opt: add decoder Option NoValidateJSON for skipping JSON faster (#696)

Co-authored-by: liuqiang.06 <[email protected]>

.github/workflows/benchmark.yml

@@ -27,11 +27,12 @@ jobs:
             ${{ runner.os }}-go-
 
       - name: Benchmark Target
+        continue-on-error: true
         run: |
           export SONIC_NO_ASYNC_GC=1
-          export SONIC_BENCH_SINGLE=1
-          go test -run ^$ -count=10 -benchmem -bench 'Benchmark(Encoder|Decoder)_(Generic|Binding)_Sonic' ./decoder >> /var/tmp/sonic_bench_target.out
-          go test -run ^$ -count=10 -benchmem -bench 'Benchmark(Get|Set)One_Sonic|BenchmarkParseSeven_Sonic' ./ast >> /var/tmp/sonic_bench_target.out
+          go test -run ^$ -count=100 -benchmem -bench 'BenchmarkDecoder_(Generic|Binding)_Sonic' ./decoder >> /var/tmp/sonic_bench_target_${{ github.run_id }}.out
+          go test -run ^$ -count=100 -benchmem -bench 'BenchmarkEncoder_(Generic|Binding)_Sonic' ./encoder >> /var/tmp/sonic_bench_target_${{ github.run_id }}.out
+          go test -run ^$ -count=100 -benchmem -bench 'Benchmark(Get|Set)One_Sonic|BenchmarkParseSeven_Sonic' ./ast >> /var/tmp/sonic_bench_target_${{ github.run_id }}.out
 
       - name: Clear repository
         run: sudo rm -fr $GITHUB_WORKSPACE && mkdir $GITHUB_WORKSPACE
@@ -42,14 +43,15 @@ jobs:
           ref: main
 
       - name: Benchmark main
+        continue-on-error: true
         run: |
           export SONIC_NO_ASYNC_GC=1
-          export SONIC_BENCH_SINGLE=1
-          go test -run ^$ -count=10 -benchmem -bench 'Benchmark(Encoder|Decoder)_(Generic|Binding)_Sonic' ./decoder >> /var/tmp/sonic_bench_main.out
-          go test -run ^$ -count=10 -benchmem -bench 'Benchmark(Get|Set)One_Sonic|BenchmarkParseSeven_Sonic' ./ast >> /var/tmp/sonic_bench_main.out
+          UNIQUE_ID=${{ github.run_id }}
+          go test -run ^$ -count=100 -benchmem -bench 'BenchmarkDecoder_(Generic|Binding)_Sonic' ./decoder >> /var/tmp/sonic_bench_main_${{ github.run_id }}.out
+          go test -run ^$ -count=100 -benchmem -bench 'BenchmarkEncoder_(Generic|Binding)_Sonic' ./encoder >> /var/tmp/sonic_bench_main_${{ github.run_id }}.out
+          go test -run ^$ -count=100 -benchmem -bench 'Benchmark(Get|Set)One_Sonic|BenchmarkParseSeven_Sonic' ./ast >> /var/tmp/sonic_bench_main_${{ github.run_id }}.out
 
       - name: Diff bench
+        continue-on-error: true
         run: |
-          go get golang.org/x/perf/cmd/benchstat && go install golang.org/x/perf/cmd/benchstat
-          benchstat -format=csv /var/tmp/sonic_bench_target.out /var/tmp/sonic_bench_main.out
-      #   run: ./scripts/bench.py -t 0.05 -d /var/tmp/sonic_bench_target.out,/var/tmp/sonic_bench_main.out x
+          ./scripts/bench.py -t 0.20 -d /var/tmp/sonic_bench_target_${{ github.run_id }}.out,/var/tmp/sonic_bench_main_${{ github.run_id }}.out x

.gitignore

@@ -52,3 +52,4 @@ fuzz/testdata
 *__debug_bin*
 *pprof
 *coverage.txt
+tools/venv/*

README.md

@@ -283,19 +283,21 @@ sub := root.Get("key3").Index(2).Int64() // == 3
 **Tip**: since `Index()` uses offset to locate data, which is much faster than scanning like `Get()`, we suggest you use it as much as possible. And sonic also provides another API `IndexOrGet()` to underlying use offset as well as ensure the key is matched.
 
