Strings in Redis

Analogy categoryAnalogies

What Redis Strings Are

A Redis string is a binary-safe sequence of bytes, meaning it can store text, serialized JSON, images, or any raw payload up to a maximum size of 512MB. Unlike higher-level types like lists or hashes, a string is Redis's most basic unit of storage — SET key value overwrites whatever was there, and GET key retrieves it in O(1) time regardless of the string's length.

🏏

Cricket analogy: Think of a Redis string like a single scorecard slot for a batter's total runs — Sachin Tendulkar's tally is just one number you overwrite each ball, and reading it back takes the same instant whether he's on 0 or 200.

Basic Operations: SET, GET, and Expiration

Beyond simple SET and GET, Redis strings support conditional writes with NX (only if the key does not exist) and XX (only if it does exist), and can carry a time-to-live via EX seconds or PX milliseconds, turning a string into a self-expiring cache entry. Commands like MSET and MGET let you write or read multiple keys atomically in a single round trip, cutting network latency for bulk operations.

🏏

Cricket analogy: SET NX is like a scorer only recording a new batting record if the slot is empty — MS Dhoni's finishing-not-out tally stays untouched unless no prior record exists — while EX 3600 is like a rain-delay rule that auto-clears a provisional score after an hour.

bash

SET user:1001:name "Alice"
GET user:1001:name

# Cache with a 1-hour TTL
SET session:abc123 "token-data" EX 3600

# Only set if key does not already exist
SET counter:pageviews 0 NX

# Atomic increments
INCR counter:pageviews
INCRBY counter:pageviews 5
INCRBYFLOAT price:product:42 0.75

# Bulk write/read in one round trip
MSET key1 "v1" key2 "v2"
MGET key1 key2

# Bit operations for compact flags
SETBIT user:active:20260710 1001 1
BITCOUNT user:active:20260710

Atomic Counters with INCR and DECR

Because strings can hold integers, Redis provides INCR, DECRBY, and INCRBYFLOAT that atomically modify a numeric string in place without a read-modify-write race condition — critical for counters like page views, rate limiters, or inventory stock levels shared across many concurrent clients. INCR is O(1) and safe under concurrency because Redis executes commands single-threaded, so two clients calling INCR simultaneously never lose an update.

🏏

Cricket analogy: INCR is like the official run counter ticking up by exactly one every single run scored in an India vs Australia ODI, even if two scorers in the box are watching simultaneously — no run ever gets double-counted or dropped.

Redis's single-threaded command execution model means INCR, DECR, and similar read-modify-write commands are atomic without any external locking — this is one of the biggest reasons developers reach for Redis strings over an application-level counter backed by a relational database.

Binary Safety, Bit Operations, and Growth

Redis strings are binary-safe, so they can store raw bytes including null characters, which makes them suitable for serialized data, images, or bitmaps. Commands like SETBIT, GETBIT, and BITCOUNT let you treat a string as a compact bit array — useful for feature flags across millions of users or tracking daily active users with minimal memory — while APPEND lets you grow a string without a full rewrite, though it still copies memory once capacity is exceeded.

🏏

Cricket analogy: BITCOUNT is like tallying how many overs in a Test match had at least one boundary by scanning a compact bitmap of boundary/no-boundary flags per ball, far cheaper than replaying full ball-by-ball commentary.

APPEND and repeated small writes to a growing string can cause memory fragmentation over time, since Redis may need to allocate a larger contiguous block each time capacity is exceeded. For write-heavy, ever-growing data, consider a List or Stream instead of continuously appending to a single String.

Redis strings are binary-safe byte sequences up to 512MB, used for caching, tokens, counters, and serialized data.
SET/GET are O(1); NX and XX flags give conditional writes, and EX/PX attach expiration in one command.
MSET and MGET batch multiple key writes/reads into a single round trip, reducing network latency.
INCR, DECRBY, and INCRBYFLOAT are atomic thanks to Redis's single-threaded execution model, making them ideal for counters.
SETBIT/GETBIT/BITCOUNT let you use a string as a compact bitmap for flags across millions of entities.
APPEND grows a string in place but can trigger memory reallocation, so ever-growing data is often better modeled as a List or Stream.