rfyiamcool/redka

Redka aims to reimplement the good parts of Redis with SQLite, while remaining compatible with Redis API.

Notable features:

Data does not have to fit in RAM.
ACID transactions.
SQL views for better introspection and reporting.
Both in-process (Go API) and standalone (RESP) servers.
Redis-compatible commands and wire protocol.

This is a work in progress. See below for the current status and roadmap.

Commands • Installation • Usage • Persistence • Performance • Roadmap • More

Commands

Redka aims to support five core Redis data types: strings, lists, sets, hashes, and sorted sets.

Strings

Strings are the most basic Redis type, representing a sequence of bytes. Redka supports the following string-related commands:

Command	Go API	Description
`DECR`	`DB.Str().Incr`	Decrements the integer value of a key by one.
`DECRBY`	`DB.Str().Incr`	Decrements a number from the integer value of a key.
`GET`	`DB.Str().Get`	Returns the value of a key.
`GETSET`	`DB.Str().GetSet`	Sets the key to a new value and returns the prev value.
`INCR`	`DB.Str().Incr`	Increments the integer value of a key by one.
`INCRBY`	`DB.Str().Incr`	Increments the integer value of a key by a number.
`INCRBYFLOAT`	`DB.Str().IncrFloat`	Increments the float value of a key by a number.
`MGET`	`DB.Str().GetMany`	Returns the values of one or more keys.
`MSET`	`DB.Str().SetMany`	Sets the values of one or more keys.
`MSETNX`	`DB.Str().SetManyNX`	Sets the values of one or more keys when all keys don't exist.
`PSETEX`	`DB.Str().SetExpires`	Sets the value and expiration time (in ms) of a key.
`SET`	`DB.Str().Set`	Sets the value of a key.
`SETEX`	`DB.Str().SetExpires`	Sets the value and expiration (in sec) time of a key.
`SETNX`	`DB.Str().SetNotExists`	Sets the value of a key when the key doesn't exist.

The following string-related commands are not planned for 1.0:

APPEND  GETDEL  GETEX  GETRANGE  LCS  SETRANGE  STRLEN  SUBSTR

Lists

Lists are lists of strings sorted by insertion order. Redka aims to support the following list-related commands in 1.0:

LINDEX  LINSERT  LLEN  LPOP  LPUSHX  LRANGE  LREM  LSET
LTRIM  RPOP  RPOPLPUSH  RPUSH  RPUSHX

Sets

Sets are unordered collections of unique strings. Redka aims to support the following set-related commands in 1.0:

SADD  SCARD  SDIFF  SDIFFSTORE  SINTER  SINTERSTORE
SISMEMBER  SMEMBERS  SMOVE  SPOP  SRANDMEMBER  SREM
SUNION  SUNIONSTORE

Hashes

Hashes are field-value (hash)maps. Redka supports the following hash-related commands:

Command	Go API	Description
`HDEL`	`DB.Hash().Delete`	Deletes one or more fields and their values.
`HEXISTS`	`DB.Hash().Exists`	Determines whether a field exists.
`HGET`	`DB.Hash().Get`	Returns the value of a field.
`HGETALL`	`DB.Hash().Items`	Returns all fields and values.
`HINCRBY`	`DB.Hash().Incr`	Increments the integer value of a field.
`HINCRBYFLOAT`	`DB.Hash().IncrFloat`	Increments the float value of a field.
`HKEYS`	`DB.Hash().Keys`	Returns all fields.
`HLEN`	`DB.Hash().Len`	Returns the number of fields.
`HMGET`	`DB.Hash().GetMany`	Returns the values of multiple fields.
`HMSET`	`DB.Hash().SetMany`	Sets the values of multiple fields.
`HSCAN`	`DB.Hash().Scanner`	Iterates over fields and values.
`HSET`	`DB.Hash().SetMany`	Sets the values of one ore more fields.
`HSETNX`	`DB.Hash().SetNotExists`	Sets the value of a field when it doesn't exist.
`HVALS`	`DB.Hash().Exists`	Returns all values.

The following hash-related commands are not planned for 1.0:

HRANDFIELD  HSTRLEN

Sorted sets

Sorted sets are collections of unique strings ordered by each string's associated score. Redka aims to support the following sorted set related commands in 1.0:

ZADD  ZCARD  ZCOUNT  ZINCRBY  ZINTERSTORE  ZRANGE
ZRANK  ZREM  ZSCORE

Key management

Redka supports the following key management (generic) commands:

Command	Go API	Description
`DEL`	`DB.Key().Delete`	Deletes one or more keys.
`EXISTS`	`DB.Key().Count`	Determines whether one or more keys exist.
`EXPIRE`	`DB.Key().Expire`	Sets the expiration time of a key (in seconds).
`EXPIREAT`	`DB.Key().ExpireAt`	Sets the expiration time of a key to a Unix timestamp.
`KEYS`	`DB.Key().Keys`	Returns all key names that match a pattern.
`PERSIST`	`DB.Key().Persist`	Removes the expiration time of a key.
`PEXPIRE`	`DB.Key().Expire`	Sets the expiration time of a key in ms.
`PEXPIREAT`	`DB.Key().ExpireAt`	Sets the expiration time of a key to a Unix ms timestamp.
`RANDOMKEY`	`DB.Key().Random`	Returns a random key name from the database.
`RENAME`	`DB.Key().Rename`	Renames a key and overwrites the destination.
`RENAMENX`	`DB.Key().RenameNotExists`	Renames a key only when the target key name doesn't exist.
`SCAN`	`DB.Key().Scanner`	Iterates over the key names in the database.