 #### SearchOption
+
 `Searcher` provides some options for user to meet different needs:
+
 ```go
 opts := ast.SearchOption{ CopyReturn: true ... }
 val, err := sonic.GetWithOptions(JSON, opts, "key")
 ```
+
 - CopyReturn
 Indicate the searcher to copy the result JSON string instead of refer from the input. This can help to reduce memory usage if you cache the results
 - ConcurentRead
 Since `ast.Node` use `Lazy-Load` design, it doesn't support Concurrently-Read by default. If you want to read it concurrently, please specify it.
 - ValidateJSON
 Indicate the searcher to validate the entire JSON. This option is enabled by default, which slow down the search speed a little.
 
-
 #### Set/Unset
 
 Modify the json content by Set()/Unset()
@@ -314,6 +316,7 @@ println(alias1 == alias2) // true
 exist, err := root.UnsetByIndex(1) // exist == true
 println(root.Get("key4").Check()) // "value not exist"
 ```
+
 #### Serialize
 
 To encode `ast.Node` as json, use `MarshalJson()` or `json.Marshal()` (MUST pass the node's pointer)
@@ -381,16 +384,12 @@ See [ast/visitor.go](https://github.com/bytedance/sonic/blob/main/ast/visitor.go
 
 ## Compatibility
 
-Sonic **DOES NOT** ensure to support all environments, due to the difficulty of developing high-performance codes. For developers who use sonic to build their applications in different environments, we have the following suggestions:
-
-- Developing on **Mac M1**: Make sure you have Rosetta 2 installed on your machine, and set `GOARCH=amd64` when building your application. Rosetta 2 can automatically translate x86 binaries to arm64 binaries and run x86 applications on Mac M1.
-- Developing on **Linux arm64**: You can install qemu and use the `qemu-x86_64 -cpu max` command to convert x86 binaries to amr64 binaries for applications built with sonic. The qemu can achieve a similar transfer effect to Rosetta 2 on Mac M1.
+For developers who want to use sonic to meet diffirent scenarios, we provide some integrated configs as `sonic.API`
 
-For developers who want to use sonic on Linux arm64 without qemu, or those who want to handle JSON strictly consistent with `encoding/json`, we provide some compatible APIs as `sonic.API`
-
-- `ConfigDefault`: the sonic's default config (`EscapeHTML=false`,`SortKeys=false`...) to run on sonic-supporting environment. It will fall back to `encoding/json` with the corresponding config, and some options like `SortKeys=false` will be invalid.
-- `ConfigStd`: the std-compatible config (`EscapeHTML=true`,`SortKeys=true`...) to run on sonic-supporting environment. It will fall back to `encoding/json`.
-- `ConfigFastest`: the fastest config (`NoQuoteTextMarshaler=true`) to run on sonic-supporting environment. It will fall back to `encoding/json` with the corresponding config, and some options will be invalid.
+- `ConfigDefault`: the sonic's default config (`EscapeHTML=false`,`SortKeys=false`...) to run sonic fast meanwhile ensure security.
+- `ConfigStd`: the std-compatible config (`EscapeHTML=true`,`SortKeys=true`...)
+- `ConfigFastest`: the fastest config (`NoQuoteTextMarshaler=true`) to run on sonic as fast as possible.
+Sonic **DOES NOT** ensure to support all environments, due to the difficulty of developing high-performance codes. On non-sonic-supporting environment, the implementation will fall back to `encoding/json`. Thus beflow configs will all equal to `ConfigStd`.
 
 ## Tips
 
@@ -480,8 +479,17 @@ For better performance, in previous case the `ast.Visitor` will be the better ch
