My thinking is a recursive function which takes as input a list of node IDs. It runs a db call to get all of the children of each of the nodes. If the nodes have no children, they are deleted. Otherwise, the function is called on the childrens’ node IDs.

I think you'll find it conceptually easier if the function just takes a single node ID. If you want one for lists, just write a wrapper that calls the recursive one once for each node in the list. Other than that, it seems entirely reasonable to me.

One caveat: you'll need to be real sure you can't have any circular references. (It sounds like you won't easily be able to but you'll need to be careful if you allow node re-parenting.)

-sauoq
"My two cents aren't worth a dime.";