Reduced number of queries for BlogView.get_entries_by_tag. Corrected waldo's AccountM...
[philo.git] / models / base.py
index 5ca782a..8140520 100644 (file)
@@ -2,14 +2,17 @@ from django import forms
 from django.db import models
 from django.contrib.contenttypes.models import ContentType
 from django.contrib.contenttypes import generic
-from django.utils import simplejson as json
 from django.core.exceptions import ObjectDoesNotExist
+from django.core.validators import RegexValidator
+from django.utils import simplejson as json
+from django.utils.encoding import smart_str
 from philo.exceptions import AncestorDoesNotExist
 from philo.models.fields import JSONField
 from philo.utils import ContentTypeRegistryLimiter, ContentTypeSubclassLimiter
 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):
@@ -21,6 +24,7 @@ class Tag(models.Model):
        
        class Meta:
                app_label = 'philo'
+               ordering = ('name',)
 
 
 class Titled(models.Model):
@@ -46,6 +50,12 @@ def unregister_value_model(model):
 
 
 class AttributeValue(models.Model):
+       attribute_set = generic.GenericRelation('Attribute', content_type_field='value_content_type', object_id_field='value_object_id')
+       
+       @property
+       def attribute(self):
+               return self.attribute_set.all()[0]
+       
        def apply_data(self, data):
                raise NotImplementedError
        
@@ -66,7 +76,7 @@ class JSONValue(AttributeValue):
        value = JSONField() #verbose_name='Value (JSON)', help_text='This value must be valid JSON.')
        
        def __unicode__(self):
-               return self.value_json
+               return smart_str(self.value)
        
        def value_formfield(self, **kwargs):
                kwargs['initial'] = self.value_json
@@ -80,7 +90,7 @@ class JSONValue(AttributeValue):
 
 
 class ForeignKeyValue(AttributeValue):
-       content_type = models.ForeignKey(ContentType, related_name='foreign_key_value_set', limit_choices_to=value_content_type_limiter, verbose_name='Value type', null=True, blank=True)
+       content_type = models.ForeignKey(ContentType, limit_choices_to=value_content_type_limiter, verbose_name='Value type', null=True, blank=True)
        object_id = models.PositiveIntegerField(verbose_name='Value ID', null=True, blank=True)
        value = generic.GenericForeignKey()
        
@@ -103,14 +113,14 @@ class ForeignKeyValue(AttributeValue):
 
 
 class ManyToManyValue(AttributeValue):
-       content_type = models.ForeignKey(ContentType, related_name='many_to_many_value_set', limit_choices_to=value_content_type_limiter, verbose_name='Value type', null=True, blank=True)
-       object_ids = models.CommaSeparatedIntegerField(max_length=300, verbose_name='Value IDs', null=True, blank=True)
+       content_type = models.ForeignKey(ContentType, limit_choices_to=value_content_type_limiter, verbose_name='Value type', null=True, blank=True)
+       values = models.ManyToManyField(ForeignKeyValue, blank=True, null=True)
        
        def get_object_id_list(self):
-               if not self.object_ids:
+               if not self.values.count():
                        return []
                else:
-                       return self.object_ids.split(',')
+                       return self.values.values_list('object_id', flat=True)
        
        def get_value(self):
                if self.content_type is None:
@@ -119,13 +129,28 @@ class ManyToManyValue(AttributeValue):
                return self.content_type.model_class()._default_manager.filter(id__in=self.get_object_id_list())
        
        def set_value(self, value):
-               if value is None:
-                       self.object_ids = ""
-                       return
-               if not isinstance(value, models.query.QuerySet):
-                       raise TypeError("Value must be a QuerySet.")
-               self.content_type = ContentType.objects.get_for_model(value.model)
-               self.object_ids = ','.join([`value` for value in value.values_list('id', flat=True)])
+               # Value is probably a queryset - but allow any iterable.
+               
+               # These lines shouldn't be necessary; however, if value is an EmptyQuerySet,
+               # 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)
+               
+               self.values.filter(~models.Q(object_id__in=value)).delete()
+               current = self.get_object_id_list()
+               
+               for v in value:
+                       if v in current:
+                               continue
+                       self.values.create(content_type=self.content_type, object_id=v)
        
        value = property(get_value, set_value)
        
