            next = lhs.get_lookup(lookup)
+            if next:
+                if len(lookups) == 1:
+                    # This was the last lookup, so return value lookup.
+                    if issubclass(next, Transform):
+                        lookups.append('exact')
+                        lhs = next(lhs, lookups)
+                    else:
+                        return next(lhs, rhs)
+                else:
+                    lhs = next(lhs, lookups)
+            # A field's get_lookup() can return None to opt for backwards
+            # compatibility path.
+            elif len(lookups) > 2:
+                raise FieldError(
+                    "Unsupported lookup for field '%s'" % lhs.output_type.name)
+            else:
+                return None
+            lookups = lookups[1:]
 
     def build_filter(self, filter_expr, branch_negated=False, current_negated=False,
                      can_reuse=None, connector=AND):
@@ -1118,21 +1136,24 @@ class Query(object):
         is responsible for unreffing the joins used.
         """
         arg, value = filter_expr
-        lookup_type, parts = self.solve_lookup_type(arg)
-        if not parts:
+        if not arg:
             raise FieldError("Cannot parse keyword query %r" % arg)
+        lookups, parts, reffed_aggregate = self.solve_lookup_type(arg)
 
         # Work out the lookup type and remove it from the end of 'parts',
         # if necessary.
-        value, lookup_type = self.prepare_lookup_value(value, lookup_type, can_reuse)
+        value, lookups = self.prepare_lookup_value(value, lookups, can_reuse)
         used_joins = getattr(value, '_used_joins', [])
 
         clause = self.where_class()
-        if self._aggregates:
-            for alias, aggregate in self.aggregates.items():
-                if alias in (parts[0], LOOKUP_SEP.join(parts)):
-                    clause.add((aggregate, lookup_type, value), AND)
-                    return clause, []
+        if reffed_aggregate:
+            condition = self.build_lookup(lookups, reffed_aggregate, value)
+            if not condition:
+                # Backwards compat for custom lookups
+                assert len(lookups) == 1
+                condition = (reffed_aggregate, lookups[0], value)
+            clause.add(condition, AND)
+            return clause, []
 
         opts = self.get_meta()
         alias = self.get_initial_alias()
@@ -1154,11 +1175,31 @@ class Query(object):
         targets, alias, join_list = self.trim_joins(sources, join_list, path)
 
         if hasattr(field, 'get_lookup_constraint'):
-            constraint = field.get_lookup_constraint(self.where_class, alias, targets, sources,
-                                                     lookup_type, value)
+            # For now foreign keys get special treatment. This should be
+            # refactored when composite fields lands.
+            condition = field.get_lookup_constraint(self.where_class, alias, targets, sources,
+                                                    lookups, value)
+            lookup_type = lookups[-1]
         else:
-            constraint = (Constraint(alias, targets[0].column, field), lookup_type, value)
-        clause.add(constraint, AND)
+            assert(len(targets) == 1)
+            col = Col(alias, targets[0], field)
+            condition = self.build_lookup(lookups, col, value)
+            if not condition:
+                # Backwards compat for custom lookups
+                if lookups[0] not in self.query_terms:
+                    raise FieldError(
+                        "Join on field '%s' not permitted. Did you "
+                        "misspell '%s' for the lookup type?" %
+                        (col.output_type.name, lookups[0]))
+                if len(lookups) > 1:
+                    raise FieldError("Nested lookup '%s' not supported." %
+                                     LOOKUP_SEP.join(lookups))
+                condition = (Constraint(alias, targets[0].column, field), lookups[0], value)
+                lookup_type = lookups[-1]
+            else:
+                lookup_type = condition.lookup_name
+
+        clause.add(condition, AND)
 