The following generic commands are not planned for 1.0:

COPY  DUMP  EXPIRETIME  MIGRATE  MOVE  OBJECT  PEXPIRETIME
PTTL  RESTORE  SORT  SORT_RO  TOUCH  TTL  TYPE  UNLINK
WAIT  WAITAOF

Transactions

Redka supports the following transaction commands:

Command	Go API	Description
`DISCARD`	`DB.View` / `DB.Update`	Discards a transaction.
`EXEC`	`DB.View` / `DB.Update`	Executes all commands in a transaction.
`MULTI`	`DB.View` / `DB.Update`	Starts a transaction.

Unlike Redis, Redka's transactions are fully ACID, providing automatic rollback in case of failure.

The following transaction commands are not planned for 1.0:

UNWATCH  WATCH

Server/connection management

Redka supports only a couple of server and connection management commands:

Command	Go API	Description
`ECHO`	—	Returns the given string.
`FLUSHDB`	`DB.Key().DeleteAll`	Remove all keys from the database.

The rest of the server and connection management commands are not planned for 1.0.

Installation

Redka can be installed as a standalone Redis-compatible server, or as a Go module for in-process use.

Standalone server

Redka server is a single-file binary. Download it from the releases.

Linux (x86 CPU only):

curl -L -O "https://github.com/nalgeon/redka/releases/download/v0.2.0/redka_linux_amd64.zip"
unzip redka_linux_amd64.zip
chmod +x redka

macOS (both x86 and ARM/Apple Silicon CPU):

curl -L -O "https://github.com/nalgeon/redka/releases/download/v0.2.0/redka_darwin_amd64.zip"
unzip redka_darwin_amd64.zip
# remove the build from quarantine
# (macOS disables unsigned binaries)
xattr -d com.apple.quarantine redka
chmod +x redka

Or pull with Docker as follows (x86/ARM):

docker pull nalgeon/redka

Or build from source (requires Go 1.22 and GCC):

git clone https://github.com/nalgeon/redka.git
cd redka
make setup build
# the path to the binary after the build
# will be ./build/redka

Go module

Install the module as follows:

go get github.com/nalgeon/redka

You'll also need an SQLite driver. Use github.com/mattn/go-sqlite3 if you don't mind CGO. Otherwise use a pure Go driver modernc.org/sqlite. Install either with go get like this:

go get github.com/mattn/go-sqlite3

Usage

Redka can be used as a standalone Redis-compatible server, or as an embeddable in-process server with Go API.

Standalone server

Redka server is a single-file binary. After downloading and unpacking the release asset, run it as follows:

redka [-h host] [-p port] [db-path]

For example:

./redka
./redka data.db
./redka -h 0.0.0.0 -p 6379 data.db

Server defaults are host localhost, port 6379 and empty DB path.

Running without a DB path creates an in-memory database. The data is not persisted in this case, and will be gone when the server is stopped.

You can also run Redka with Docker as follows:

# database inside the container
# will be lost when the container stops
docker run --rm -p 6379:6379 nalgeon/redka

# persistent database
# using the /path/to/data host directory
docker run --rm -p 6379:6379 -v /path/to/data:/data nalgeon/redka

# in-memory database, custom post
docker run --rm -p 6380:6380 nalgeon/redka redka -h 0.0.0.0 -p 6380

Server defaults in Docker are host 0.0.0.0, port 6379 and DB path /data/redka.db.

Once the server is running, connect to it using redis-cli or an API client like redis-py or go-redis — just as you would with Redis.

redis-cli -h localhost -p 6379

127.0.0.1:6379> echo hello
"hello"
127.0.0.1:6379> set name alice
OK
127.0.0.1:6379> get name
"alice"

In-process server

The primary object in Redka is the DB. To open or create your database, use the redka.Open() function:

package main

import (
    "log"

    _ "github.com/mattn/go-sqlite3"
    "github.com/nalgeon/redka"
)

func main() {
    // Open or create the data.db file.
    db, err := redka.Open("data.db", nil)
    if err != nil {
        log.Fatal(err)
    }
    // Always close the database when you are finished.
    defer db.Close()
    // ...
}

Don't forget to import the driver (here I use github.com/mattn/go-sqlite3). Using modernc.org/sqlite is slightly different, see example/modernc/main.go for details.

