2192 All Ancestors of a Node in a Directed Acyclic Graph

Intuition

An ancestor in a graph is a node that can be used to traverse further to some other distinct node. It depends on the problem, whether they mention it if a node can be an ancestor for itself.

So, since graphs are recursive structures, you may find some sub-problem that would solve our purpose.

Such as, the number of ancestors a node u has, is the number of ancestors all of it’s past ancestors have. For this one would require to know the parents of a child node. Luckily, we have a DAG and are given an edge list - we can just store the graph in a reverse order.

But you can imagine, when we calculate an ancestor list anc_list_u for a node u and when u will be used as an ancestor for some other node v. The ancestor list anc_list_v for node v would consist duplicates of all ancestors in anc_list_u and anc_list_v.

So to tackle that, our sub-problem computes ancestors in a set to avoid duplicates. The problem also asks us that every thing is sorted, so we throw in a sorted() there too.

Code

Python3

def getAncestors(self, n: int, edges: List[List[int]]) -> List[List[int]]:
 
    revAdj = collections.defaultdict(list)
    children = set()
 
    for u, v in edges:
        revAdj[v].append(u)
        children.add(u)
        children.add(v)
    '''
    store the graph in a reverse order, meaning, every child is pointing to it's parent
    '''
 
    rootNodes = set(revAdj.keys()) - children
    '''
    rootNodes are the nodes with in-degree == 0
    '''
 
    def helper(g, n: int, m) -> List[int]:
        '''
        recursively compute all the ancestors of a node in a reversed DAG
 
        returns: list of all ancestors
        '''
        if n in m: return m[n]
        if n not in g: return []
 
        anc = set()
 
        for parent in g[n]:
            if parent not in rootNodes:
                anc |= set(helper(g, parent, m))
                # or use the in-built set function set.update() -> it also accepts a list
                # anc.update(helper(g, parent, m))
                anc.add(parent)
 
        m[n] = sorted(list(anc))
 
        return m[n]
 
    memo = collections.defaultdict(int)
 
    for val in range(n):
        ancestors = helper(revAdj, val, memo)
        if len(ancestors) == 0:
            memo[val] = ancestors
 
 
    '''
    subproblem:
    return a list with all ancestors
 
    A - arrive on a node
    B - ancestors = ar.extend(ancestors_of_neighbor1).extend(ancestors_of_neighbor2).extend(so_on)
 
    base_case: traverse till you find a starting node
 
    what's a starting node?
    starting_nodes = set(children) - set(revAdj.keys())
 
 
    '''
 
    return [anc for node, anc in sorted(memo.items(), key=lambda x: x[0])]

Big O Analysis

• Runtime

  The runtime complexity here is $O(N\ast KlogK)$ where K is the number of ancestors for every node and N is the number of vertices.

• Memory

  The memory usage is $O(V+E)$ since we store an adjacency list that is based on a edge list.

— A

GitHub | Twitter