Hashes for Structured Data

Where we left off

A plain string holds one value. A session isn't one value, it's a small object: a user ID, when it started, how many turns it's had. You could serialize that into JSON and store it as one string, but Redis has a value type built for exactly this shape.

Hashes: a key holding a small field/value map

r.hset("session:abc123", mapping={
    "user_id": "u42",
    "started_at": "2026-08-07T10:00:00Z",
    "turn_count": 0,
})

A hash is one Redis key that maps to multiple field/value pairs, like a small dict living under a single key. HSET key mapping={...} sets several fields at once; a single field can also be set with HSET key field value.

Reading it back

session = r.hgetall("session:abc123")
turn_count = r.hget("session:abc123", "turn_count")

HGETALL key returns the whole hash as a Python dict. HGET key field reads just one field, cheaper than pulling the whole hash back when only one piece is needed.

Updating one field without touching the rest

r.hincrby("session:abc123", "turn_count", 1)

HINCRBY is INCR scoped to one field of a hash, atomic the same way: incrementing turn_count doesn't require reading the hash, changing it in Python, and writing the whole thing back, which would risk two concurrent turns clobbering each other's count.

Why a hash, and not a JSON string

SET key '{"user_id": "u42", "turn_count": 0}' would work, but every change would mean deserializing the whole string, editing one field in Python, and re-serializing all of it, and Redis itself couldn't increment turn_count atomically, it's just a string to Redis at that point. A hash lets Redis understand the structure, so HINCRBY and HGET operate on one field directly. (Lesson 13's RedisJSON module adds a third option: real nested JSON Redis understands the shape of, past a flat field/value map.)