         require_outer = lookup_type == 'isnull' and value is True and not current_negated
         if current_negated and (lookup_type != 'isnull' or value is False):
@@ -1175,7 +1216,8 @@ class Query(object):
                 # (col IS NULL OR col != someval)
                 #   <=>
                 # NOT (col IS NOT NULL AND col = someval).
-                clause.add((Constraint(alias, targets[0].column, None), 'isnull', False), AND)
+                lookup_class = targets[0].get_lookup('isnull')
+                clause.add(lookup_class(Col(alias, targets[0], sources[0]), False), AND)
         return clause, used_joins if not require_outer else ()
 
     def add_filter(self, filter_clause):
@@ -1189,7 +1231,7 @@ class Query(object):
         if not self._aggregates:
             return False
         if not isinstance(obj, Node):
-            return (refs_aggregate(obj[0].split(LOOKUP_SEP), self.aggregates)
+            return (refs_aggregate(obj[0].split(LOOKUP_SEP), self.aggregates)[0]
                     or (hasattr(obj[1], 'contains_aggregate')
                         and obj[1].contains_aggregate(self.aggregates)))
         return any(self.need_having(c) for c in obj.children)
@@ -1277,7 +1319,7 @@ class Query(object):
         needed_inner = joinpromoter.update_join_types(self)
         return target_clause, needed_inner
 
-    def names_to_path(self, names, opts, allow_many):
+    def names_to_path(self, names, opts, allow_many=True, fail_on_missing=False):
         """
         Walks the names path and turns them PathInfo tuples. Note that a
         single name in 'names' can generate multiple PathInfos (m2m for
@@ -1297,9 +1339,10 @@ class Query(object):
             try:
                 field, model, direct, m2m = opts.get_field_by_name(name)
             except FieldDoesNotExist:
-                available = opts.get_all_field_names() + list(self.aggregate_select)
-                raise FieldError("Cannot resolve keyword %r into field. "
-                                 "Choices are: %s" % (name, ", ".join(available)))
+                # We didn't found the current field, so move position back
+                # one step.
+                pos -= 1
+                break
             # Check if we need any joins for concrete inheritance cases (the
             # field lives in parent, but we are currently in one of its
             # children)
@@ -1334,15 +1377,14 @@ class Query(object):
                 final_field = field
                 targets = (field,)
                 break
+        if pos == -1 or (fail_on_missing and pos + 1 != len(names)):
+            self.raise_field_error(opts, name)
+        return path, final_field, targets, names[pos + 1:]
 
-        if pos != len(names) - 1:
-            if pos == len(names) - 2:
-                raise FieldError(
-                    "Join on field %r not permitted. Did you misspell %r for "
-                    "the lookup type?" % (name, names[pos + 1]))
-            else:
-                raise FieldError("Join on field %r not permitted." % name)
-        return path, final_field, targets
+    def raise_field_error(self, opts, name):
+        available = opts.get_all_field_names() + list(self.aggregate_select)
+        raise FieldError("Cannot resolve keyword %r into field. "
+                         "Choices are: %s" % (name, ", ".join(available)))
 
     def setup_joins(self, names, opts, alias, can_reuse=None, allow_many=True):
         """
@@ -1371,8 +1413,9 @@ class Query(object):
         """
         joins = [alias]
         # First, generate the path for the names
-        path, final_field, targets = self.names_to_path(
-            names, opts, allow_many)
+        path, final_field, targets, rest = self.names_to_path(
+            names, opts, allow_many, fail_on_missing=True)
+
         # Then, add the path to the query's joins. Note that we can't trim
         # joins at this stage - we will need the information about join type
         # of the trimmed joins.
@@ -1387,8 +1430,6 @@ class Query(object):
             alias = self.join(
                 connection, reuse=reuse, nullable=nullable, join_field=join.join_field)
             joins.append(alias)
-        if hasattr(final_field, 'field'):
-            final_field = final_field.field
         return final_field, targets, opts, joins, path
 
