Redis

What it is

Redis is the most common in-memory cache and quick-lookup store, and it usually sits one network hop away from data you already keep in DuckDB. The normal way to bridge the two is to write a script — pull keys out of Redis in Python or Node, reshape them, load them into a table. This extension removes that middle layer: it exposes Redis operations as redis_* SQL functions so the cache becomes just another source you can SELECT from, JOIN against, and write back to. It is scoped deliberately to the SQL-side workflows where data integration matters — draining a keyspace into a table, joining cached lookups against bulk data, priming a cache from analytical output, auditing keys — rather than to being a complete Redis client.

How it works

Connection details live in a DuckDB secret of TYPE redis, and every redis_* function takes a trailing argument naming that secret. This keeps credentials out of query text and lets dev / staging / prod live as separate secrets you select per query. There is no ATTACH interface and no virtual catalog — this is purely a function-based extension, so Redis access composes inline anywhere a scalar function or table function is allowed.

• • Secret-based configuration: A redis secret holds host, port, and an optional password. See CREATE SECRET; parameterize the password from the environment so it never lands in a cached query plan.
• • RESP over TCP: The extension speaks the native Redis Serialization Protocol directly over a plain TCP socket — no separate Redis daemon, sidecar, or driver process is involved.
• • Connection reuse: Connections to a given secret are pooled and reused across calls rather than reopened per row, so per-row lookups inside a SELECT and streaming table functions amortize the connection cost.

Production caveats

Be aware before pointing this at a busy Redis instance. The extension is honest-by-design about what it doesn't yet do.

• • Use SCAN, not KEYS: KEYS blocks the Redis server for the duration of the scan — fine in dev, dangerous on production keyspaces. Prefer the cursor-paginated SCAN-backed functions, which iterate without stalling traffic.
• • Strings, hashes, and lists only: No SET / ZSET / STREAM / PUB-SUB / TRANSACTIONS / SCRIPTING today. See Redis data types for what's supported elsewhere.
• • Single-node connections: No Redis Cluster topology awareness. Point the secret at a single endpoint.
• • Plaintext auth: The password field in the secret is sent over plain TCP — TLS is not yet wired up. Use the extension on trusted networks until that lands. See Redis security for the upstream guidance.
• • Experimental status: The function surface may change as more Redis primitives land. Pin a known-good extension version in production CTAS jobs.

Connect via a secret

CREATE SECRET IF NOT EXISTS redis (
  TYPE redis,
  PROVIDER config,
  host 'localhost',
  port '6379'
);

Read and write strings

CREATE SECRET IF NOT EXISTS redis (
  TYPE redis,
  PROVIDER config,
  host 'localhost',
  port '6379'
);

SELECT redis_set('user:1', 'John Doe', 'redis') AS ok;
SELECT redis_get('user:1', 'redis') AS name;

Discover keys matching a pattern

CREATE SECRET IF NOT EXISTS redis (
  TYPE redis,
  PROVIDER config,
  host 'localhost',
  port '6379'
);

SELECT * FROM redis_keys('user:*', 'redis');