Fixed a problem when constructing complex select_related() calls.

Avoids joining with the wrong tables when connecting select_related() tables to
the main query. This also leads to slightly more efficient (meaning less tables
are joined) SQL queries in some other cases, too. Some unnecessary tables are
now trimmed that were not previously.


git-svn-id: http://code.djangoproject.com/svn/django/trunk@7741 bcc190cf-cafb-0310-a4f2-bffc1f526a37

2 files changed, 57 insertions, 13 deletions

diff --git a/django/db/models/sql/query.py b/django/db/models/sql/query.py
index e6a3cd2679..e3e3585a8e 100644
--- a/django/db/models/sql/query.py
+++ b/django/db/models/sql/query.py
@@ -631,8 +631,10 @@ class Query(object):
             # We have to do the same "final join" optimisation as in
             # add_filter, since the final column might not otherwise be part of
             # the select set (so we can't order on it).
-            join = self.alias_map[alias]
-            if col == join[RHS_JOIN_COL]:
+            while 1:
+                join = self.alias_map[alias]
+                if col != join[RHS_JOIN_COL]:
+                    break
                 self.unref_alias(alias)
                 alias = join[LHS_ALIAS]
                 col = join[LHS_JOIN_COL]
@@ -830,6 +832,10 @@ class Query(object):
         if not always_create:
             for alias in self.join_map.get(t_ident, ()):
                 if alias not in exclusions:
+                    if lhs_table and not self.alias_refcount[self.alias_map[alias][LHS_ALIAS]]:
+                        # The LHS of this join tuple is no longer part of the
+                        # query, so skip this possibility.
+                        continue
                     self.ref_alias(alias)
                     if promote:
                         self.promote_alias(alias)
@@ -989,20 +995,22 @@ class Query(object):
             col = target.column
         alias = join_list[-1]
 
-        if final > 1:
+        while final > 1:
             # An optimization: if the final join is against the same column as
             # we are comparing against, we can go back one step in the join
-            # chain and compare against the lhs of the join instead. The result
-            # (potentially) involves one less table join.
+            # chain and compare against the lhs of the join instead (and then
+            # repeat the optimization). The result, potentially, involves less
+            # table joins.
             join = self.alias_map[alias]
-            if col == join[RHS_JOIN_COL]:
-                self.unref_alias(alias)
-                alias = join[LHS_ALIAS]
-                col = join[LHS_JOIN_COL]
-                join_list = join_list[:-1]
-                final -= 1
-                if final == penultimate:
-                    penultimate = last.pop()
+            if col != join[RHS_JOIN_COL]:
+                break
+            self.unref_alias(alias)
+            alias = join[LHS_ALIAS]
+            col = join[LHS_JOIN_COL]
+            join_list = join_list[:-1]
+            final -= 1
+            if final == penultimate:
+                penultimate = last.pop()
 
         if (lookup_type == 'isnull' and value is True and not negate and
                 final > 1):
diff --git a/tests/regressiontests/queries/models.py b/tests/regressiontests/queries/models.py
index 5af48741d6..2e84d89ea8 100644
--- a/tests/regressiontests/queries/models.py
+++ b/tests/regressiontests/queries/models.py
@@ -135,6 +135,24 @@ class ManagedModel(models.Model):
     def __unicode__(self):
         return self.data
 
+# An inter-related setup with multiple paths from Child to Detail.
+class Detail(models.Model):
+    data = models.CharField(max_length=10)
+
+class MemberManager(models.Manager):
+    def get_query_set(self):
+        return super(MemberManager, self).get_query_set().select_related("details")
+
+class Member(models.Model):
+    name = models.CharField(max_length=10)
+    details = models.OneToOneField(Detail, primary_key=True)
+
+    objects = MemberManager()
+
+class Child(models.Model):
+    person = models.OneToOneField(Member, primary_key=True)
+    parent = models.ForeignKey(Member, related_name="children")
+
 
 __test__ = {'API_TESTS':"""
 >>> t1 = Tag(name='t1')
@@ -720,5 +738,23 @@ appropriately.
 >>> Report.objects.values_list("creator__extra__info", flat=True).order_by("name")
 [u'e1', u'e2', None]
 
+Similarly for select_related(), joins beyond an initial nullable join must
+use outer joins so that all results are included.
+>>> Report.objects.select_related("creator", "creator__extra").order_by("name")
+[<Report: r1>, <Report: r2>, <Report: r3>]
+
+When there are multiple paths to a table from another table, we have to be
+careful not to accidentally reuse an inappropriate join when using
+select_related(). We used to return the parent's Detail record here by mistake.
+
+>>> d1 = Detail.objects.create(data="d1")
+>>> d2 = Detail.objects.create(data="d2")
+>>> m1 = Member.objects.create(name="m1", details=d1)
+>>> m2 = Member.objects.create(name="m2", details=d2)
+>>> c1 = Child.objects.create(person=m2, parent=m1)
+>>> obj = m1.children.select_related("person__details")[0]
+>>> obj.person.details.data
+u'd2'
+
 """}