 But `ast.Visitor` is not a very handy API. You might need to write a lot of code to implement your visitor and carefully maintain the tree hierarchy during decoding. Please read the comments in [ast/visitor.go](https://github.com/bytedance/sonic/blob/main/ast/visitor.go) carefully if you decide to use this API.
 
 ### Buffer Size
+
 Sonic use memory pool in many places like `encoder.Encode`, `ast.Node.MarshalJSON` to improve performace, which may produce more memory usage (in-use) when server's load is high. See [issue 614](https://github.com/bytedance/sonic/issues/614). Therefore, we introduce some options to let user control the behavior of memory pool. See [option](https://pkg.go.dev/github.com/bytedance/[email protected]/option#pkg-variables) package.
 
+### Faster JSON skip
+
+For security, sonic use [FSM](native/skip_one.c) algorithm to  validate JSON when decoding raw JSON or encoding `json.Marshaler`, which is much slower (1~10x) than [SIMD-searching-pair](native/skip_one_fast.c) algorithm. If user has many redundant JSON value and DO NOT NEED to strictly validate JSON correctness, you can enable below options:
+
+- `Config.NoValidateSkipJSON`: for faster skipping JSON when decoding, such as unknown fields, json.Unmarshaler(json.RawMessage), mismatched values, and redundant array elements
+- `Config.NoValidateJSONMarshaler`: avoid validating JSON when encoding `json.Marshaler`
+- `SearchOption.ValidateJSON`: indicates if validate located JSON value when `Get`
+
 ## Community
 
 Sonic is a subproject of [CloudWeGo](https://www.cloudwego.io/). We are committed to building a cloud native ecosystem.

README_ZH_CN.md

@@ -283,11 +283,14 @@ sub := root.Get("key3").Index(2).Int64() // == 3
 **注意**：由于 `Index()` 使用偏移量来定位数据，比使用扫描的 `Get()` 要快的多，建议尽可能的使用 `Index` 。 Sonic 也提供了另一个 API， `IndexOrGet()` ，以偏移量为基础并且也确保键的匹配。
 
 #### 查找选项
+
 `ast.Searcher`提供了一些选项，以满足用户的不同需求:
+
 ```
 opts:= ast.SearchOption{CopyReturn: true…}
 Val, err:= sonic。gettwithoptions (JSON, opts， "key")
 ```
+
 - CopyReturn
 指示搜索器复制结果JSON字符串，而不是从输入引用。如果用户缓存结果，这有助于减少内存使用
 - ConcurentRead
@@ -381,16 +384,12 @@ type Visitor interface {
 
 ## 兼容性
 
-由于开发高性能代码的困难性， Sonic **不**保证对所有环境的支持。对于在不同环境中使用 Sonic 构建应用程序的开发者，我们有以下建议：
-
-- 在 **Mac M1** 上开发：确保在您的计算机上安装了 Rosetta 2，并在构建时设置 `GOARCH=amd64` 。 Rosetta 2 可以自动将 x86 二进制文件转换为 arm64 二进制文件，并在 Mac M1 上运行 x86 应用程序。
-- 在 **Linux arm64** 上开发：您可以安装 qemu 并使用 `qemu-x86_64 -cpu max` 命令来将 x86 二进制文件转换为 arm64 二进制文件。qemu可以实现与Mac M1上的Rosetta 2类似的转换效果。
-
-对于希望在不使用 qemu 下使用 sonic 的开发者，或者希望处理 JSON 时与 `encoding/JSON` 严格保持一致的开发者，我们在 `sonic.API` 中提供了一些兼容性 API
+对于想要使用sonic来满足不同场景的开发人员，我们提供了一些集成配置:
 
-- `ConfigDefault`: 在支持 sonic 的环境下 sonic 的默认配置（`EscapeHTML=false`，`SortKeys=false`等）。行为与具有相应配置的 `encoding/json` 一致，一些选项，如 `SortKeys=false` 将无效。
-- `ConfigStd`: 在支持 sonic 的环境下与标准库兼容的配置（`EscapeHTML=true`，`SortKeys=true`等）。行为与 `encoding/json` 一致。
-- `ConfigFastest`: 在支持 sonic 的环境下运行最快的配置（`NoQuoteTextMarshaler=true`）。行为与具有相应配置的 `encoding/json` 一致，某些选项将无效。
+- `ConfigDefault`: sonic的默认配置 (`EscapeHTML=false`， `SortKeys=false`…) 保证性能同时兼顾安全性。
+- `ConfigStd`: 与 `encoding/json` 保证完全兼容的配置
+- `ConfigFastest`: 最快的配置(`NoQuoteTextMarshaler=true...`) 保证性能最优但是会缺少一些安全性检查（validate UTF8 等）
+Sonic **不**确保支持所有环境，由于开发高性能代码的困难。在不支持声音的环境中，实现将回落到 `encoding/json`。因此上述配置将全部等于`ConfigStd`。
 
 ## 注意事项
 
@@ -478,8 +477,17 @@ go someFunc(user)
 但是，`ast.Visitor` 并不是一个很易用的 API。你可能需要写大量的代码去实现自己的 `ast.Visitor`，并且需要在解析过程中仔细维护树的层级。如果你决定要使用这个 API，请先仔细阅读 [ast/visitor.go](https://github.com/bytedance/sonic/blob/main/ast/visitor.go) 中的注释。
 
 ### 缓冲区大小
+
