Golang JSON parser which strong?

This is a creation in Article, where the information may have evolved or changed.

Full text link:https://github.com/json-iterator/go-benchmark

The aim is not to sell json-iterator. Instead, it proves that json-iterator is no slower than the rest of the library, so that you can put the spit point in other ways: For example, the feature is not complete, the API is not friendly. The re-inventing of the JSON parser is because it is often necessary to deal with the strange format of JSON, and do not want to turn the data two times. There is no API in the market to meet my needs, I will write a special article on the API to demonstrate the uniqueness of json-iterator. (Https://github.com/json-itera ... ）

• Jsonparser:https://github.com/buger/jsonparser

• Jsoniter Pull-api:https://github.com/json-iterator/go

• Jsoniter Reflect-api:https://github.com/json-iterator/go/blob/master/jsoniter_reflect.go

• Encoding/json:golang Standard Lib

• Easy JSON:https://github.com/mailru/easyjson

Test equipment

• cpu:i7-6700k @ 4.0G

• Level 1 Cache size:4 x KB 8-way set associative instruction caches

• Level 2 cache Size:4 x KB 4-way set associative caches

• Level 3 Cache Size:8 MB 16-way set associative shared cache

• Go:1.8beta1

Small payload

Https://github.com/json-itera ...

Jsonparser	Jsoniter Pull-api	Jsoniter Reflect-api	Encoding/json	Easyjson
599 NS/OP	515 Ns/op	684 ns/op	2453 Ns/op	687 Ns/op
B/op	B/op	B/op	864 b/op	B/op
2 allocs/op	2 allocs/op	4 allocs/op	Allocs/op	2 allocs/op

Encoding/json has a good performance on i7-6700k, but it's much slower to cache on smaller CPUs.

Medium payload

Jsonparser	Jsoniter Pull-api	Jsoniter Reflect-api	Encoding/json	Easyjson
5238 ns/op	4111 ns/op	4708 ns/op	24939 ns/op	7361 ns/op
104 B/op	104 B/op	368 b/op	808 B/op	248 b/op
4 allocs/op	4 allocs/op	Allocs/op	Allocs/op	8 Allocs/op

Https://github.com/json-itera ...

Json-iterator's reflection API is also fairly fast.

Large payload

Https://github.com/json-itera ...

Jsonparser	Jsoniter Pull-api	Encoding/json
38334 ns/op	38463 ns/op	290778 ns/op
0 b/op	0 b/op	2128 B/op
0 allocs/op	0 allocs/op	Allocs/op

Jsonparser is a little faster when most fields are not used.

Large file

Test file Used:https://github.com/json-itera ...

Jsonparser	Jsoniter Pull-api	Encoding/json
42698634 ns/op	37760014 ns/op	235354502 ns/op
67107104 b/op	4248 B/op	71467896 b/op
Allocs/op	5 allocs/op	272477 allocs/op

Jsonparser and other a whole bunch of JSON parsers are "]byte" as input, just for running points. On this point Jackson's author also has a spit groove (https://www.infoq.com/news/20 ... ）。 and jsoniter can support IO. Reader as input, very friendly for large file handling.