node_id parent_id
1       0
2       1
3       1
4       2
...

import pandas as pd
from collections import defaultdict

# phase 1
nodes_df = pd.read_csv('input.csv')

# phase 2
node_ancestors = defaultdict(set)
node_ancestors[0] = set([0])
for id, ndata in nodes_df1.iterrows():
    node_ancestors[ndata['node_id']].add(ndata['node_id'])
    node_ancestors[ndata['node_id']].update(node_ancestors[ndata['parent_id']])

# phase 3
node_descendants = defaultdict(set)
node_descendants[0] = set([0])
for id, ndata in nodes_df1[::-1].iterrows():
    node_descendants[ndata['node_id']].add(ndata['node_id'])
    node_descendants[ndata['parent_id']].
      update(node_descendants[ndata['node_id']])

node_ancestors = defaultdict(set)
node_ancestors[0] = set([0])
node_descendants = defaultdict(set)
node_descendants[0] = set([0])
for _, node, parent in df.itertuples():
    node_ancestors[node].add(node)
    node_ancestors[node].update(node_ancestors[parent])

    for ancestor in node_ancestors[node]:
        node_descendants[ancestor].add(node)

def parse_tree(df):
    node_ancestors = defaultdict(set)
    node_ancestors[0] = set([0])
    node_descendants = defaultdict(set)
    node_descendants[0] = set([0])
    for _, node, parent in df.itertuples():
        node_ancestors[node].add(node)
        node_ancestors[node].update(node_ancestors[parent])

        for ancestor in node_ancestors[node]:
            node_descendants[ancestor].add(node)

    return node_ancestors, node_descendants