The high performance GC-free cache library for Golang.
- Fast set/get, scales well with the number of goroutines
- High hit rate, benefit from LRU & custom eviction policies
- GC friendly, almost zero GC overhead
- Zero-copy access
It requires Go 1.12 or newer.
go get github.com/qianbin/directcache// capacity will be set to directcache.MinCapacity
c := directcache.New(0)
key := []byte("DirectCache")
val := []byte("is awesome,")
c.Set(key, val)
got, ok := c.Get(key)
fmt.Println(string(key), string(got), ok)
// Output: DirectCache is awesome, truezero-copy access
c.AdvGet(
key,
func(val []byte){
// consume the value
},
true, // peek, which means the recently-used flag won't be added to the accessed entry.
)in-place value assignment
c.AdvSet(
key,
len(val),
func(_val []byte) {
copy(_val, val) // in-place assign value
},
)custom eviction policy
shouldEvict := func(entry directcache.Entry) bool {
if entry.RecentlyUsed() {
return false
}
// custom rules...
return true
}
c.SetEvictionPolicy(shouldEvict)dump entries
c.Dump(func(e directcache.Entry) bool {
// do something to e
return true
})The performance is compared with FreeCache, FastCache and BigCache. The code of benchmarks can be found under ./benches/.
$ cd benches
$ go test -run=^$ -bench ^Benchmark benches -benchtime=5s
goos: linux
goarch: amd64
pkg: benches
cpu: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
BenchmarkGet/DirectCache-8 39183045 154.3 ns/op 155.50 MB/s
BenchmarkGet/FreeCache-8 23686060 253.4 ns/op 94.71 MB/s
BenchmarkGet/FastCache-8 34900104 170.7 ns/op 140.59 MB/s
BenchmarkGet/BigCache-8 27393757 217.9 ns/op 110.13 MB/s
BenchmarkParallelGet/DirectCache-8 168825828 35.39 ns/op 678.14 MB/s
BenchmarkParallelGet/FreeCache-8 100000000 56.37 ns/op 425.77 MB/s
BenchmarkParallelGet/FastCache-8 170344753 35.95 ns/op 667.52 MB/s
BenchmarkParallelGet/BigCache-8 135971068 46.87 ns/op 512.09 MB/s
BenchmarkSet/DirectCache-8 32157844 348.8 ns/op 68.82 MB/s
BenchmarkSet/FreeCache-8 33666162 410.0 ns/op 58.54 MB/s
BenchmarkSet/FastCache-8 28295484 222.3 ns/op 107.98 MB/s
BenchmarkSet/BigCache-8 18730782 471.1 ns/op 50.95 MB/s
BenchmarkParallelSet/DirectCache-8 92843146 80.81 ns/op 296.98 MB/s
BenchmarkParallelSet/FreeCache-8 76079636 99.11 ns/op 242.16 MB/s
BenchmarkParallelSet/FastCache-8 97058389 81.22 ns/op 295.50 MB/s
BenchmarkParallelSet/BigCache-8 68715756 99.88 ns/op 240.30 MB/s
BenchmarkParallelSetGet/DirectCache-8 34167814 202.3 ns/op 118.65 MB/s
BenchmarkParallelSetGet/FreeCache-8 23563779 294.8 ns/op 81.40 MB/s
BenchmarkParallelSetGet/FastCache-8 33523118 185.4 ns/op 129.46 MB/s
BenchmarkParallelSetGet/BigCache-8 25198658 248.0 ns/op 96.77 MB/s
PASS
ok benches 172.131s$ go test -timeout 30s -run ^TestHitrate$ benches -v
=== RUN TestHitrate
=== RUN TestHitrate/DirectCache
benches_test.go:181: hits: 741644 misses: 258356 hitrate: 74.16%
=== RUN TestHitrate/DirectCache(custom_policy)
benches_test.go:181: hits: 784456 misses: 215544 hitrate: 78.45%
=== RUN TestHitrate/FreeCache
benches_test.go:181: hits: 727308 misses: 272692 hitrate: 72.73%
=== RUN TestHitrate/FastCache
benches_test.go:181: hits: 690139 misses: 309861 hitrate: 69.01%
=== RUN TestHitrate/BigCache
benches_test.go:181: hits: 697831 misses: 302169 hitrate: 69.78%
--- PASS: TestHitrate (5.22s)
--- PASS: TestHitrate/DirectCache (0.87s)
--- PASS: TestHitrate/DirectCache(custom_policy) (1.08s)
--- PASS: TestHitrate/FreeCache (1.01s)
--- PASS: TestHitrate/FastCache (1.12s)
--- PASS: TestHitrate/BigCache (1.14s)
PASS
ok benches 5.225sThe MIT License