Add internal/vm package

This commit is contained in:
Masaaki Goshima 2021-03-16 02:50:19 +09:00
parent 2385cfcdbf
commit c45f1e8b2c
13 changed files with 6060 additions and 71 deletions

View File

@ -26,3 +26,7 @@ golangci-lint: | $(BIN_DIR)
GOBIN=$(BIN_DIR) go get github.com/golangci/golangci-lint/cmd/golangci-lint@v1.36.0; \ GOBIN=$(BIN_DIR) go get github.com/golangci/golangci-lint/cmd/golangci-lint@v1.36.0; \
rm -rf $$GOLANGCI_LINT_TMP_DIR; \ rm -rf $$GOLANGCI_LINT_TMP_DIR; \
} }
.PHONY: generate
generate:
go generate ./internal/...

View File

@ -12,6 +12,8 @@ import (
"strings" "strings"
"sync" "sync"
"unsafe" "unsafe"
_ "github.com/goccy/go-json/internal/encoder/vm"
) )
// An Encoder writes JSON values to an output stream. // An Encoder writes JSON values to an output stream.

View File

@ -66,18 +66,18 @@ func createOpType(op, code string) opType {
} }
func _main() error { func _main() error {
tmpl, err := template.New("").Parse(`// Code generated by cmd/generator. DO NOT EDIT! tmpl, err := template.New("").Parse(`// Code generated by internal/cmd/generator. DO NOT EDIT!
package json package encoder
import ( import (
"strings" "strings"
) )
type codeType int type CodeType int
const ( const (
{{- range $index, $type := .CodeTypes }} {{- range $index, $type := .CodeTypes }}
code{{ $type }} codeType = {{ $index }} Code{{ $type }} CodeType = {{ $index }}
{{- end }} {{- end }}
) )
@ -87,92 +87,86 @@ var opTypeStrings = [{{ .OpLen }}]string{
{{- end }} {{- end }}
} }
type opType int type OpType int
const ( const (
{{- range $index, $type := .OpTypes }} {{- range $index, $type := .OpTypes }}
op{{ $type.Op }} opType = {{ $index }} Op{{ $type.Op }} OpType = {{ $index }}
{{- end }} {{- end }}
) )
func (t opType) String() string { func (t OpType) String() string {
if int(t) >= {{ .OpLen }} { if int(t) >= {{ .OpLen }} {
return "" return ""
} }
return opTypeStrings[int(t)] return opTypeStrings[int(t)]
} }
func (t opType) codeType() codeType { func (t OpType) CodeType() CodeType {
if strings.Contains(t.String(), "Struct") { if strings.Contains(t.String(), "Struct") {
if strings.Contains(t.String(), "End") { if strings.Contains(t.String(), "End") {
return codeStructEnd return CodeStructEnd
} }
return codeStructField return CodeStructField
} }
if t.String() == "Array" || t.String() == "ArrayPtr" { switch t {
return codeArrayHead case OpArray, OpArrayPtr:
} return CodeArrayHead
if strings.Contains(t.String(), "ArrayElem") { case OpArrayElem:
return codeArrayElem return CodeArrayElem
} case OpSlice, OpSlicePtr:
if t.String() == "Slice" || t.String() == "SlicePtr" { return CodeSliceHead
return codeSliceHead case OpSliceElem:
} return CodeSliceElem
if strings.Contains(t.String(), "SliceElem") { case OpMap, OpMapPtr:
return codeSliceElem return CodeMapHead
} case OpMapKey:
if t.String() == "Map" || t.String() == "MapPtr" { return CodeMapKey
return codeMapHead case OpMapValue:
} return CodeMapValue
if strings.Contains(t.String(), "MapKey") { case OpMapEnd:
return codeMapKey return CodeMapEnd
}
if strings.Contains(t.String(), "MapValue") {
return codeMapValue
}
if strings.Contains(t.String(), "MapEnd") {
return codeMapEnd
} }
return codeOp return CodeOp
} }
func (t opType) headToPtrHead() opType { func (t OpType) HeadToPtrHead() OpType {
if strings.Index(t.String(), "PtrHead") > 0 { if strings.Index(t.String(), "PtrHead") > 0 {
return t return t
} }
idx := strings.Index(t.String(), "Field") idx := strings.Index(t.String(), "Head")
if idx == -1 { if idx == -1 {
return t return t
} }
suffix := "Ptr"+t.String()[idx+len("Field"):] suffix := "Ptr"+t.String()[idx+len("Head"):]
const toPtrOffset = 3 const toPtrOffset = 3
if strings.Contains(opType(int(t) + toPtrOffset).String(), suffix) { if strings.Contains(OpType(int(t) + toPtrOffset).String(), suffix) {
return opType(int(t) + toPtrOffset) return OpType(int(t) + toPtrOffset)
} }
return t return t
} }
func (t opType) headToOmitEmptyHead() opType { func (t OpType) HeadToOmitEmptyHead() OpType {
const toOmitEmptyOffset = 1 const toOmitEmptyOffset = 1
if strings.Contains(opType(int(t) + toOmitEmptyOffset).String(), "OmitEmpty") { if strings.Contains(OpType(int(t) + toOmitEmptyOffset).String(), "OmitEmpty") {
return opType(int(t) + toOmitEmptyOffset) return OpType(int(t) + toOmitEmptyOffset)
} }
return t return t
} }
func (t opType) headToStringTagHead() opType { func (t OpType) HeadToStringTagHead() OpType {
const toStringTagOffset = 2 const toStringTagOffset = 2
if strings.Contains(opType(int(t) + toStringTagOffset).String(), "StringTag") { if strings.Contains(OpType(int(t) + toStringTagOffset).String(), "StringTag") {