@@ -136,7 +161,12 @@ class ManyToManyValue(AttributeValue):
                return form_class(self.content_type.model_class()._default_manager.all(), **kwargs)
        
        def apply_data(self, cleaned_data):
-               self.value = cleaned_data.get('value', None)
+               if 'value' in cleaned_data and cleaned_data['value'] is not None:
+                       self.value = cleaned_data['value']
+               else:
+                       self.content_type = cleaned_data.get('content_type', None)
+                       # If there is no value set in the cleaned data, clear the stored value.
+                       self.value = []
        
        class Meta:
                app_label = 'philo'
@@ -151,30 +181,7 @@ class Attribute(models.Model):
        value_object_id = models.PositiveIntegerField(verbose_name='Value ID', null=True, blank=True)
        value = generic.GenericForeignKey('value_content_type', 'value_object_id')
        
-       key = models.CharField(max_length=255)
-       
-       def get_value_class(self, value):
-               if isinstance(value, models.query.QuerySet):
-                       return ManyToManyValue
-               elif isinstance(value, models.Model) or (value is None and self.value_content_type.model_class() is ForeignKeyValue):
-                       return ForeignKeyValue
-               else:
-                       return JSONValue
-       
-       def set_value(self, value):
-               # is this useful? The best way of doing it?
-               value_class = self.get_value_class(value)
-               
-               if self.value is None or value_class != self.value_content_type.model_class():
-                       if self.value is not None:
-                               self.value.delete()
-                       new_value = value_class()
-                       new_value.value = value
-                       new_value.save()
-                       self.value = new_value
-               else:
-                       self.value.value = value
-                       self.value.save()
+       key = models.CharField(max_length=255, validators=[RegexValidator("\w+")], help_text="Must contain one or more alphanumeric characters or underscores.")
        
        def __unicode__(self):
                return u'"%s": %s' % (self.key, self.value)
@@ -191,11 +198,15 @@ class QuerySetMapper(object, DictMixin):
        
        def __getitem__(self, key):
                try:
-                       return self.queryset.get(key__exact=key).value
+                       value = self.queryset.get(key__exact=key).value
                except ObjectDoesNotExist:
                        if self.passthrough is not None:
                                return self.passthrough.__getitem__(key)
                        raise KeyError
+               else:
+                       if value is not None:
+                               return value.value
+                       return value
        
        def keys(self):
                keys = set(self.queryset.values_list('key', flat=True).distinct())
@@ -232,7 +243,7 @@ class Entity(models.Model):
        
        @property
        def attributes(self):
-               return QuerySetMapper(self.attribute_set)
+               return QuerySetMapper(self.attribute_set.all())
        
        @property
        def _added_attribute_registry(self):
@@ -253,14 +264,15 @@ class Entity(models.Model):
                        self.attribute_set.filter(key__exact=key).delete()
                del self._removed_attribute_registry[:]
                
-               for key, value in self._added_attribute_registry.items():
+               for field, value in self._added_attribute_registry.items():
                        try:
-                               attribute = self.attribute_set.get(key__exact=key)
+                               attribute = self.attribute_set.get(key__exact=field.key)
                        except Attribute.DoesNotExist:
                                attribute = Attribute()
                                attribute.entity = self
-                               attribute.key = key
-                       attribute.set_value(value)
+                               attribute.key = field.key
+                       
+                       field.set_attribute_value(attribute, value)
                        attribute.save()
                self._added_attribute_registry.clear()
        
@@ -271,69 +283,142 @@ class Entity(models.Model):
 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.
+                                       if root is not None and deepest_found is None:
+                                               return root, build_path(segments)
+                                       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 == self:
+                       return ''
+               
+               if root is not None and not self.is_descendant_of(root):
+                       raise AncestorDoesNotExist(root)
+               
+               qs = self.get_ancestors()
+               
                if root is not None:
-                       if not self.has_ancestor(root):
-                               raise AncestorDoesNotExist(root)
-                       path = ''
-                       parent = self
-                       while parent and parent != root:
-                               path = getattr(parent, field, '?') + pathsep + path
-                               parent = parent.parent
-                       return path
-               else:
-                       path = getattr(self, field, '?')
-                       parent = self.parent
-                       while parent and parent != root:
-                               path = getattr(parent, field, '?') + pathsep + path
-                               parent = parent.parent
-                       return path
+                       qs = qs.filter(**{'%s__gt' % self._mptt_meta.level_attr: root.get_level()})
+               
+               return pathsep.join([getattr(parent, field, '?') for parent in list(qs) + [self]])
        path = property(get_path)
        
        def __unicode__(self):
@@ -344,18 +429,22 @@ class TreeModel(models.Model):
                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
        
-       @property
-       def relationships(self):
-               if self.parent:
-                       return QuerySetMapper(self.relationship_set, passthrough=self.parent.relationships)
-               return super(TreeEntity, self).relationships
-       
        class Meta:
                abstract = True
\ No newline at end of file