Lyft's 3-Part Rating Service Question

Lyft coding rounds favor practical modeling over puzzles. This rating service hands you the real domain (riders and drivers rate each other after a ride) and asks you to model it, ingest a JSON feed, and answer two queries. The hidden test is what received rating actually means. Here is the full question, Java for all three parts, and a detailed rubric.

The question

getRatingForRideAndUser(3, 9012) == 1
getAverageRatingOfUser(5678)    == (4 + 2) / 2 == 3.0

Part 1 and 2 · Model and ingest

Model a ride and a rating as plain records. Ingestion deserializes the JSON feed into lists, then indexes them: ratings by ride, and the set of rides each user took. Those indexes are what make the Part 3 queries fast.

import java.util.*;

class RatingService {
    record Ride(int rideId, int driverId, int riderId, long timestamp) {}
    record Rating(int rideId, int fromUserId, int rating) {}

    private final Map<Integer, List<Rating>> ratingsByRide = new HashMap<>();
    private final Map<Integer, List<Integer>> ridesByUser = new HashMap<>();

    // Part 2: ingest the parsed feed (e.g. deserialized with Jackson) and index it.
    void load(List<Ride> rides, List<Rating> ratings) {
        for (Ride r : rides) {
            ridesByUser.computeIfAbsent(r.driverId(), k -> new ArrayList<>()).add(r.rideId());
            ridesByUser.computeIfAbsent(r.riderId(), k -> new ArrayList<>()).add(r.rideId());
        }
        for (Rating rt : ratings)
            ratingsByRide.computeIfAbsent(rt.rideId(), k -> new ArrayList<>()).add(rt);
    }
}

Part 3 · The two APIs

Here is the crux. The rating a user received on a ride is the one the other participant submitted, the rating whose author is not that user. The average is taken over every ride the user was part of.

    // The rating the user RECEIVED = submitted by the counterparty on that ride.
    Integer getRatingForRideAndUser(int rideId, int userId) {
        for (Rating rt : ratingsByRide.getOrDefault(rideId, List.of()))
            if (rt.fromUserId() != userId) return rt.rating();
        return null;   // counterparty never rated
    }

    // Average of all ratings the user received across their rides.
    double getAverageRatingOfUser(int userId) {
        int sum = 0, count = 0;
        for (int rideId : ridesByUser.getOrDefault(userId, List.of())) {
            Integer received = getRatingForRideAndUser(rideId, userId);
            if (received != null) { sum += received; count++; }
        }
        return count == 0 ? 0.0 : (double) sum / count;
    }

Check it against the examples: on ride 3, user 9012 received the rating authored by 6543, which is 1; user 5678 drove rides 1 and 2 and received 4 and 2, averaging 3.0. Read received as the counterparty submitted and the rest follows.

How Lyft scores it

Lyft interview FAQ