return opType(int(t) + toStringTagOffset) return OpType(int(t) + toStringTagOffset)
} }
return t return t
} }
func (t opType) ptrHeadToHead() opType { func (t OpType) PtrHeadToHead() OpType {
idx := strings.Index(t.String(), "Ptr") idx := strings.Index(t.String(), "Ptr")
if idx == -1 { if idx == -1 {
return t return t
@ -180,36 +174,37 @@ func (t opType) ptrHeadToHead() opType {
suffix := t.String()[idx+len("Ptr"):] suffix := t.String()[idx+len("Ptr"):]
const toPtrOffset = 3 const toPtrOffset = 3
if strings.Contains(opType(int(t) - toPtrOffset).String(), suffix) { if strings.Contains(OpType(int(t) - toPtrOffset).String(), suffix) {
return opType(int(t) - toPtrOffset) return OpType(int(t) - toPtrOffset)
} }
return t return t
} }
func (t opType) fieldToEnd() opType { func (t OpType) FieldToEnd() OpType {
switch t { idx := strings.Index(t.String(), "Field")
{{- range $type := .OpTypes }} if idx == -1 {
{{- if $type.IsFieldToEnd }} return t
case op{{ $type.Op }}: }
return op{{ call $type.FieldToEnd }} suffix := t.String()[idx+len("Field"):]
{{- end }} const toEndOffset = 3
{{- end }} if strings.Contains(OpType(int(t) + toEndOffset).String(), "End"+suffix) {
return OpType(int(t) + toEndOffset)
} }
return t return t
} }
func (t opType) fieldToOmitEmptyField() opType { func (t OpType) FieldToOmitEmptyField() OpType {
const toOmitEmptyOffset = 1 const toOmitEmptyOffset = 1
if strings.Contains(opType(int(t) + toOmitEmptyOffset).String(), "OmitEmpty") { if strings.Contains(OpType(int(t) + toOmitEmptyOffset).String(), "OmitEmpty") {
return opType(int(t) + toOmitEmptyOffset) return OpType(int(t) + toOmitEmptyOffset)
} }
return t return t
} }
func (t opType) fieldToStringTagField() opType { func (t OpType) FieldToStringTagField() OpType {
const toStringTagOffset = 2 const toStringTagOffset = 2
if strings.Contains(opType(int(t) + toStringTagOffset).String(), "StringTag") { if strings.Contains(OpType(int(t) + toStringTagOffset).String(), "StringTag") {
return opType(int(t) + toStringTagOffset) return OpType(int(t) + toStringTagOffset)
} }
return t return t
} }
@ -269,7 +264,7 @@ func (t opType) fieldToStringTagField() opType {
typ := typ typ := typ
op := fmt.Sprintf( op := fmt.Sprintf(
"StructField%sHead%s%s", "Struct%sHead%s%s",
ptrOrNot, ptrOrNot,
opt, opt,
typ, typ,
@ -279,28 +274,28 @@ func (t opType) fieldToStringTagField() opType {
Code: "StructField", Code: "StructField",
HeadToPtrHead: func() string { HeadToPtrHead: func() string {
return fmt.Sprintf( return fmt.Sprintf(
"StructFieldPtrHead%s%s", "StructPtrHead%s%s",
opt, opt,
typ, typ,
) )
}, },
HeadToOmitEmptyHead: func() string { HeadToOmitEmptyHead: func() string {
return fmt.Sprintf( return fmt.Sprintf(
"StructField%sHeadOmitEmpty%s", "Struct%sHeadOmitEmpty%s",
ptrOrNot, ptrOrNot,
typ, typ,
) )
}, },
HeadToStringTagHead: func() string { HeadToStringTagHead: func() string {
return fmt.Sprintf( return fmt.Sprintf(
"StructField%sHeadStringTag%s", "Struct%sHeadStringTag%s",
ptrOrNot, ptrOrNot,
typ, typ,
) )
}, },
PtrHeadToHead: func() string { PtrHeadToHead: func() string {
return fmt.Sprintf( return fmt.Sprintf(
"StructFieldHead%s%s", "StructHead%s%s",
opt, opt,
typ, typ,
) )
@ -354,8 +349,6 @@ func (t opType) fieldToStringTagField() opType {
}, },
}) })
} }
}
for _, typ := range append(primitiveTypesUpper, "") {
for _, opt := range []string{"", "OmitEmpty", "StringTag"} { for _, opt := range []string{"", "OmitEmpty", "StringTag"} {
opt := opt opt := opt
typ := typ typ := typ
@ -390,7 +383,7 @@ func (t opType) fieldToStringTagField() opType {
}); err != nil { }); err != nil {
return err return err
} }
path := filepath.Join(repoRoot(), "encode_optype.go") path := filepath.Join(repoRoot(), "internal", "encoder", "optype.go")
buf, err := format.Source(b.Bytes()) buf, err := format.Source(b.Bytes())
if err != nil { if err != nil {
return err return err
@ -400,10 +393,11 @@ func (t opType) fieldToStringTagField() opType {
func repoRoot() string { func repoRoot() string {
_, file, _, _ := runtime.Caller(0) _, file, _, _ := runtime.Caller(0)
relativePathFromRepoRoot := filepath.Join("cmd", "generator") relativePathFromRepoRoot := filepath.Join("internal", "cmd", "generator")
return strings.TrimSuffix(filepath.Dir(file), relativePathFromRepoRoot) return strings.TrimSuffix(filepath.Dir(file), relativePathFromRepoRoot)
} }
//go:generate go run main.go
func main() { func main() {
if err := _main(); err != nil { if err := _main(); err != nil {
panic(err) panic(err)

