from philo.signals import entity_class_prepared
from philo.validators import json_validator
from UserDict import DictMixin
+from mptt.models import MPTTModel, MPTTModelBase, MPTTOptions
class Tag(models.Model):
# Value is probably a queryset - but allow any iterable.
# These lines shouldn't be necessary; however, if value is an EmptyQuerySet,
- # the code won't work without them. Unclear why...
+ # the code (specifically the object_id__in query) won't work without them. Unclear why...
if not value:
value = []
+ # Before we can fiddle with the many-to-many to foreignkeyvalues, we need
+ # a pk.
+ if self.pk is None:
+ self.save()
+
if isinstance(value, models.query.QuerySet):
value = value.values_list('id', flat=True)
@property
def attributes(self):
- return QuerySetMapper(self.attribute_set)
+ return QuerySetMapper(self.attribute_set.all())
@property
def _added_attribute_registry(self):
class TreeManager(models.Manager):
use_for_related_fields = True
- def roots(self):
- return self.filter(parent__isnull=True)
-
- def get_with_path(self, path, root=None, absolute_result=True, pathsep='/'):
+ def get_with_path(self, path, root=None, absolute_result=True, pathsep='/', field='slug'):
"""
- Returns the object with the path, or None if there is no object with that path,
- unless absolute_result is set to False, in which case it returns a tuple containing
- the deepest object found along the path, and the remainder of the path after that
- object as a string (or None in the case that there is no remaining path).
+ Returns the object with the path, unless absolute_result is set to False, in which
+ case it returns a tuple containing the deepest object found along the path, and the
+ remainder of the path after that object as a string (or None if there is no remaining
+ path). Raises a DoesNotExist exception if no object is found with the given path.
+
+ If the path you're searching for is known to exist, it is always faster to use
+ absolute_result=True - unless the path depth is over ~40, in which case the high cost
+ of the absolute query makes a binary search (i.e. non-absolute) faster.
"""
- slugs = path.split(pathsep)
- obj = root
- remaining_slugs = list(slugs)
- remainder = None
- for slug in slugs:
- remaining_slugs.remove(slug)
- if slug: # ignore blank slugs, handles for multiple consecutive pathseps
+ # Note: SQLite allows max of 64 tables in one join. That means the binary search will
+ # only work on paths with a max depth of 127 and the absolute fetch will only work
+ # to a max depth of (surprise!) 63. Although this could be handled, chances are your
+ # tree structure won't be that deep.
+ segments = path.split(pathsep)
+
+ # Check for a trailing pathsep so we can restore it later.
+ trailing_pathsep = False
+ if segments[-1] == '':
+ trailing_pathsep = True
+
+ # Clean out blank segments. Handles multiple consecutive pathseps.
+ while True:
+ try:
+ segments.remove('')
+ except ValueError:
+ break
+
+ # Special-case a lack of segments. No queries necessary.
+ if not segments:
+ if root is not None:
+ if absolute_result:
+ return root
+ return root, None
+ else:
+ raise self.model.DoesNotExist('%s matching query does not exist.' % self.model._meta.object_name)
+
+ def make_query_kwargs(segments, root):
+ kwargs = {}
+ prefix = ""
+ revsegs = list(segments)
+ revsegs.reverse()
+
+ for segment in revsegs:
+ kwargs["%s%s__exact" % (prefix, field)] = segment
+ prefix += "parent__"
+
+ if prefix:
+ kwargs[prefix[:-2]] = root
+
+ return kwargs
+
+ def build_path(segments):
+ path = pathsep.join(segments)
+ if trailing_pathsep and segments and segments[-1] != '':
+ path += pathsep
+ return path
+
+ def find_obj(segments, depth, deepest_found=None):
+ if deepest_found is None:
+ deepest_level = 0
+ elif root is None:
+ deepest_level = deepest_found.get_level() + 1
+ else:
+ deepest_level = deepest_found.get_level() - root.get_level()
+ try:
+ obj = self.get(**make_query_kwargs(segments[deepest_level:depth], deepest_found or root))
+ except self.model.DoesNotExist:
+ if not deepest_level and depth > 1:
+ # make sure there's a root node...
