XADD & XREAD — Redis Streams

Pub/Sub is fire-and-forget — if no one's listening, messages vanish. Lists work as queues but lose history once consumed. Streams (added in Redis 5) are the missing piece: an append-only log that persists every entry, supports multiple consumers, and remembers what each consumer has seen.

This is Redis answering "what would Kafka look like, embedded?"

Data Model

A stream is an ordered sequence of entries. Each entry has:

• A monotonically increasing ID (<milliseconds>-<sequence>, e.g. 1700000000123-0)
• A set of field/value pairs (like a hash)

events:  [ id=1-0  {action: signup, user: alice} ]
         [ id=2-0  {action: login,  user: alice} ]
         [ id=3-0  {action: signup, user: bob}   ]

The ID is the cursor — readers say "give me everything after 2-0" and they get the new entries.

Core Commands

Write side:

• XADD <stream> <id> field value [field value ...] — append entry. <id> can be * (auto-generate based on current time + counter) or explicit (123-0). Returns the assigned ID.
• XLEN <stream> — total entry count.

Read side:

• XRANGE <stream> <start> <end> — historical slice. - = first, + = last. Returns all entries in range.
• XREAD COUNT <n> STREAMS <stream> <last_id> — read entries newer than last_id. Returns up to n entries.
• XREAD COUNT <n> BLOCK <ms> STREAMS <stream> $ — blocking read for new entries (the $ means "from the current end"). This is the pub/sub-like mode.

Consumer groups (advanced — not in this lesson):

• XGROUP CREATE / XREADGROUP / XACK — multiple workers cooperatively process entries, like Kafka consumer groups.

ID Monotonicity

IDs must be strictly increasing. If you try XADD s 5-0 ... then XADD s 3-0 ..., you get:

-ERR The ID specified in XADD is equal or smaller than the target stream top item

This guarantee is what makes streams safe for replay: every consumer can rely on IDs as a strict total order.

Why Streams Beat Lists for Messaging

Real-World Uses

• Event sourcing — append every domain event, derive read models by replay
• Activity feeds — Twitter-style timelines
• Job queues with audit trail — even after a job runs, the history is queryable
• CDC pipelines — capture changes from a primary store, downstream consumers replay

What You'll Build

XADD, XLEN, XRANGE, and a basic XREAD. Consumer groups are a follow-on you can layer on top once these primitives are in place.