View File

@ -0,0 +1 @@
package compiler

View File

@ -0,0 +1,11 @@
// +build !race
package compiler
import (
"github.com/goccy/go-json/internal/encoder"
)
func CompileToGetCodeSet(typeptr uintptr) (*encoder.OpcodeSet, error) {
return nil, nil
}

407
internal/encoder/encoder.go Normal file
View File

@ -0,0 +1,407 @@
package encoder
import (
"bytes"
"encoding"
"encoding/base64"
"encoding/json"
"fmt"
"math"
"reflect"
"strconv"
"sync"
"unsafe"
"github.com/goccy/go-json/internal/errors"
"github.com/goccy/go-json/internal/runtime"
)
type Option int
const (
HTMLEscapeOption Option = 1 << iota
IndentOption
UnorderedMapOption
)
type Opcode struct {
Op OpType // operation type
Type *runtime.Type // go type
DisplayIdx int // opcode index
Key []byte // struct field key
EscapedKey []byte // struct field key ( HTML escaped )
PtrNum int // pointer number: e.g. double pointer is 2.
DisplayKey string // key text to display
IsTaggedKey bool // whether tagged key
AnonymousKey bool // whether anonymous key
AnonymousHead bool // whether anonymous head or not
Indirect bool // whether indirect or not
Nilcheck bool // whether needs to nilcheck or not
AddrForMarshaler bool // whether needs to addr for marshaler or not
RshiftNum uint8 // use to take bit for judging whether negative integer or not
Mask uint64 // mask for number
Indent int // indent number
Idx uintptr // offset to access ptr
HeadIdx uintptr // offset to access slice/struct head
ElemIdx uintptr // offset to access array/slice/map elem
Length uintptr // offset to access slice/map length or array length
MapIter uintptr // offset to access map iterator
MapPos uintptr // offset to access position list for sorted map
Offset uintptr // offset size from struct header
Size uintptr // array/slice elem size
MapKey *Opcode // map key
MapValue *Opcode // map value
Elem *Opcode // array/slice elem
End *Opcode // array/slice/struct/map end
PrevField *Opcode // prev struct field
NextField *Opcode // next struct field
Next *Opcode // next opcode
Jmp *CompiledCode // for recursive call
}
type OpcodeSet struct {
Code *Opcode
CodeLength int
}
type CompiledCode struct {
Code *Opcode
Linked bool // whether recursive code already have linked
CurLen uintptr
NextLen uintptr
}
const StartDetectingCyclesAfter = 1000
func Load(base uintptr, idx uintptr) uintptr {
addr := base + idx
return **(**uintptr)(unsafe.Pointer(&addr))
}
func Store(base uintptr, idx uintptr, p uintptr) {
addr := base + idx
**(**uintptr)(unsafe.Pointer(&addr)) = p
}
func LoadAndStoreNPtr(base uintptr, idx uintptr, ptrNum int) {
addr := base + idx
p := **(**uintptr)(unsafe.Pointer(&addr))
for i := 0; i < ptrNum; i++ {
if p == 0 {
**(**uintptr)(unsafe.Pointer(&addr)) = 0
return
}
p = PtrToPtr(p)
}
**(**uintptr)(unsafe.Pointer(&addr)) = p
}
func PtrToUint64(p uintptr) uint64 { return **(**uint64)(unsafe.Pointer(&p)) }
func PtrToFloat32(p uintptr) float32 { return **(**float32)(unsafe.Pointer(&p)) }
func PtrToFloat64(p uintptr) float64 { return **(**float64)(unsafe.Pointer(&p)) }
func PtrToBool(p uintptr) bool { return **(**bool)(unsafe.Pointer(&p)) }
func PtrToBytes(p uintptr) []byte { return **(**[]byte)(unsafe.Pointer(&p)) }
func PtrToNumber(p uintptr) json.Number { return **(**json.Number)(unsafe.Pointer(&p)) }
func PtrToString(p uintptr) string { return **(**string)(unsafe.Pointer(&p)) }
func PtrToSlice(p uintptr) *runtime.SliceHeader { return *(**runtime.SliceHeader)(unsafe.Pointer(&p)) }
func PtrToPtr(p uintptr) uintptr {
return uintptr(**(**unsafe.Pointer)(unsafe.Pointer(&p)))
}
func PtrToNPtr(p uintptr, ptrNum int) uintptr {
for i := 0; i < ptrNum; i++ {
if p == 0 {
return 0
}
p = PtrToPtr(p)
}
return p
}
func PtrToUnsafePtr(p uintptr) unsafe.Pointer {
return *(*unsafe.Pointer)(unsafe.Pointer(&p))
}
func PtrToInterface(code *Opcode, p uintptr) interface{} {
return *(*interface{})(unsafe.Pointer(&emptyInterface{
typ: code.Type,
ptr: *(*unsafe.Pointer)(unsafe.Pointer(&p)),
}))
}
func ErrUnsupportedValue(code *Opcode, ptr uintptr) *errors.UnsupportedValueError {
v := *(*interface{})(unsafe.Pointer(&emptyInterface{
typ: code.Type,
ptr: *(*unsafe.Pointer)(unsafe.Pointer(&ptr)),
}))
return &errors.UnsupportedValueError{
Value: reflect.ValueOf(v),
Str: fmt.Sprintf("encountered a cycle via %s", code.Type),
}
}
func ErrUnsupportedFloat(v float64) *errors.UnsupportedValueError {
return &errors.UnsupportedValueError{
Value: reflect.ValueOf(v),
Str: strconv.FormatFloat(v, 'g', -1, 64),
}
}
func ErrMarshalerWithCode(code *Opcode, err error) *errors.MarshalerError {
return &errors.MarshalerError{
Type: runtime.RType2Type(code.Type),
Err: err,
}
}
type emptyInterface struct {
typ *runtime.Type
ptr unsafe.Pointer
}
type MapItem struct {
Key []byte
Value []byte
}
type Mapslice struct {
Items []MapItem
}
func (m *Mapslice) Len() int {
return len(m.Items)
}
func (m *Mapslice) Less(i, j int) bool {