 Sonic在许多地方使用内存池，如`encoder.Encode`, `ast.Node.MarshalJSON`等来提高性能，这可能会在服务器负载高时产生更多的内存使用(in-use)。参见[issue 614](https://github.com/bytedance/sonic/issues/614)。因此，我们引入了一些选项来让用户配置内存池的行为。参见[option](https://pkg.go.dev/github.com/bytedance/[email protected]/option#pkg-variables)包。
 
+### 更快的 JSON Skip
+
+为了安全起见，在跳过原始JSON 时，sonic decoder 默认使用[FSM](native/skip_one.c)算法扫描来跳过同时校验 JSON。它相比[SIMD-searching-pair](native/skip_one_fast.c)算法跳过要慢得多(1~10倍)。如果用户有很多冗余的JSON值，并且不需要严格验证JSON的正确性，你可以启用以下选项:
+
+- `Config.NoValidateSkipJSON`: 用于在解码时更快地跳过JSON，例如未知字段，`json.RawMessage`，不匹配的值和冗余的数组元素等
+- `Config.NoValidateJSONMarshaler`: 编码JSON时避免验证JSON。封送拆收器
+- `SearchOption.ValidateJSON`: 指示当`Get`时是否验证定位的JSON值
+
 ## 社区
 
 Sonic 是 [CloudWeGo](https://www.cloudwego.io/) 下的一个子项目。我们致力于构建云原生生态系统。

api.go

@@ -84,6 +84,10 @@ type Config struct {
     // NoValidateJSONMarshaler indicates that the encoder should not validate the output string
     // after encoding the JSONMarshaler to JSON.
     NoValidateJSONMarshaler       bool
+
+    // NoValidateJSONSkip indicates the decoder should not validate the JSON value when skipping it,
+    // such as unknown-fields, mismatched-type, redundant elements..
+    NoValidateJSONSkip bool
 
     // NoEncoderNewline indicates that the encoder should not add a newline after every message
     NoEncoderNewline bool
@@ -109,6 +113,7 @@ var (
     ConfigFastest = Config{
         NoQuoteTextMarshaler: true,
         NoValidateJSONMarshaler: true,
+        NoValidateJSONSkip: true,
     }.Froze()
 )
 

decoder/decoder_compat.go

@@ -42,6 +42,7 @@ const (
      _F_use_number      = types.B_USE_NUMBER
      _F_validate_string = types.B_VALIDATE_STRING
      _F_allow_control   = types.B_ALLOW_CONTROL
+     _F_no_validate_json   = types.B_NO_VALIDATE_JSON
 )
 
 type Options uint64
@@ -53,6 +54,7 @@ const (
      OptionDisableUnknown   Options = 1 << _F_disable_unknown
      OptionCopyString       Options = 1 << _F_copy_string
      OptionValidateString   Options = 1 << _F_validate_string
+     OptionNoValidateJSON   Options = 1 << _F_no_validate_json
 )
 
 func (self *Decoder) SetOptions(opts Options) {

decoder/decoder_native.go

@@ -43,6 +43,7 @@ const (
     OptionDisableUnknown   Options = api.OptionDisableUnknown
     OptionCopyString       Options = api.OptionCopyString
     OptionValidateString   Options = api.OptionValidateString
+    OptionNoValidateJSON   Options = api.OptionNoValidateJSON
 )
 
 // StreamDecoder is the decoder context object for streaming input.

decoder/decoder_native_test.go

@@ -1,7 +1,6 @@
 //go:build (amd64 && go1.17 && !go1.24) || (arm64 && go1.20 && !go1.24)
 // +build amd64,go1.17,!go1.24 arm64,go1.20,!go1.24
 
-
 /*
  * Copyright 2021 ByteDance Inc.
  *
@@ -21,15 +20,90 @@
 package decoder
 
 import (
-    `encoding/json`
-    _`strings`
-    `testing`
-    _`reflect`
+	"encoding/json"
+	"fmt"
+	_ "reflect"
+	"strings"
+	_ "strings"
+	"testing"
+	"time"
 
-    `github.com/bytedance/sonic/internal/rt`
-    `github.com/stretchr/testify/assert`
+	"github.com/bytedance/sonic/internal/rt"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
+
+func BenchmarkSkipValidate(b *testing.B) {
+    type skiptype struct {
+        A int `json:"a"` // mismatched
+        B string `json:"-"` // ommited
+        C [1]int `json:"c"` // fast int
+        D struct {} `json:"d"` // empty struct
+        E map[string]int `json:"e"` // mismatched elem
+        F json.RawMessage `json:"f"` // unmarshaler
+        // Unknonwn
+    }
+    type C struct {
+        name string
+        json string
+        expTime float64
+    }
+    var sam = map[int]interface{}{}