     def trim_joins(self, targets, joins, path):
@@ -1451,17 +1492,19 @@ class Query(object):
         # nothing
         alias, col = query.select[0].col
         if self.is_nullable(query.select[0].field):
-            query.where.add((Constraint(alias, col, query.select[0].field), 'isnull', False), AND)
+            lookup_class = query.select[0].field.get_lookup('isnull')
+            lookup = lookup_class(Col(alias, query.select[0].field, query.select[0].field), False)
+            query.where.add(lookup, AND)
         if alias in can_reuse:
-            pk = query.select[0].field.model._meta.pk
+            select_field = query.select[0].field
+            pk = select_field.model._meta.pk
             # Need to add a restriction so that outer query's filters are in effect for
             # the subquery, too.
             query.bump_prefix(self)
-            query.where.add(
-                (Constraint(query.select[0].col[0], pk.column, pk),
-                 'exact', Col(alias, pk.column)),
-                AND
-            )
+            lookup_class = select_field.get_lookup('exact')
+            lookup = lookup_class(Col(query.select[0].col[0], pk, pk),
+                                  Col(alias, pk, pk))
+            query.where.add(lookup, AND)
 
         condition, needed_inner = self.build_filter(
             ('%s__in' % trimmed_prefix, query),
diff --git a/django/db/models/sql/subqueries.py b/django/db/models/sql/subqueries.py
index e9e292e787..86b1efd3f8 100644
--- a/django/db/models/sql/subqueries.py
+++ b/django/db/models/sql/subqueries.py
@@ -5,12 +5,12 @@ Query subclasses which provide extra functionality beyond simple data retrieval.
 from django.conf import settings
 from django.core.exceptions import FieldError
 from django.db import connections
+from django.db.models.query_utils import Q
 from django.db.models.constants import LOOKUP_SEP
 from django.db.models.fields import DateField, DateTimeField, FieldDoesNotExist
 from django.db.models.sql.constants import GET_ITERATOR_CHUNK_SIZE, SelectInfo
 from django.db.models.sql.datastructures import Date, DateTime
 from django.db.models.sql.query import Query
-from django.db.models.sql.where import AND, Constraint
 from django.utils import six
 from django.utils import timezone
 
@@ -42,10 +42,10 @@ class DeleteQuery(Query):
         if not field:
             field = self.get_meta().pk
         for offset in range(0, len(pk_list), GET_ITERATOR_CHUNK_SIZE):
-            where = self.where_class()
-            where.add((Constraint(None, field.column, field), 'in',
-                       pk_list[offset:offset + GET_ITERATOR_CHUNK_SIZE]), AND)
-            self.do_query(self.get_meta().db_table, where, using=using)
+            self.where = self.where_class()
+            self.add_q(Q(
+                **{field.attname + '__in': pk_list[offset:offset + GET_ITERATOR_CHUNK_SIZE]}))
+            self.do_query(self.get_meta().db_table, self.where, using=using)
 
     def delete_qs(self, query, using):
         """
@@ -80,9 +80,8 @@ class DeleteQuery(Query):
                     SelectInfo((self.get_initial_alias(), pk.column), None)
                 ]
                 values = innerq
-            where = self.where_class()
-            where.add((Constraint(None, pk.column, pk), 'in', values), AND)
-            self.where = where
+            self.where = self.where_class()
+            self.add_q(Q(pk__in=values))
         self.get_compiler(using).execute_sql(None)
 
 
@@ -113,13 +112,10 @@ class UpdateQuery(Query):
                 related_updates=self.related_updates.copy(), **kwargs)
 
     def update_batch(self, pk_list, values, using):
-        pk_field = self.get_meta().pk
         self.add_update_values(values)
         for offset in range(0, len(pk_list), GET_ITERATOR_CHUNK_SIZE):
             self.where = self.where_class()
-            self.where.add((Constraint(None, pk_field.column, pk_field), 'in',
-                            pk_list[offset:offset + GET_ITERATOR_CHUNK_SIZE]),
-                           AND)
+            self.add_q(Q(pk__in=pk_list[offset: offset + GET_ITERATOR_CHUNK_SIZE]))
             self.get_compiler(using).execute_sql(None)
 