return bytes.Compare(m.Items[i].Key, m.Items[j].Key) < 0
}
func (m *Mapslice) Swap(i, j int) {
m.Items[i], m.Items[j] = m.Items[j], m.Items[i]
}
type MapContext struct {
Pos []int
Slice *Mapslice
Buf []byte
}
var mapContextPool = sync.Pool{
New: func() interface{} {
return &MapContext{}
},
}
func NewMapContext(mapLen int) *MapContext {
ctx := mapContextPool.Get().(*MapContext)
if ctx.Slice == nil {
ctx.Slice = &Mapslice{
Items: make([]MapItem, 0, mapLen),
}
}
if cap(ctx.Pos) < (mapLen*2 + 1) {
ctx.Pos = make([]int, 0, mapLen*2+1)
ctx.Slice.Items = make([]MapItem, 0, mapLen)
} else {
ctx.Pos = ctx.Pos[:0]
ctx.Slice.Items = ctx.Slice.Items[:0]
}
ctx.Buf = ctx.Buf[:0]
return ctx
}
func ReleaseMapContext(c *MapContext) {
mapContextPool.Put(c)
}
//go:linkname MapIterInit reflect.mapiterinit
//go:noescape
func MapIterInit(mapType *runtime.Type, m unsafe.Pointer) unsafe.Pointer
//go:linkname MapIterKey reflect.mapiterkey
//go:noescape
func MapIterKey(it unsafe.Pointer) unsafe.Pointer
//go:linkname MapIterNext reflect.mapiternext
//go:noescape
func MapIterNext(it unsafe.Pointer)
//go:linkname MapLen reflect.maplen
//go:noescape
func MapLen(m unsafe.Pointer) int
type RuntimeContext struct {
Buf []byte
Ptrs []uintptr
KeepRefs []unsafe.Pointer
SeenPtr []uintptr
BaseIndent int
Prefix []byte
IndentStr []byte
}
func (c *RuntimeContext) Init(p uintptr, codelen int) {
if len(c.Ptrs) < codelen {
c.Ptrs = make([]uintptr, codelen)
}
c.Ptrs[0] = p
c.KeepRefs = c.KeepRefs[:0]
c.SeenPtr = c.SeenPtr[:0]
c.BaseIndent = 0
}
func (c *RuntimeContext) Ptr() uintptr {
header := (*runtime.SliceHeader)(unsafe.Pointer(&c.Ptrs))
return uintptr(header.Data)
}
func AppendByteSlice(b []byte, src []byte) []byte {
if src == nil {
return append(b, `null`...)
}
encodedLen := base64.StdEncoding.EncodedLen(len(src))
b = append(b, '"')
pos := len(b)
remainLen := cap(b[pos:])
var buf []byte
if remainLen > encodedLen {
buf = b[pos : pos+encodedLen]
} else {
buf = make([]byte, encodedLen)
}
base64.StdEncoding.Encode(buf, src)
return append(append(b, buf...), '"')
}
func AppendFloat32(b []byte, v float32) []byte {
f64 := float64(v)
abs := math.Abs(f64)
fmt := byte('f')
// Note: Must use float32 comparisons for underlying float32 value to get precise cutoffs right.
if abs != 0 {
f32 := float32(abs)
if f32 < 1e-6 || f32 >= 1e21 {
fmt = 'e'
}
}
return strconv.AppendFloat(b, f64, fmt, -1, 32)
}
func AppendFloat64(b []byte, v float64) []byte {
abs := math.Abs(v)
fmt := byte('f')
// Note: Must use float32 comparisons for underlying float32 value to get precise cutoffs right.
if abs != 0 {
if abs < 1e-6 || abs >= 1e21 {
fmt = 'e'
}
}
return strconv.AppendFloat(b, v, fmt, -1, 64)
}
func AppendBool(b []byte, v bool) []byte {
if v {
return append(b, "true"...)
}
return append(b, "false"...)
}
var (
floatTable = [256]bool{
'0': true,
'1': true,
'2': true,
'3': true,
'4': true,
'5': true,
'6': true,
'7': true,
'8': true,
'9': true,
'.': true,
'e': true,
'E': true,
'+': true,
'-': true,
}
)
func AppendNumber(b []byte, n json.Number) ([]byte, error) {
if len(n) == 0 {
return append(b, '0'), nil
}
for i := 0; i < len(n); i++ {
if !floatTable[n[i]] {
return nil, fmt.Errorf("json: invalid number literal %q", n)
}
}
b = append(b, n...)
return b, nil
}
func AppendMarshalJSON(code *Opcode, b []byte, v interface{}, escape bool) ([]byte, error) {
rv := reflect.ValueOf(v) // convert by dynamic interface type
if code.AddrForMarshaler {
if rv.CanAddr() {
rv = rv.Addr()
} else {
newV := reflect.New(rv.Type())
newV.Elem().Set(rv)
rv = newV
}
}
v = rv.Interface()
marshaler, ok := v.(json.Marshaler)
if !ok {
return AppendNull(b), nil
}
bb, err := marshaler.MarshalJSON()
if err != nil {
return nil, &errors.MarshalerError{Type: reflect.TypeOf(v), Err: err}
}
return bb, nil
//buf := bytes.NewBuffer(b)
//TODO: we should validate buffer with `compact`
// if err := compact(buf, bb, escape); err != nil {
// return nil, &errors.MarshalerError{Type: reflect.TypeOf(v), Err: err}
// }
//return buf.Bytes(), nil
}
func AppendMarshalText(code *Opcode, b []byte, v interface{}, escape bool) ([]byte, error) {
rv := reflect.ValueOf(v) // convert by dynamic interface type
if code.AddrForMarshaler {
if rv.CanAddr() {
rv = rv.Addr()
} else {
newV := reflect.New(rv.Type())
newV.Elem().Set(rv)
rv = newV
}
}
v = rv.Interface()
marshaler, ok := v.(encoding.TextMarshaler)
if !ok {
return AppendNull(b), nil
}
bytes, err := marshaler.MarshalText()
if err != nil {
return nil, &errors.MarshalerError{Type: reflect.TypeOf(v), Err: err}
}
if escape {
return AppendEscapedString(b, *(*string)(unsafe.Pointer(&bytes))), nil
}
return AppendString(b, *(*string)(unsafe.Pointer(&bytes))), nil
}
func AppendNull(b []byte) []byte {
return append(b, "null"...)
}
func AppendComma(b []byte) []byte {
return append(b, ',')
}
func AppendStructEnd(b []byte) []byte {
return append(b, '}', ',')
}

