encode.go

// The majority of this code was vastly derived from Go Authors
// https://github.com/golang/go/tree/master/src/encoding/json.
//
// Copyright 2017 Manfred Touron. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// Package pycodec implements encoding and decoding of PYCODEC.
// The mapping between PYCODEC and Go values is described
// in the documentation for the Marshal and Unmarshal functions.
package pycodec

import (
	"bytes"
	"encoding"
	"encoding/base64"
	"fmt"
	"math"
	"reflect"
	"runtime"
	"sort"
	"strconv"
	"strings"
	"sync"
	"sync/atomic"
	"unicode"
	"unicode/utf8"
)

// Marshal returns the PYCODEC encoding of v.
//
// Marshal traverses the value v recursively.
// If an encountered value implements the Marshaler interface
// and is not a nil pointer, Marshal calls its MarshalPYCODEC method
// to produce PYCODEC. If no MarshalPYCODEC method is present but the
// value implements encoding.TextMarshaler instead, Marshal calls
// its MarshalText method and encodes the result as a PYCODEC string.
// The nil pointer exception is not strictly necessary
// but mimics a similar, necessary exception in the behavior of
// UnmarshalPYCODEC.
//
// Otherwise, Marshal uses the following type-dependent default encodings:
//
// Boolean values encode as PYCODEC booleans.
//
// Floating point, integer, and Number values encode as PYCODEC numbers.
//
// String values encode as PYCODEC strings coerced to valid UTF-8,
// replacing invalid bytes with the Unicode replacement rune.
// The angle brackets "<" and ">" are escaped to "\u003c" and "\u003e"
// to keep some browsers from misinterpreting PYCODEC output as HTML.
// Ampersand "&" is also escaped to "\u0026" for the same reason.
// This escaping can be disabled using an Encoder that had SetEscapeHTML(false)
// called on it.
//
// Array and slice values encode as PYCODEC arrays, except that
// []byte encodes as a base64-encoded string, and a nil slice
// encodes as the null PYCODEC value.
//
// Struct values encode as PYCODEC objects.
// Each exported struct field becomes a member of the object, using the
// field name as the object key, unless the field is omitted for one of the
// reasons given below.
//
// The encoding of each struct field can be customized by the format string
// stored under the "pycodec" key in the struct field's tag.
// The format string gives the name of the field, possibly followed by a
// comma-separated list of options. The name may be empty in order to
// specify options without overriding the default field name.
//
// The "omitempty" option specifies that the field should be omitted
// from the encoding if the field has an empty value, defined as
// false, 0, a nil pointer, a nil interface value, and any empty array,
// slice, map, or string.
//
// As a special case, if the field tag is "-", the field is always omitted.
// Note that a field with name "-" can still be generated using the tag "-,".
//
// Examples of struct field tags and their meanings:
//
//   // Field appears in PYCODEC as key "myName".
//   Field int `pycodec:"myName"`
//
//   // Field appears in PYCODEC as key "myName" and
//   // the field is omitted from the object if its value is empty,
//   // as defined above.
//   Field int `pycodec:"myName,omitempty"`
//
//   // Field appears in PYCODEC as key "Field" (the default), but
//   // the field is skipped if empty.
//   // Note the leading comma.
//   Field int `pycodec:",omitempty"`
//
//   // Field is ignored by this package.
//   Field int `pycodec:"-"`
//
//   // Field appears in PYCODEC as key "-".
//   Field int `pycodec:"-,"`
//
// The "string" option signals that a field is stored as PYCODEC inside a
// PYCODEC-encoded string. It applies only to fields of string, floating point,
// integer, or boolean types. This extra level of encoding is sometimes used
// when communicating with JavaScript programs:
//
//    Int64String int64 `pycodec:",string"`
//
// The key name will be used if it's a non-empty string consisting of
// only Unicode letters, digits, and ASCII punctuation except quotation
// marks, backslash, and comma.
//
// Anonymous struct fields are usually marshaled as if their inner exported fields
// were fields in the outer struct, subject to the usual Go visibility rules amended
// as described in the next paragraph.
// An anonymous struct field with a name given in its PYCODEC tag is treated as
// having that name, rather than being anonymous.
// An anonymous struct field of interface type is treated the same as having
// that type as its name, rather than being anonymous.
//
// The Go visibility rules for struct fields are amended for PYCODEC when
// deciding which field to marshal or unmarshal. If there are
// multiple fields at the same level, and that level is the least
// nested (and would therefore be the nesting level selected by the
// usual Go rules), the following extra rules apply:
//
// 1) Of those fields, if any are PYCODEC-tagged, only tagged fields are considered,
// even if there are multiple untagged fields that would otherwise conflict.
//
// 2) If there is exactly one field (tagged or not according to the first rule), that is selected.
//
// 3) Otherwise there are multiple fields, and all are ignored; no error occurs.
//
// Handling of anonymous struct fields is new in Go 1.1.
// Prior to Go 1.1, anonymous struct fields were ignored. To force ignoring of
// an anonymous struct field in both current and earlier versions, give the field
// a PYCODEC tag of "-".
//
// Map values encode as PYCODEC objects. The map's key type must either be a
// string, an integer type, or implement encoding.TextMarshaler. The map keys
// are sorted and used as PYCODEC object keys by applying the following rules,
// subject to the UTF-8 coercion described for string values above:
//   - string keys are used directly
//   - encoding.TextMarshalers are marshaled
//   - integer keys are converted to strings
//
// Pointer values encode as the value pointed to.
// A nil pointer encodes as the null PYCODEC value.
//
// Interface values encode as the value contained in the interface.
// A nil interface value encodes as the null PYCODEC value.
//
// Channel, complex, and function values cannot be encoded in PYCODEC.
// Attempting to encode such a value causes Marshal to return
// an UnsupportedTypeError.
//
// PYCODEC cannot represent cyclic data structures and Marshal does not
// handle them. Passing cyclic structures to Marshal will result in
// an infinite recursion.
//
func Marshal(v interface{}) ([]byte, error) {
	e := &encodeState{}
	err := e.marshal(v, encOpts{escapeHTML: true})
	if err != nil {
		return nil, err
	}
	return e.Bytes(), nil
}

