sugarcane

sugarcane

The sweet embedded database in Go

Sugarcane is a persistent data-store for Go structures. Data is saved with the output of encoding/gob to disk, from which it can be loaded as byte buffer and ultimately used again as a native structure. You could for example save application progress in native Go structure, so you don't need to encode your data to different format, say JSON.

Databases saved by sugarcane are also lightweight; one million lines of three field structures weight only around 50-100MB.

Sugarcane neither has any dependencies outside of the Go standard library.

Saving object into disk is as easy as

type Person struct {
	Name string
	Age  int
}

var p Person
p.Name = "foo"
p.Visits = 3

db, _ := sugarcane.Open("./person_table")

db.Insert(p)

You can then read a single structure with

db.Scan(&p)

You can also read the whole file with for loop

var persons []Person

for i := 0; ; i++ {
	var q Person
	db, err := db.Scan(&q)
	if err == io.EOF {
		break
	}
	persons = append(persons, p)
}

fmt.Println(persons)

##Performance

At the moment, sugarcane is rather naive implementation and does not include any optimizations whatsoever. Therefore, it should be no surprise that sugarcane is not as fast as, say PostgreSQL. To get a better look at the performance, git clone the repository and run go test -bench=".*". On my Macbook I get the following results:

BenchmarkInsert	  200000	      8642 ns/op
BenchmarkRead	   50000	     50118 ns/op
ok	4.873s

Inserting million lines of five field structures (two integers, three strings) takes:

real	0m20.372s
user	0m17.434s
sys		0m5.098s

Reading however...

real	0m57.875s
user	1m4.876s
sys		0m1.281s

This makes sugarcane about 50 times slower than PostgreSQL. Although, some simple benchmarks indicate that sugarcane would be up to two minutes faster than MongoDB when writing 10M rows, and equally fast to write data as Redis. Whether that is anything to believe, you will have to run the tests yourself.

##License

MIT