… a Redis client for the Go programming language.
- Boring API with full type-safety
- Automatic pipelining on concurrency
- High throughput despite low footprint
- Efficient OS and network utilization
- Robust error recovery
Network I/O is executed in the same goroutine that invoked the Client. There is no internal error reporting/logging by design.
This is free and unencumbered software released into the public domain.
Version 2 [github.com/pascaldekloe/redis/v2] utilizes generics.
// Redis is a thread-safe connection establishment.
var Redis = redis.NewDefaultClient[string,string]("rds1.example.com")
// Grow adds a string to a list.
func Grow() {
newLen, err := Redis.RPUSH("demo_list", "hello")
if err != nil {
log.Print("demo_list update error: ", err)
return
}
log.Printf("demo_list has %d elements", newLen)
}
// Ping pushes a message to a publish–subscribe channel.
func Ping() {
clientCount, err := Redis.PUBLISH("demo_channel", "ping")
if err != nil {
log.Print("demo_channel publish error: ", err)
return
}
log.Printf("pinged %d clients in demo_channel", clientCount)
}Zero-Copy Pub/Sub Reception
// RedisListener is a thread-safe connection establishment.
var RedisListener = redis.NewListener(redis.ListenerConfig{
Func: func(channel string, message []byte, err error) {
switch err {
case nil:
log.Printf("received %q on %q", message, channel)
case redis.ErrClosed:
log.Print("subscription establishment terminated")
case io.ErrShortBuffer: // see ListenerConfig BufferSize
log.Printf("message on %q skipped due size", channel)
default:
log.Print("subscription error: ", err)
// recovery attempts follow automatically
}
},
Addr: "rds1.example.com:6379",
})
// Peek listens shortly to a publish–subscribe channel.
func Peek() {
RedisListener.SUBSCRIBE("demo_channel")
time.Sleep(time.Second)
RedisListener.UNSUBSCRIBE("demo_channel")
}- Memory allocations are limited to user data only.
- Subscription receival does not allocate any memory.
- Unix domain sockets significantly reduce latency.
- TCP performs on large payloads and high concurrency.
Configure Redis with client-output-buffer-limit pubsub 512mb 512mb 60 for the benchmarks.
The following results were measured with Redis version 7, and Go version 1.20, on an Apple M1.
goos: darwin
goarch: arm64
pkg: github.com/pascaldekloe/redis/v2
│ tcp.txt │ unix.txt │
│ sec/op │ sec/op vs base │
SimpleString/sequential-8 18.507µ ± 1% 6.452µ ± 1% -65.14% (p=0.000 n=8)
SimpleString/parallel-8 3.603µ ± 0% 1.631µ ± 1% -54.74% (p=0.000 n=8)
Integer/sequential-8 18.211µ ± 0% 6.327µ ± 0% -65.26% (p=0.000 n=8)
Integer/parallel-8 3.499µ ± 0% 1.618µ ± 0% -53.77% (p=0.000 n=8)
Bulk/8B/sequential-8 18.419µ ± 0% 6.418µ ± 0% -65.15% (p=0.000 n=8)
Bulk/8B/parallel-8 3.541µ ± 0% 1.631µ ± 1% -53.95% (p=0.000 n=8)
Bulk/800B/sequential-8 18.724µ ± 0% 6.717µ ± 0% -64.13% (p=0.000 n=8)
Bulk/800B/parallel-8 3.620µ ± 4% 2.034µ ± 0% -43.80% (p=0.000 n=8)