124
internal/encoder/int.go Normal file
View File

@ -0,0 +1,124 @@
package encoder
import (
"unsafe"
)
var endianness int
func init() {
var b [2]byte
*(*uint16)(unsafe.Pointer(&b)) = uint16(0xABCD)
switch b[0] {
case 0xCD:
endianness = 0 // LE
case 0xAB:
endianness = 1 // BE
default:
panic("could not determine endianness")
}
}
// "00010203...96979899" cast to []uint16
var intLELookup = [100]uint16{
0x3030, 0x3130, 0x3230, 0x3330, 0x3430, 0x3530, 0x3630, 0x3730, 0x3830, 0x3930,
0x3031, 0x3131, 0x3231, 0x3331, 0x3431, 0x3531, 0x3631, 0x3731, 0x3831, 0x3931,
0x3032, 0x3132, 0x3232, 0x3332, 0x3432, 0x3532, 0x3632, 0x3732, 0x3832, 0x3932,
0x3033, 0x3133, 0x3233, 0x3333, 0x3433, 0x3533, 0x3633, 0x3733, 0x3833, 0x3933,
0x3034, 0x3134, 0x3234, 0x3334, 0x3434, 0x3534, 0x3634, 0x3734, 0x3834, 0x3934,
0x3035, 0x3135, 0x3235, 0x3335, 0x3435, 0x3535, 0x3635, 0x3735, 0x3835, 0x3935,
0x3036, 0x3136, 0x3236, 0x3336, 0x3436, 0x3536, 0x3636, 0x3736, 0x3836, 0x3936,
0x3037, 0x3137, 0x3237, 0x3337, 0x3437, 0x3537, 0x3637, 0x3737, 0x3837, 0x3937,
0x3038, 0x3138, 0x3238, 0x3338, 0x3438, 0x3538, 0x3638, 0x3738, 0x3838, 0x3938,
0x3039, 0x3139, 0x3239, 0x3339, 0x3439, 0x3539, 0x3639, 0x3739, 0x3839, 0x3939,
}
var intBELookup = [100]uint16{
0x3030, 0x3031, 0x3032, 0x3033, 0x3034, 0x3035, 0x3036, 0x3037, 0x3038, 0x3039,
0x3130, 0x3131, 0x3132, 0x3133, 0x3134, 0x3135, 0x3136, 0x3137, 0x3138, 0x3139,
0x3230, 0x3231, 0x3232, 0x3233, 0x3234, 0x3235, 0x3236, 0x3237, 0x3238, 0x3239,
0x3330, 0x3331, 0x3332, 0x3333, 0x3334, 0x3335, 0x3336, 0x3337, 0x3338, 0x3339,
0x3430, 0x3431, 0x3432, 0x3433, 0x3434, 0x3435, 0x3436, 0x3437, 0x3438, 0x3439,
0x3530, 0x3531, 0x3532, 0x3533, 0x3534, 0x3535, 0x3536, 0x3537, 0x3538, 0x3539,
0x3630, 0x3631, 0x3632, 0x3633, 0x3634, 0x3635, 0x3636, 0x3637, 0x3638, 0x3639,
0x3730, 0x3731, 0x3732, 0x3733, 0x3734, 0x3735, 0x3736, 0x3737, 0x3738, 0x3739,
0x3830, 0x3831, 0x3832, 0x3833, 0x3834, 0x3835, 0x3836, 0x3837, 0x3838, 0x3839,
0x3930, 0x3931, 0x3932, 0x3933, 0x3934, 0x3935, 0x3936, 0x3937, 0x3938, 0x3939,
}
var intLookup = [2]*[100]uint16{&intLELookup, &intBELookup}
func AppendInt(out []byte, u64 uint64, code *Opcode) []byte {
n := u64 & code.Mask
negative := (u64>>code.RshiftNum)&1 == 1
if !negative {
if n < 10 {
return append(out, byte(n+'0'))
} else if n < 100 {
u := intLELookup[n]
return append(out, byte(u), byte(u>>8))
}
} else {
n = -n & code.Mask
}
lookup := intLookup[endianness]
var b [22]byte
u := (*[11]uint16)(unsafe.Pointer(&b))
i := 11
for n >= 100 {
j := n % 100
n /= 100
i--
u[i] = lookup[j]
}
i--
u[i] = lookup[n]
i *= 2 // convert to byte index
if n < 10 {
i++ // remove leading zero
}
if negative {
i--
b[i] = '-'
}
return append(out, b[i:]...)
}
func AppendUint(out []byte, u64 uint64, code *Opcode) []byte {
n := u64 & code.Mask
if n < 10 {
return append(out, byte(n+'0'))
} else if n < 100 {
u := intLELookup[n]
return append(out, byte(u), byte(u>>8))
}
lookup := intLookup[endianness]
var b [22]byte
u := (*[11]uint16)(unsafe.Pointer(&b))
i := 11
for n >= 100 {
j := n % 100
n /= 100
i--
u[i] = lookup[j]
}
i--
u[i] = lookup[n]
i *= 2 // convert to byte index
if n < 10 {
i++ // remove leading zero
}
return append(out, b[i:]...)
}