To open an in-memory database that doesn't persist to disk, use the following path:

// All data is lost when the database is closed.
redka.Open("file:redka?mode=memory&cache=shared")

After opening the database, call redka.DB methods to run individual commands:

db.Str().Set("name", "alice")
db.Str().Set("age", 25)

count, err := db.Key().Count("name", "age", "city")
slog.Info("count", "count", count, "err", err)

name, err := db.Str().Get("name")
slog.Info("get", "name", name, "err", err)

count count=2 err=<nil>
get name="alice" err=<nil>

See the full example in example/simple/main.go.

Use transactions to batch commands. There are View (read-only transaction) and Update (writable transaction) methods for this:

updCount := 0
err := db.Update(func(tx *redka.Tx) error {
    err := tx.Str().Set("name", "bob")
    if err != nil {
        return err
    }
    updCount++

    err = tx.Str().Set("age", 50)
    if err != nil {
        return err
    }
    updCount++
    return nil
})
slog.Info("updated", "count", updCount, "err", err)

updated count=2 err=<nil>

See the full example in example/tx/main.go.

See the package documentation for API reference.

Persistence

Redka stores data in a SQLite database using the following tables:

rkey
---
id       integer primary key
key      text not null
type     integer not null    -- 1 string, 2 list, 3 set, 4 hash, 5 sorted set
version  integer not null    -- incremented when the key value is updated
etime    integer             -- expiration timestamp in unix milliseconds
mtime    integer not null    -- modification timestamp in unix milliseconds

rstring
---
key_id   integer not null    -- FK -> rkey.id
value    blob not null

rhash
---
key_id   integer not null    -- FK -> rkey.id
field    text not null
value    blob not null

To access the data with SQL, use views instead of tables:

select * from vstring;

┌────────┬──────┬───────┬───────┬─────────────────────┐
│ key_id │ key  │ value │ etime │        mtime        │
├────────┼──────┼───────┼───────┼─────────────────────┤
│ 1      │ name │ alice │       │ 2024-04-03 16:58:14 │
│ 2      │ age  │ 50    │       │ 2024-04-03 16:34:52 │
└────────┴──────┴───────┴───────┴─────────────────────┘

etime and mtime are in UTC.

There is a separate view for every data type:

vstring  vhash

Performance

I've compared Redka with Redis using redis-benchmark with the following parameters:

10 parallel connections
1000000 requests
10000 randomized keys
GET/SET commands

SQLite settings:

pragma journal_mode = wal;
pragma synchronous = normal;
pragma temp_store = memory;
pragma mmap_size = 268435456;
pragma foreign_keys = on;

Hardware: Apple M1 8-core CPU, 16GB RAM

Results:

redis-server --appendonly no
redis-benchmark -p 6379 -q -c 10 -n 1000000 -r 10000 -t get,set

SET: 133262.25 requests per second, p50=0.055 msec
GET: 139217.59 requests per second, p50=0.055 msec

./redka -p 6380 data.db
redis-benchmark -p 6380 -q -c 10 -n 1000000 -r 10000 -t get,set

SET: 22551.47 requests per second, p50=0.255 msec
GET: 56802.05 requests per second, p50=0.119 msec

So while Redka is 2-6 times slower than Redis (not surprising, since we are comparing a relational database to a key-value data store), it can still do 23K writes/sec and 57K reads/sec, which is pretty good if you ask me.

Note that running in a container may result in poorer performance.

Roadmap

The project is on its way to 1.0.

The 1.0 release will include the following features from Redis 2.x (which I consider the "golden age" of the Redis API):

✅ Strings.
⬜ Lists.
⬜ Sets.
✅ Hashes.
⏳ Sorted sets.
✅ Key management.
✅ Transactions.

✅ = done, ⏳ = in progress, ⬜ = next in line

Future versions may include additional data types (such as streams, HyperLogLog or geo), features like publish/subscribe, and more commands for existing types.

Features I'd rather not implement even in future versions:

Lua scripting.
Authentication and ACLs.
Multiple databases.
Watch/unwatch.

Features I definitely don't want to implement:

Cluster.
Sentinel.

More information

Contributing

Contributions are welcome. For anything other than bugfixes, please first open an issue to discuss what you want to change.

Be sure to add or update tests as appropriate.

Acknowledgements

Redka would not be possible without these great projects and their creators:

Redis (Salvatore Sanfilippo). It's such an amazing idea to go beyond the get-set paradigm and provide a convenient API for more complex data structures.
SQLite (D. Richard Hipp). The in-process database powering the world.
Redcon (Josh Baker). A very clean and convenient implementation of a RESP server.

Logo font by Ek Type.

License

The software is available under the BSD-3-Clause license.

Stay tuned

★ Subscribe to stay on top of new features.

rfyiamcool / redka

Commands

Strings

Lists

Sets

Hashes

Sorted sets

Key management

Transactions

Server/connection management

Installation

Standalone server

Go module

Usage

Standalone server

In-process server

Persistence

Performance

Roadmap

More information

Contributing

Acknowledgements

License

Stay tuned

About

Languages