Bulk/8000B/sequential-8 21.159µ ± 1% 8.904µ ± 0% -57.92% (p=0.000 n=8)
Bulk/8000B/parallel-8 5.226µ ± 5% 6.201µ ± 0% +18.65% (p=0.000 n=8)
Array/1×8B/sequential-8 18.841µ ± 0% 6.659µ ± 0% -64.66% (p=0.000 n=8)
Array/1×8B/parallel-8 3.946µ ± 4% 1.722µ ± 2% -56.36% (p=0.000 n=8)
Array/12×8B/sequential-8 19.996µ ± 1% 7.755µ ± 0% -61.22% (p=0.000 n=8)
Array/12×8B/parallel-8 4.172µ ± 4% 2.233µ ± 5% -46.49% (p=0.000 n=8)
Array/144×8B/sequential-8 31.16µ ± 0% 19.37µ ± 1% -37.85% (p=0.000 n=8)
Array/144×8B/parallel-8 9.729µ ± 3% 9.110µ ± 3% -6.36% (p=0.000 n=8)
PubSub/8B/1publishers-8 4.572µ ± 0% 1.928µ ± 3% -57.84% (p=0.000 n=8)
PubSub/8B/2publishers-8 2.631µ ± 0% 1.530µ ± 4% -41.83% (p=0.000 n=8)
PubSub/8B/16publishers-8 1.411µ ± 1% 1.335µ ± 8% -5.39% (p=0.000 n=8)
PubSub/800B/1publishers-8 4.845µ ± 1% 2.241µ ± 2% -53.75% (p=0.000 n=8)
PubSub/800B/2publishers-8 2.815µ ± 2% 1.831µ ± 3% -34.96% (p=0.000 n=8)
PubSub/800B/16publishers-8 1.435µ ± 1% 1.933µ ± 5% +34.67% (p=0.000 n=8)
PubSub/8000B/1publishers-8 5.471µ ± 0% 13.661µ ± 16% +149.71% (p=0.000 n=8)
PubSub/8000B/2publishers-8 5.185µ ± 2% 11.883µ ± 13% +129.20% (p=0.000 n=8)
PubSub/8000B/16publishers-8 5.009µ ± 2% 13.477µ ± 11% +169.05% (p=0.000 n=8)
geomean 6.465µ 4.096µ -36.65%
│ tcp.txt │ unix.txt │
│ B/s │ B/s vs base │
Bulk/8B/sequential-8 419.9Ki ± 0% 1220.7Ki ± 1% +190.70% (p=0.000 n=8)
Bulk/8B/parallel-8 2.155Mi ± 0% 4.678Mi ± 1% +117.04% (p=0.000 n=8)
Bulk/800B/sequential-8 40.75Mi ± 0% 113.59Mi ± 0% +178.78% (p=0.000 n=8)
Bulk/800B/parallel-8 210.7Mi ± 3% 375.0Mi ± 0% +77.96% (p=0.000 n=8)
Bulk/8000B/sequential-8 360.6Mi ± 0% 856.9Mi ± 0% +137.64% (p=0.000 n=8)
Bulk/8000B/parallel-8 1.426Gi ± 3% 1.202Gi ± 1% -15.72% (p=0.000 n=8)
Array/1×8B/sequential-8 410.2Ki ± 2% 1171.9Ki ± 0% +185.71% (p=0.000 n=8)
Array/1×8B/parallel-8 1.931Mi ± 2% 4.435Mi ± 3% +129.63% (p=0.000 n=8)
Array/12×8B/sequential-8 4.578Mi ± 0% 11.806Mi ± 0% +157.92% (p=0.000 n=8)
Array/12×8B/parallel-8 21.94Mi ± 2% 41.01Mi ± 4% +86.90% (p=0.000 n=8)
Array/144×8B/sequential-8 35.26Mi ± 0% 56.72Mi ± 2% +60.87% (p=0.000 n=8)
Array/144×8B/parallel-8 112.9Mi ± 1% 120.6Mi ± 3% +6.79% (p=0.000 n=8)
PubSub/8B/1publishers-8 1.669Mi ± 1% 3.958Mi ± 8% +137.14% (p=0.000 n=8)
PubSub/8B/2publishers-8 2.899Mi ± 0% 4.988Mi ± 4% +72.04% (p=0.000 n=8)
PubSub/8B/16publishers-8 5.407Mi ± 1% 5.717Mi ± 1% +5.73% (p=0.000 n=8)
PubSub/800B/1publishers-8 157.5Mi ± 1% 340.5Mi ± 9% +116.22% (p=0.000 n=8)
PubSub/800B/2publishers-8 271.1Mi ± 2% 416.8Mi ± 4% +53.75% (p=0.000 n=8)
PubSub/800B/16publishers-8 531.7Mi ± 0% 394.8Mi ± 7% -25.75% (p=0.000 n=8)
PubSub/8000B/1publishers-8 1394.7Mi ± 2% 558.9Mi ± 15% -59.92% (p=0.000 n=8)
PubSub/8000B/2publishers-8 1471.6Mi ± 7% 642.3Mi ± 16% -56.35% (p=0.000 n=8)
PubSub/8000B/16publishers-8 1523.2Mi ± 2% 566.2Mi ± 11% -62.83% (p=0.000 n=8)
geomean 38.78Mi 56.13Mi +44.74%