[Fixed]-Efficient function to retrieve a queryset of ancestors of an mptt queryset


How about:

def qs_ancestors(queryset):
    if isinstance(queryset, EmptyQuerySet):
        return queryset
    new_queryset = queryset.none()
    for obj in queryset:
        new_queryset = new_queryset | obj.get_ancestors()
return new_queryset

It’s still len(queryset) clauses. You could potentially reduce the number of clauses a bit by doing a preprocess pass that removes objects that are ancestors of other objects in the queryset, something like:

min_obj_set = []
for obj in queryset.order_by('tree_id', '-level'):
    for obj2 in min_obj_set:
        if obj.is_ancestor_of(obj2):

Although the above snippet is only an example, you’ll probably want to use a BST if your querset contains a significant amount of objects.

You’ll have to test if this yeilds an increase in speed vs. the larger DB query, though.


I had to write a similar algorithm once. I had a view displaying a MPTT tree, it’s was a very large tree so I couldn’t load all of it’s data in the HTML template. So I displayed only the root nodes in the initial load and used Ajax to load the other nodes.

It was working good until my boss asked me to implement a ‘search’ option. The search had to look in all nodes and explode the tree were it found a match. It took me a while to figure this out, but I finally got it. Here is the solution a came up with:

from django.db.models import Q

def get_parents(self, qs):
    tree_list = {}
    query = Q()
    for node in qs:
        if node.tree_id not in tree_list:
            tree_list[node.tree_id] = []

        parent =  node.parent.pk if node.parent is not None else None,

        if parent not in tree_list[node.tree_id]:

            query |= Q(lft__lt=node.lft, rght__gt=node.rght, tree_id=node.tree_id)

    return YourModel.objects.filter(query)

It needs only two queries to run, the initial qs passed as an argument and the final queryset returned by the function. The tree_list is a dictionary that stores nodes that were already added to the queryset, it’s an optimization and it’s not necessary for the algorithm to work. But since I was working with a relative big tree I had to include it.

I guess you could turn this method into a manager and make it more generic, i.e. make it work for any MPTT model and not only YourModel

Leave a comment