View File

@ -0,0 +1,8 @@
// +build !go1.13
package encoder
import "unsafe"
//go:linkname MapIterValue reflect.mapitervalue
func MapIterValue(it unsafe.Pointer) unsafe.Pointer

View File

@ -0,0 +1,8 @@
// +build go1.13
package encoder
import "unsafe"
//go:linkname MapIterValue reflect.mapiterelem
func MapIterValue(it unsafe.Pointer) unsafe.Pointer

1025
internal/encoder/optype.go Normal file

File diff suppressed because it is too large Load Diff

637
internal/encoder/string.go Normal file
View File

@ -0,0 +1,637 @@
package encoder
import (
"math/bits"
"reflect"
"unicode/utf8"
"unsafe"
)
const (
lsb = 0x0101010101010101
msb = 0x8080808080808080
)
var needEscapeWithHTML = [256]bool{
'"': true,
'&': true,
'<': true,
'>': true,
'\\': true,
0x00: true,
0x01: true,
0x02: true,
0x03: true,
0x04: true,
0x05: true,
0x06: true,
0x07: true,
0x08: true,
0x09: true,
0x0a: true,
0x0b: true,
0x0c: true,
0x0d: true,
0x0e: true,
0x0f: true,
0x10: true,
0x11: true,
0x12: true,
0x13: true,
0x14: true,
0x15: true,
0x16: true,
0x17: true,
0x18: true,
0x19: true,
0x1a: true,
0x1b: true,
0x1c: true,
0x1d: true,
0x1e: true,
0x1f: true,
/* 0x20 - 0x7f */
0x80: true,
0x81: true,
0x82: true,
0x83: true,
0x84: true,
0x85: true,
0x86: true,
0x87: true,
0x88: true,
0x89: true,
0x8a: true,
0x8b: true,
0x8c: true,
0x8d: true,
0x8e: true,
0x8f: true,
0x90: true,
0x91: true,
0x92: true,
0x93: true,
0x94: true,
0x95: true,
0x96: true,
0x97: true,
0x98: true,
0x99: true,
0x9a: true,
0x9b: true,
0x9c: true,
0x9d: true,
0x9e: true,
0x9f: true,
0xa0: true,
0xa1: true,
0xa2: true,
0xa3: true,
0xa4: true,
0xa5: true,
0xa6: true,
0xa7: true,
0xa8: true,
0xa9: true,
0xaa: true,
0xab: true,
0xac: true,
0xad: true,
0xae: true,
0xaf: true,
0xb0: true,
0xb1: true,
0xb2: true,
0xb3: true,
0xb4: true,
0xb5: true,
0xb6: true,
0xb7: true,
0xb8: true,
0xb9: true,
0xba: true,
0xbb: true,
0xbc: true,
0xbd: true,
0xbe: true,
0xbf: true,
0xc0: true,
0xc1: true,
0xc2: true,
0xc3: true,
0xc4: true,
0xc5: true,
0xc6: true,
0xc7: true,
0xc8: true,
0xc9: true,
0xca: true,
0xcb: true,
0xcc: true,
0xcd: true,
0xce: true,
0xcf: true,
0xd0: true,
0xd1: true,
0xd2: true,
0xd3: true,
0xd4: true,
0xd5: true,
0xd6: true,
0xd7: true,
0xd8: true,
0xd9: true,
0xda: true,
0xdb: true,
0xdc: true,
0xdd: true,
0xde: true,
0xdf: true,
0xe0: true,
0xe1: true,
0xe2: true,
0xe3: true,
0xe4: true,
0xe5: true,
0xe6: true,
0xe7: true,
0xe8: true,
0xe9: true,
0xea: true,
0xeb: true,
0xec: true,
0xed: true,
0xee: true,
0xef: true,
0xf0: true,
0xf1: true,
0xf2: true,
0xf3: true,
0xf4: true,
0xf5: true,
0xf6: true,
0xf7: true,
0xf8: true,
0xf9: true,
0xfa: true,
0xfb: true,
0xfc: true,
0xfd: true,
0xfe: true,
0xff: true,
}
var needEscape = [256]bool{
'"': true,
'\\': true,
0x00: true,
0x01: true,
0x02: true,
0x03: true,
0x04: true,
0x05: true,
0x06: true,
0x07: true,
0x08: true,
0x09: true,
0x0a: true,
0x0b: true,
0x0c: true,
0x0d: true,
0x0e: true,
0x0f: true,
0x10: true,
0x11: true,
0x12: true,
0x13: true,
0x14: true,
0x15: true,
0x16: true,
0x17: true,
0x18: true,
0x19: true,
0x1a: true,
0x1b: true,
0x1c: true,
0x1d: true,
0x1e: true,
0x1f: true,
/* 0x20 - 0x7f */
0x80: true,
0x81: true,
0x82: true,
0x83: true,
0x84: true,
0x85: true,
0x86: true,
0x87: true,
0x88: true,
0x89: true,
0x8a: true,
0x8b: true,
0x8c: true,
0x8d: true,
0x8e: true,
0x8f: true,
0x90: true,
0x91: true,
0x92: true,
0x93: true,
0x94: true,
0x95: true,
0x96: true,
0x97: true,
0x98: true,
0x99: true,
0x9a: true,
0x9b: true,
0x9c: true,
0x9d: true,
0x9e: true,
0x9f: true,
0xa0: true,
0xa1: true,
0xa2: true,
0xa3: true,
0xa4: true,
0xa5: true,
0xa6: true,
0xa7: true,
0xa8: true,
0xa9: true,
0xaa: true,
0xab: true,
0xac: true,
0xad: true,
0xae: true,
0xaf: true,
0xb0: true,
0xb1: true,