+ depth = 1
+ else:
+ # Try finding one with half the path since the deepest find.
+ depth = (deepest_level + depth)/2
+
+ if deepest_level == depth:
+ # This should happen if nothing is found with any part of the given path.
+ raise
+
+ return find_obj(segments, depth, deepest_found)
+ else:
+ # Yay! Found one!
+ if root is None:
+ deepest_level = obj.get_level() + 1
+ else:
+ deepest_level = obj.get_level() - root.get_level()
+
+ # Could there be a deeper one?
+ if obj.is_leaf_node():
+ return obj, build_path(segments[deepest_level:]) or None
+
+ depth += (len(segments) - depth)/2 or len(segments) - depth
+
+ if depth > deepest_level + obj.get_descendant_count():
+ depth = deepest_level + obj.get_descendant_count()
+
+ if deepest_level == depth:
+ return obj, build_path(segments[deepest_level:]) or None
+
try:
- obj = self.get(slug__exact=slug, parent__exact=obj)
+ return find_obj(segments, depth, obj)
except self.model.DoesNotExist:
- if absolute_result:
- obj = None
- remaining_slugs.insert(0, slug)
- remainder = pathsep.join(remaining_slugs)
- break
- if obj:
- if absolute_result:
- return obj
- else:
- return (obj, remainder)
- raise self.model.DoesNotExist('%s matching query does not exist.' % self.model._meta.object_name)
+ # Then this was the deepest.
+ return obj, build_path(segments[deepest_level:])
+
+ if absolute_result:
+ return self.get(**make_query_kwargs(segments, root))
+
+ # Try a modified binary search algorithm. Feed the root in so that query complexity
+ # can be reduced. It might be possible to weight the search towards the beginning
+ # of the path, since short paths are more likely, but how far forward? It would
+ # need to shift depending on len(segments) - perhaps logarithmically?
+ return find_obj(segments, len(segments)/2 or len(segments))
-class TreeModel(models.Model):
+class TreeModel(MPTTModel):
objects = TreeManager()
parent = models.ForeignKey('self', related_name='children', null=True, blank=True)
slug = models.SlugField(max_length=255)
- def has_ancestor(self, ancestor):
- parent = self
- while parent:
- if parent == ancestor:
- return True
- parent = parent.parent
- return False
-
def get_path(self, root=None, pathsep='/', field='slug'):
- if root is not None and not self.has_ancestor(root):
+ if root == self:
+ return ''
+
+ if root is not None and not self.is_descendant_of(root):
raise AncestorDoesNotExist(root)
- path = getattr(self, field, '?')
- parent = self.parent
- while parent and parent != root:
- path = getattr(parent, field, '?') + pathsep + path
- parent = parent.parent
- return path
+ qs = self.get_ancestors()
+
+ if root is not None:
+ qs = qs.filter(level__gt=root.level)
+
+ return pathsep.join([getattr(parent, field, '?') for parent in list(qs) + [self]])
path = property(get_path)
def __unicode__(self):
abstract = True
+class TreeEntityBase(MPTTModelBase, EntityBase):
+ def __new__(meta, name, bases, attrs):
+ attrs['_mptt_meta'] = MPTTOptions(attrs.pop('MPTTMeta', None))
+ cls = EntityBase.__new__(meta, name, bases, attrs)
+
+ return meta.register(cls)
+
+
class TreeEntity(Entity, TreeModel):
+ __metaclass__ = TreeEntityBase
+
@property
def attributes(self):
if self.parent:
- return QuerySetMapper(self.attribute_set, passthrough=self.parent.attributes)
+ return QuerySetMapper(self.attribute_set.all(), passthrough=self.parent.attributes)
return super(TreeEntity, self).attributes
class Meta:
git.ithinksw.org / philo.git / blobdiff