+    for i := 0; i < 1; i++ {
+        sam[i] = _BindingValue
+    }
+    comptd, err := json.Marshal(sam)
+    if err != nil {
+        b.Fatal("invalid json")
+    }
+    compt := string(comptd)
+    var cases = []C{
+        {"mismatched", `{"a":`+compt+`}`, 5},
+        {"ommited", `{"b":`+compt+`}`, 5},
+        {"number", `{"c":[`+strings.Repeat("-1.23456e-19,", 1000)+`1]}`, 1.5},
+        {"unknown", `{"unknown":`+compt+`}`, 5},
+        {"empty", `{"d":`+compt+`}`, 5},
+        {"mismatched elem", `{"e":`+compt+`}`, 5},
+        {"unmarshaler", `{"f":`+compt+`}`, 3},
+    }
+    _ = NewDecoder(`{}`).Decode(&skiptype{})
+
+    var avg1, avg2 time.Duration
+    for _, c := range cases {
+        b.Run(c.name, func(b *testing.B) {
+            b.Run("validate", func(b *testing.B) {
+                b.ResetTimer()
+                t1 := time.Now()
+                for i := 0; i < b.N; i++ {
+                    var obj1 = &skiptype{}
+                    // validate skip
+                    d := NewDecoder(c.json)
+                    _ = d.Decode(obj1)
+                }
+                d1 := time.Since(t1)
+                avg1 = d1/time.Duration(b.N)
+            })
+            b.Run("fast", func(b *testing.B) {
+                b.ResetTimer()
+                t2 := time.Now()
+                for i := 0; i < b.N; i++ {
+                    var obj2 = &skiptype{}
+                    // fask skip
+                    d := NewDecoder(c.json)
+                    d.SetOptions(OptionNoValidateJSON)
+                    _ = d.Decode(obj2)
+                }
+                d2 := time.Since(t2)
+                avg2 = d2/time.Duration(b.N)
+            })
+            // fast skip must be expTime x faster
+            require.True(b, float64(avg1)/float64(avg2) >  c.expTime, fmt.Sprintf("%v/%v=%v", avg1, avg2, float64(avg1)/float64(avg2)))
+        })
+    }
+}
+
+
 func TestSkipMismatchTypeAmd64Error(t *testing.T) {
     // t.Run("struct", func(t *testing.T) {
     //     println("TestSkipError")

decoder/decoder_test.go

@@ -17,16 +17,16 @@
 package decoder
 
 import (
-    `encoding/json`
-    `runtime`
-    `runtime/debug`
-    `strings`
-    `sync`
-    `testing`
-    `time`
-
-    `github.com/stretchr/testify/assert`
-    `github.com/stretchr/testify/require`
+	"encoding/json"
+	"runtime"
+	"runtime/debug"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestMain(m *testing.M) {
@@ -85,7 +85,6 @@ func init() {
     _ = json.Unmarshal([]byte(TwitterJson), &_BindingValue)
 }
 
-
 func TestSkipMismatchTypeError(t *testing.T) {
     t.Run("struct", func(t *testing.T) {
         println("TestSkipError")

internal/decoder/api/decoder.go

@@ -44,6 +44,7 @@ const (
     OptionDisableUnknown   = consts.OptionDisableUnknown
     OptionCopyString       = consts.OptionCopyString
     OptionValidateString   = consts.OptionValidateString
+    OptionNoValidateJSON   = consts.OptionNoValidateJSON
 )
 
 type (

internal/decoder/consts/option.go

@@ -12,9 +12,11 @@ const (
     F_disable_unknown = 3
     F_copy_string     = 4
 
+
     F_use_number      = types.B_USE_NUMBER
     F_validate_string = types.B_VALIDATE_STRING
     F_allow_control   = types.B_ALLOW_CONTROL
+    F_no_validate_json = types.B_NO_VALIDATE_JSON
 )
 
 type Options uint64
@@ -26,6 +28,7 @@ const (
     OptionDisableUnknown   Options = 1 << F_disable_unknown
     OptionCopyString       Options = 1 << F_copy_string
     OptionValidateString   Options = 1 << F_validate_string
+    OptionNoValidateJSON   Options = 1 << F_no_validate_json
 )
 
 const (