0xb2: true,
0xb3: true,
0xb4: true,
0xb5: true,
0xb6: true,
0xb7: true,
0xb8: true,
0xb9: true,
0xba: true,
0xbb: true,
0xbc: true,
0xbd: true,
0xbe: true,
0xbf: true,
0xc0: true,
0xc1: true,
0xc2: true,
0xc3: true,
0xc4: true,
0xc5: true,
0xc6: true,
0xc7: true,
0xc8: true,
0xc9: true,
0xca: true,
0xcb: true,
0xcc: true,
0xcd: true,
0xce: true,
0xcf: true,
0xd0: true,
0xd1: true,
0xd2: true,
0xd3: true,
0xd4: true,
0xd5: true,
0xd6: true,
0xd7: true,
0xd8: true,
0xd9: true,
0xda: true,
0xdb: true,
0xdc: true,
0xdd: true,
0xde: true,
0xdf: true,
0xe0: true,
0xe1: true,
0xe2: true,
0xe3: true,
0xe4: true,
0xe5: true,
0xe6: true,
0xe7: true,
0xe8: true,
0xe9: true,
0xea: true,
0xeb: true,
0xec: true,
0xed: true,
0xee: true,
0xef: true,
0xf0: true,
0xf1: true,
0xf2: true,
0xf3: true,
0xf4: true,
0xf5: true,
0xf6: true,
0xf7: true,
0xf8: true,
0xf9: true,
0xfa: true,
0xfb: true,
0xfc: true,
0xfd: true,
0xfe: true,
0xff: true,
}
var hex = "0123456789abcdef"
// escapeIndex finds the index of the first char in `s` that requires escaping.
// A char requires escaping if it's outside of the range of [0x20, 0x7F] or if
// it includes a double quote or backslash.
// If no chars in `s` require escaping, the return value is -1.
func escapeIndex(s string) int {
chunks := stringToUint64Slice(s)
for _, n := range chunks {
// combine masks before checking for the MSB of each byte. We include
// `n` in the mask to check whether any of the *input* byte MSBs were
// set (i.e. the byte was outside the ASCII range).
mask := n | below(n, 0x20) | contains(n, '"') | contains(n, '\\')
if (mask & msb) != 0 {
return bits.TrailingZeros64(mask&msb) / 8
}
}
valLen := len(s)
for i := len(chunks) * 8; i < valLen; i++ {
if needEscape[s[i]] {
return i
}
}
return -1
}
// below return a mask that can be used to determine if any of the bytes
// in `n` are below `b`. If a byte's MSB is set in the mask then that byte was
// below `b`. The result is only valid if `b`, and each byte in `n`, is below
// 0x80.
func below(n uint64, b byte) uint64 {
return n - expand(b)
}
// contains returns a mask that can be used to determine if any of the
// bytes in `n` are equal to `b`. If a byte's MSB is set in the mask then
// that byte is equal to `b`. The result is only valid if `b`, and each
// byte in `n`, is below 0x80.
func contains(n uint64, b byte) uint64 {
return (n ^ expand(b)) - lsb
}
// expand puts the specified byte into each of the 8 bytes of a uint64.
func expand(b byte) uint64 {
return lsb * uint64(b)
}
//nolint:govet
func stringToUint64Slice(s string) []uint64 {
return *(*[]uint64)(unsafe.Pointer(&reflect.SliceHeader{
Data: ((*reflect.StringHeader)(unsafe.Pointer(&s))).Data,
Len: len(s) / 8,
Cap: len(s) / 8,
}))
}
func AppendEscapedString(buf []byte, s string) []byte {
valLen := len(s)
if valLen == 0 {
return append(buf, `""`...)
}
buf = append(buf, '"')
var (
i, j int
)
if valLen >= 8 {
chunks := stringToUint64Slice(s)
for _, n := range chunks {
// combine masks before checking for the MSB of each byte. We include
// `n` in the mask to check whether any of the *input* byte MSBs were
// set (i.e. the byte was outside the ASCII range).
mask := n | (n - (lsb * 0x20)) |
((n ^ (lsb * '"')) - lsb) |
((n ^ (lsb * '\\')) - lsb) |
((n ^ (lsb * '<')) - lsb) |
((n ^ (lsb * '>')) - lsb) |
((n ^ (lsb * '&')) - lsb)
if (mask & msb) != 0 {
j = bits.TrailingZeros64(mask&msb) / 8
goto ESCAPE_END
}
}
for i := len(chunks) * 8; i < valLen; i++ {
if needEscapeWithHTML[s[i]] {
j = i
goto ESCAPE_END
}
}
// no found any escape characters.
return append(append(buf, s...), '"')
}
ESCAPE_END:
for j < valLen {
c := s[j]
if !needEscapeWithHTML[c] {
// fast path: most of the time, printable ascii characters are used
j++
continue
}
switch c {
