This package provides a fast way to load large amounts of data into Go structs. Memdump can load datasets containing millions of small structs at over 1 GB/s (compared to ~30 MB/s for gob or json).

The price you pay is:

• you cannot load structs that contain maps or interfaces
• your data is not portable across machine architectures (64 bit vs 32 bit, big-endian vs small-endian)

The benchmarks were measured by encoding and decoding a tree containing 2,097,151 small structs. Code is under the bench dir.

Decode

                 gob    28.17 MB/s      (39.8 MB in 1.41s)
                json    30.17 MB/s      (113.8 MB in 3.77s)
             memdump  1031.54 MB/s      (113.2 MB in 0.11s)

Encode

                 gob    37.07 MB/s      (39.8 MB in 1.07s)
                json    77.20 MB/s      (113.8 MB in 1.47s)
             memdump    61.25 MB/s      (113.2 MB in 1.85s)

To reproduce these results:

$ go run ./bench/summarize.go

go get github.com/alexflint/go-memdump

Write data to a file:

type data struct {
	Foo string
	Bar int
}

w, err := os.Create("/tmp/data.memdump")
if err != nil {
	...
}

// note that you must pass a pointer when encoding
mydata := data{Foo: "abc", Bar: 123}
memdump.Encode(w, &mydata)

Load data from a file:

r, err := os.Open("/tmp/data.memdump")
if err != nil {
	...
}

// note that you muss pass a pointer to a pointer when decoding
var mydata *data
memdump.Decode(r, &mydata)