// MarshalIndent is like Marshal but applies Indent to format the output.
func MarshalIndent(v interface{}, prefix, indent string) ([]byte, error) {
	b, err := Marshal(v)
	if err != nil {
		return nil, err
	}
	var buf bytes.Buffer
	err = Indent(&buf, b, prefix, indent)
	if err != nil {
		return nil, err
	}
	return buf.Bytes(), nil
}

// HTMLEscape appends to dst the PYCODEC-encoded src with <, >, &, U+2028 and U+2029
// characters inside string literals changed to \u003c, \u003e, \u0026, \u2028, \u2029
// so that the PYCODEC will be safe to embed inside HTML <script> tags.
// For historical reasons, web browsers don't honor standard HTML
// escaping within <script> tags, so an alternative PYCODEC encoding must
// be used.
func HTMLEscape(dst *bytes.Buffer, src []byte) {
	// The characters can only appear in string literals,
	// so just scan the string one byte at a time.
	start := 0
	for i, c := range src {
		if c == '<' || c == '>' || c == '&' {
			if start < i {
				dst.Write(src[start:i])
			}
			dst.WriteString(`\u00`)
			dst.WriteByte(hex[c>>4])
			dst.WriteByte(hex[c&0xF])
			start = i + 1
		}
		// Convert U+2028 and U+2029 (E2 80 A8 and E2 80 A9).
		if c == 0xE2 && i+2 < len(src) && src[i+1] == 0x80 && src[i+2]&^1 == 0xA8 {
			if start < i {
				dst.Write(src[start:i])
			}
			dst.WriteString(`\u202`)
			dst.WriteByte(hex[src[i+2]&0xF])
			start = i + 3
		}
	}
	if start < len(src) {
		dst.Write(src[start:])
	}
}

// Marshaler is the interface implemented by types that
// can marshal themselves into valid PYCODEC.
type Marshaler interface {
	MarshalPYCODEC() ([]byte, error)
}

// An UnsupportedTypeError is returned by Marshal when attempting
// to encode an unsupported value type.
type UnsupportedTypeError struct {
	Type reflect.Type
}

func (e *UnsupportedTypeError) Error() string {
	return "pycodec: unsupported type: " + e.Type.String()
}

type UnsupportedValueError struct {
	Value reflect.Value
	Str   string
}

func (e *UnsupportedValueError) Error() string {
	return "pycodec: unsupported value: " + e.Str
}

// Before Go 1.2, an InvalidUTF8Error was returned by Marshal when
// attempting to encode a string value with invalid UTF-8 sequences.
// As of Go 1.2, Marshal instead coerces the string to valid UTF-8 by
// replacing invalid bytes with the Unicode replacement rune U+FFFD.
// This error is no longer generated but is kept for backwards compatibility
// with programs that might mention it.
type InvalidUTF8Error struct {
	S string // the whole string value that caused the error
}

func (e *InvalidUTF8Error) Error() string {
	return "pycodec: invalid UTF-8 in string: " + strconv.Quote(e.S)
}

type MarshalerError struct {
	Type reflect.Type
	Err  error
}

func (e *MarshalerError) Error() string {
	return "pycodec: error calling MarshalPYCODEC for type " + e.Type.String() + ": " + e.Err.Error()
}

var hex = "0123456789abcdef"