     def add_update_values(self, values):
diff --git a/django/db/models/sql/where.py b/django/db/models/sql/where.py
index 44a4ce9d1d..be0c559c1b 100644
--- a/django/db/models/sql/where.py
+++ b/django/db/models/sql/where.py
@@ -5,6 +5,7 @@ Code to manage the creation and SQL rendering of 'where' constraints.
 import collections
 import datetime
 from itertools import repeat
+import warnings
 
 from django.conf import settings
 from django.db.models.fields import DateTimeField, Field
@@ -101,7 +102,7 @@ class WhereNode(tree.Node):
         for child in self.children:
             try:
                 if hasattr(child, 'as_sql'):
-                    sql, params = child.as_sql(qn=qn, connection=connection)
+                    sql, params = qn.compile(child)
                 else:
                     # A leaf node in the tree.
                     sql, params = self.make_atom(child, qn, connection)
@@ -152,16 +153,16 @@ class WhereNode(tree.Node):
                 sql_string = '(%s)' % sql_string
         return sql_string, result_params
 
-    def get_cols(self):
+    def get_group_by_cols(self):
         cols = []
         for child in self.children:
-            if hasattr(child, 'get_cols'):
-                cols.extend(child.get_cols())
+            if hasattr(child, 'get_group_by_cols'):
+                cols.extend(child.get_group_by_cols())
             else:
                 if isinstance(child[0], Constraint):
                     cols.append((child[0].alias, child[0].col))
-                if hasattr(child[3], 'get_cols'):
-                    cols.extend(child[3].get_cols())
+                if hasattr(child[3], 'get_group_by_cols'):
+                    cols.extend(child[3].get_group_by_cols())
         return cols
 
     def make_atom(self, child, qn, connection):
@@ -174,6 +175,9 @@ class WhereNode(tree.Node):
         Returns the string for the SQL fragment and the parameters to use for
         it.
         """
+        warnings.warn(
+            "The make_atom() method will be removed in Django 1.9. Use Lookup class instead.",
+            PendingDeprecationWarning)
         lvalue, lookup_type, value_annotation, params_or_value = child
         field_internal_type = lvalue.field.get_internal_type() if lvalue.field else None
 
@@ -193,13 +197,13 @@ class WhereNode(tree.Node):
             field_sql, field_params = self.sql_for_columns(lvalue, qn, connection, field_internal_type), []
         else:
             # A smart object with an as_sql() method.
-            field_sql, field_params = lvalue.as_sql(qn, connection)
+            field_sql, field_params = qn.compile(lvalue)
 
         is_datetime_field = value_annotation is datetime.datetime
         cast_sql = connection.ops.datetime_cast_sql() if is_datetime_field else '%s'
 
         if hasattr(params, 'as_sql'):
-            extra, params = params.as_sql(qn, connection)
+            extra, params = qn.compile(params)
             cast_sql = ''
         else:
             extra = ''
@@ -282,6 +286,8 @@ class WhereNode(tree.Node):
             if hasattr(child, 'relabel_aliases'):
                 # For example another WhereNode
                 child.relabel_aliases(change_map)
+            elif hasattr(child, 'relabeled_clone'):
+                self.children[pos] = child.relabeled_clone(change_map)
             elif isinstance(child, (list, tuple)):
                 # tuple starting with Constraint
                 child = (child[0].relabeled_clone(change_map),) + child[1:]
@@ -347,10 +353,13 @@ class Constraint(object):
     pre-process itself prior to including in the WhereNode.
     """
     def __init__(self, alias, col, field):
+        warnings.warn(
+            "The Constraint class will be removed in Django 1.9. Use Lookup class instead.",
+            PendingDeprecationWarning)
         self.alias, self.col, self.field = alias, col, field
 
     def prepare(self, lookup_type, value):
-        if self.field:
+        if self.field and not hasattr(value, 'as_sql'):
             return self.field.get_prep_lookup(lookup_type, value)
         return value