representing hierarchical data with django-rest-framework
Sometimes you want to provide features that require the manipulation and/or representation of hierarchical data, maybe it's a comment system or a product categorization feature, and if you don't have many constraints a design called adjacency list is often used. It's so simple and natural that you probably have done something similar a few times, even not knowing the concept.
class Comment(models.Model): comment = models.CharField(max_length=250) parent = models.ForeignKey('self', related_name='reply_set', null=True) date = models.DateField(auto_now=True)
In a simple adjacency list model, each record contains a reference to it predecessor, when this attribute is set to null, the record become a root node. The remaining records can also be root nodes or children's.
With the simple model above, one can display comments in a tree format and basic operations will be simple and cheap:
#knowing who is the father is just a matter of consulting the model attribute: comment = Comments.objects.get(pk=1) comment.parent #accessing replies (children's) comment.reply_set.all()
But in order to explore the entire lineage of a conversation and provide the many features of a tree data structure, you may have to define some iteration/recursive methods that in practice can spawn many SQL consults and get uncomfortably slow for large amount of data.
caption: django-rest-framework browsable api
The tree representation seen above is done using a nested serializer in django-rest-framework. Unfortunately it falls in the expensiveness trap, it walk thought the data recursively doing multiple sql consults in order to build the tree. On the bright side, it can be easily achieved with the following code:
class RecursiveSerializer(serializers.Serializer): def to_representation(self, value): serializer = self.parent.parent.__class__(value, context=self.context) return serializer.data class CommentSerializer(serializers.ModelSerializer): reply_set = RecursiveSerializer(many=True, read_only=True) class Meta: model = Comment fields = ('id', 'date', 'comment', 'parent', 'reply_set')
I suggest that the data representation have an extra route to only display trees for root nodes. You should not direct
filter in the ViewSet queryset attribute or you will lose the ability to update/delete children's using this endpoint.
If you don't like this solution, you can instead implement a url-based filter with django-filter or even override the
class CommentViewSet(viewsets.ModelViewSet): # queryset = Comment.objects.filter(parent=None) # Don't queryset = Comment.objects.all() serializer_class = CommentSerializer @list_route() def roots(self, request): queryset = Comment.objects.filter(parent=None) serializer = self.get_serializer(queryset, many=True) return Response(serializer.data)
One important thing to account for is the delete behavior, by default foreign keys uses CASCADE,
so if you remove a higher level comment, all the replies will be ripped off too. To avoid this, you can change the parent
field to uses
on_delete=models.PROTECT or any of the other options. In the case of a comment system you can replicate
the reddit approach, change the comment content to something like "Comment removed", remove author information
and keep the record in place to preserve the tree:
caption: How reddit handle deleted comments
by Ricardo Pascal on March 9, 2016