case '\\', '"':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', c)
i = j + 1
j = j + 1
continue
case '\n':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 'n')
i = j + 1
j = j + 1
continue
case '\r':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 'r')
i = j + 1
j = j + 1
continue
case '\t':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 't')
i = j + 1
j = j + 1
continue
case '<', '>', '&':
buf = append(buf, s[i:j]...)
buf = append(buf, `\u00`...)
buf = append(buf, hex[c>>4], hex[c&0xF])
i = j + 1
j = j + 1
continue
}
// This encodes bytes < 0x20 except for \t, \n and \r.
if c < 0x20 {
buf = append(buf, s[i:j]...)
buf = append(buf, `\u00`...)
buf = append(buf, hex[c>>4], hex[c&0xF])
i = j + 1
j = j + 1
continue
}
r, size := utf8.DecodeRuneInString(s[j:])
if r == utf8.RuneError && size == 1 {
buf = append(buf, s[i:j]...)
buf = append(buf, `\ufffd`...)
i = j + size
j = j + size
continue
}
switch r {
case '\u2028', '\u2029':
// U+2028 is LINE SEPARATOR.
// U+2029 is PARAGRAPH SEPARATOR.
// They are both technically valid characters in JSON strings,
// but don't work in JSONP, which has to be evaluated as JavaScript,
// and can lead to security holes there. It is valid JSON to
// escape them, so we do so unconditionally.
// See http://timelessrepo.com/json-isnt-a-javascript-subset for discussion.
buf = append(buf, s[i:j]...)
buf = append(buf, `\u202`...)
buf = append(buf, hex[r&0xF])
i = j + size
j = j + size
continue
}
j += size
}
return append(append(buf, s[i:]...), '"')
}
func AppendString(buf []byte, s string) []byte {
valLen := len(s)
if valLen == 0 {
return append(buf, `""`...)
}
buf = append(buf, '"')
var escapeIdx int
if valLen >= 8 {
if escapeIdx = escapeIndex(s); escapeIdx < 0 {
return append(append(buf, s...), '"')
}
}
i := 0
j := escapeIdx
for j < valLen {
c := s[j]
if c >= 0x20 && c <= 0x7f && c != '\\' && c != '"' {
// fast path: most of the time, printable ascii characters are used
j++
continue
}
switch c {
case '\\', '"':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', c)
i = j + 1
j = j + 1
continue
case '\n':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 'n')
i = j + 1
j = j + 1
continue
case '\r':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 'r')
i = j + 1
j = j + 1
continue
case '\t':
buf = append(buf, s[i:j]...)
buf = append(buf, '\\', 't')
i = j + 1
j = j + 1
continue
case '<', '>', '&':
buf = append(buf, s[i:j]...)
buf = append(buf, `\u00`...)
buf = append(buf, hex[c>>4], hex[c&0xF])
i = j + 1
j = j + 1
continue
}
// This encodes bytes < 0x20 except for \t, \n and \r.
if c < 0x20 {
buf = append(buf, s[i:j]...)
buf = append(buf, `\u00`...)
buf = append(buf, hex[c>>4], hex[c&0xF])
i = j + 1
j = j + 1
continue
}
r, size := utf8.DecodeRuneInString(s[j:])
if r == utf8.RuneError && size == 1 {
buf = append(buf, s[i:j]...)
buf = append(buf, `\ufffd`...)
i = j + size
j = j + size
continue
}
switch r {
case '\u2028', '\u2029':
// U+2028 is LINE SEPARATOR.
// U+2029 is PARAGRAPH SEPARATOR.
// They are both technically valid characters in JSON strings,
// but don't work in JSONP, which has to be evaluated as JavaScript,
// and can lead to security holes there. It is valid JSON to
// escape them, so we do so unconditionally.
// See http://timelessrepo.com/json-isnt-a-javascript-subset for discussion.
buf = append(buf, s[i:j]...)
buf = append(buf, `\u202`...)
buf = append(buf, hex[r&0xF])
i = j + size
j = j + size
continue
}
j += size
}
return append(append(buf, s[i:]...), '"')
}

3759
internal/encoder/vm/vm.go Normal file

File diff suppressed because it is too large Load Diff

9
internal/runtime/type.go Normal file
View File

@ -0,0 +1,9 @@
package runtime
import "unsafe"
type SliceHeader struct {
Data unsafe.Pointer
Len int
Cap int
}