Coding questions · Meta

Recommend friends-of-friends

The question evaluates competency in graph traversal, set operations, and adjacency-list manipulation for social-network style data. Commonly asked in Coding & Algorithms interviews for Data Engineer roles, it gauges the ability to reason about relationship aggregation, duplicate elimination, and algorithmic efficiency when computing friends-of-friends, representing a practical application of algorithmic and data-structure knowledge rather than a purely conceptual exercise.

Question

Given a dictionary such as {A:[B,C], B:[C,D], C:[E]}, return for a user U all people followed by U’s followees but not already followed by U. Order is irrelevant and duplicates must be removed.

Recommend Friends-of-Friends

You are given a directed "follows" graph as a dictionary mapping each user to the list of users they follow. For a given user U, return everyone followed by U's followees (the friends-of-friends) who is NOT already followed by U and is NOT U themselves.

Duplicates must be removed and order does not matter. For deterministic output, return the recommendations sorted in ascending order.

Example: given {A: [B, C], B: [C, D], C: [E]} and user A:

A follows B and C.
B follows C and D; C follows E.
Candidates from followees = {C, D, E}.
Remove C (A already follows C) and A itself.
Result = [D, E].

Constraints

Each key in the graph maps to a (possibly empty) list of followees.
A user may appear as a followee without being a key in the graph (they follow no one).
Self-loops and mutual follows are possible and must be handled.
The recommended set excludes the user themselves and anyone the user already directly follows.
Results contain no duplicates.

Hints

First collect the set of users that U directly follows so membership checks are O(1).
Iterate over each direct followee, then over the people that followee follows.
Skip any candidate that equals U or is already in U's direct-follow set; collect the rest in a set to dedupe.