ocean

In-memory key value store that saves your data to disk using JSON.

Installation

go get git.akyoto.dev/go/ocean

Example

// User type
type User struct { Name string }

// Create ~/.ocean/myapp/User.dat
users := ocean.NewFile[User]("myapp")

// Write
users.Set("1", &User{Name: "User 1"})
users.Set("2", &User{Name: "User 2"})
users.Set("3", &User{Name: "User 3"})

// Read
a, err := users.Get("1")
b, err := users.Get("2")
c, err := users.Get("3")

// Iterate
for user := range users.All() {
    fmt.Println(user.Name)
}

Format

1
{"name":"User 1"}
2
{"name":"User 2"}
3
{"name":"User 3"}

Benchmarks

BenchmarkGet-12         275126157                4.462 ns/op           0 B/op          0 allocs/op
BenchmarkSet-12          4796011               251.0 ns/op            32 B/op          2 allocs/op
BenchmarkDelete-12      471913158                2.530 ns/op           0 B/op          0 allocs/op
BenchmarkNew-12         48838576                22.89 ns/op           80 B/op          1 allocs/op

Storage systems

nil

You can specify nil as the storage system which will keep data in RAM only.

ocean.File

ocean.File uses a single file to store all records. Writes using Set(key, value) are async and only mark the collection as "dirty" which is very quick. The sync to disk happens shortly afterwards. Every collection uses one goroutine to check the "dirty" flag, write the new contents to disk and reset the flag.

The biggest advantage of ocean.File is that it scales well with the number of requests:

Suppose n is the number of write requests and io is the time it takes for one write. Immediate ocean would require O(n * io) time to complete all writes but the async behavior makes it O(n).

You should use ocean.File if you have a permanently running process such as a web server where end users expect quick responses and background work can happen after the user request has already been dealt with.

Make sure you defer collection.Sync() to ensure that queued writes will be handled when the process ends.