Project 4 — Leaderboard and Notifications

What you'll walk away with

Explain the core ideas behind Project 4 — Leaderboard and Notifications
Run the sample Redis command or code and verify its output
Apply the technique correctly to the Tutorial Platform and production scenarios

Build the mental model

Leaderboard scores should derive from durable completion events and be applied idempotently to a sorted set. Pub/Sub is only a UI refresh hint; reconnecting clients fetch the canonical leaderboard API. Season keys, tie policy, privacy, and rebuild tooling are project requirements.

Connect it to a real scenario

Check each PostgreSQL completion ID against an idempotency record before incrementing a weekly sorted set. Publish a rank-changed hint. The API returns the top 100 and the requesting user's neighborhood; an admin rebuild constructs a fresh key from DB events and switches atomically.

Derive rankings from durable events
Use Pub/Sub only as a refresh hint
Provide a rebuild-and-switch procedure

Try the working example

shell

SADD processed:completion:2026-W35 completion:9001
ZINCRBY leaderboard:2026-W35 1 user:42
ZREVRANGE leaderboard:2026-W35 0 99 WITHSCORES
ZREVRANK leaderboard:2026-W35 user:42
PUBLISH leaderboard:changed '{"season":"2026-W35","userId":42}'

You should see

You get an idempotent weekly score, rank API data, and a live refresh signal.

5-minute try-it

Write tests for top-three ties, deleted users, season rollover, and rebuilds during traffic.

One important caution

Separate non-atomic SADD and ZINCRBY calls can mark an event without scoring it after a crash; use a script or transaction.

Redis — Sorted Sets — Redis

Easy traps

Separate non-atomic SADD and ZINCRBY calls can mark an event without scoring it after a crash; use a script or transaction.
Validate sample commands on a local or test instance with recoverable data before applying them to production Redis.

Exercise

Write tests for top-three ties, deleted users, season rollover, and rebuilds during traffic.

You'll know it worked when: You get an idempotent weekly score, rank API data, and a live refresh signal.