kpango / go-json-bench

golang json Unmarshal VS Decode Benchmark

This repository is Archived, no longer maintained.

Golang JSON Parse Benchmark

golang json Unmarshal VS Decode Benchmark

Description

There is Unmarshal and Decoder in golang's encoding/json package, and both can be used when parsing JSON data. In this repository, we compared the io.Reader parsing performance of both by performing a simple benchmark.

Code Example

json.Unmarshal

	var target JSON
	body, err := ioutil.ReadAll(reader)
	if err != nil {
		return err
	}
	return json.Unmarshal(body, &target)

json.Decoder

	var target JSON
	return json.NewDecoder(reader).Decode(&target)

Motivation

In golang, compare them to make a performance-oriented choice when parsing json

Installation

go get github.com/kpango/go-json-bench

Usage

Benchmark

make bench

Profiler

make profile

Cleaning

make clean

Benchmark Result

go test -count=10 -run=NONE -bench . -benchmem -o pprof/test.bin -cpuprofile pprof/cpu.out -memprofile pprof/mem.out
goos: darwin
goarch: amd64
pkg: github.com/kpango/json-bench
BenchmarkJSONUnmarshal-8           	  500000	      2925 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2958 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      3052 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2959 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2976 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2942 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2977 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2982 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2936 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshal-8           	  500000	      2933 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONDecode-8              	  500000	      2550 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2514 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2521 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2572 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2519 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2511 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2492 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2485 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2488 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecode-8              	  500000	      2476 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1194 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1236 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1243 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1268 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1251 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1344 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1360 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1277 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1264 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONUnmarshalParallel-8   	 1000000	      1312 ns/op	    2640 B/op	      10 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	       914 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	       966 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	       984 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	       992 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	      1062 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 1000000	      1039 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 1000000	      1021 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	      1026 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 1000000	      1016 ns/op	    1192 B/op	       9 allocs/op
BenchmarkJSONDecodeParallel-8      	 2000000	      1117 ns/op	    1192 B/op	       9 allocs/op
PASS
ok  	github.com/kpango/json-bench	65.207s

full result

Conclusion

In the case of io.Reader's input data, the execution speed was slightly faster using json.Decoder, and the byte transfer amount was about half.