// An encodeState encodes PYCODEC into a bytes.Buffer.
type encodeState struct {
	bytes.Buffer // accumulated output
	scratch      [64]byte
}

var encodeStatePool sync.Pool

func newEncodeState() *encodeState {
	if v := encodeStatePool.Get(); v != nil {
		e := v.(*encodeState)
		e.Reset()
		return e
	}
	return new(encodeState)
}

func (e *encodeState) marshal(v interface{}, opts encOpts) (err error) {
	defer func() {
		if r := recover(); r != nil {
			if _, ok := r.(runtime.Error); ok {
				panic(r)
			}
			if s, ok := r.(string); ok {
				panic(s)
			}
			err = r.(error)
		}
	}()
	e.reflectValue(reflect.ValueOf(v), opts)
	return nil
}

func (e *encodeState) error(err error) {
	panic(err)
}

func isEmptyValue(v reflect.Value) bool {
	switch v.Kind() {
	case reflect.Array, reflect.Map, reflect.Slice, reflect.String:
		return v.Len() == 0
	case reflect.Bool:
		return !v.Bool()
	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
		return v.Int() == 0
	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
		return v.Uint() == 0
	case reflect.Float32, reflect.Float64:
		return v.Float() == 0
	case reflect.Interface, reflect.Ptr:
		return v.IsNil()
	}
	return false
}

func (e *encodeState) reflectValue(v reflect.Value, opts encOpts) {
	valueEncoder(v)(e, v, opts)
}

type encOpts struct {
	// quoted causes primitive fields to be encoded inside PYCODEC strings.
	quoted bool
	// escapeHTML causes '<', '>', and '&' to be escaped in PYCODEC strings.
	escapeHTML bool
}

type encoderFunc func(e *encodeState, v reflect.Value, opts encOpts)

var encoderCache struct {
	sync.RWMutex
	m map[reflect.Type]encoderFunc
}

func valueEncoder(v reflect.Value) encoderFunc {
	if !v.IsValid() {
		return invalidValueEncoder
	}
	return typeEncoder(v.Type())
}

func typeEncoder(t reflect.Type) encoderFunc {
	encoderCache.RLock()
	f := encoderCache.m[t]
	encoderCache.RUnlock()
	if f != nil {
		return f
	}

	// To deal with recursive types, populate the map with an
	// indirect func before we build it. This type waits on the
	// real func (f) to be ready and then calls it. This indirect
	// func is only used for recursive types.
	encoderCache.Lock()
	if encoderCache.m == nil {
		encoderCache.m = make(map[reflect.Type]encoderFunc)
	}
	var wg sync.WaitGroup
	wg.Add(1)
	encoderCache.m[t] = func(e *encodeState, v reflect.Value, opts encOpts) {
		wg.Wait()
		f(e, v, opts)
	}
	encoderCache.Unlock()

	// Compute fields without lock.
	// Might duplicate effort but won't hold other computations back.
	f = newTypeEncoder(t, true)
	wg.Done()
	encoderCache.Lock()
	encoderCache.m[t] = f
	encoderCache.Unlock()
	return f
}

var (
	marshalerType     = reflect.TypeOf(new(Marshaler)).Elem()
	textMarshalerType = reflect.TypeOf(new(encoding.TextMarshaler)).Elem()
)

// newTypeEncoder constructs an encoderFunc for a type.
// The returned encoder only checks CanAddr when allowAddr is true.
func newTypeEncoder(t reflect.Type, allowAddr bool) encoderFunc {
	if t.Implements(marshalerType) {
		return marshalerEncoder
	}
	if t.Kind() != reflect.Ptr && allowAddr {
		if reflect.PtrTo(t).Implements(marshalerType) {
			return newCondAddrEncoder(addrMarshalerEncoder, newTypeEncoder(t, false))
		}
	}

	if t.Implements(textMarshalerType) {
		return textMarshalerEncoder
	}
	if t.Kind() != reflect.Ptr && allowAddr {
		if reflect.PtrTo(t).Implements(textMarshalerType) {
			return newCondAddrEncoder(addrTextMarshalerEncoder, newTypeEncoder(t, false))
		}
	}

	switch t.Kind() {
	case reflect.Bool:
		return boolEncoder
	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
		return intEncoder
	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
		return uintEncoder
	case reflect.Float32:
		return float32Encoder
	case reflect.Float64:
		return float64Encoder
	case reflect.String:
		return stringEncoder
	case reflect.Interface:
		return interfaceEncoder
	case reflect.Struct:
		return newStructEncoder(t)
	case reflect.Map:
		return newMapEncoder(t)
	case reflect.Slice:
		return newSliceEncoder(t)
	case reflect.Array:
		return newArrayEncoder(t)
	case reflect.Ptr:
		return newPtrEncoder(t)
	default:
		return unsupportedTypeEncoder
	}
}

func invalidValueEncoder(e *encodeState, v reflect.Value, _ encOpts) {
	e.WriteString("null")
}

func marshalerEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	if v.Kind() == reflect.Ptr && v.IsNil() {
		e.WriteString("null")
		return
	}
	m, ok := v.Interface().(Marshaler)
	if !ok {
		e.WriteString("null")
		return
	}
	b, err := m.MarshalPYCODEC()
	if err == nil {
		// copy PYCODEC into buffer, checking validity.
		err = compact(&e.Buffer, b, opts.escapeHTML)
	}
	if err != nil {
		e.error(&MarshalerError{v.Type(), err})
	}
}

func addrMarshalerEncoder(e *encodeState, v reflect.Value, _ encOpts) {
	va := v.Addr()
	if va.IsNil() {
		e.WriteString("null")
		return
	}
	m := va.Interface().(Marshaler)
	b, err := m.MarshalPYCODEC()
	if err == nil {
		// copy PYCODEC into buffer, checking validity.
		err = compact(&e.Buffer, b, true)
	}
	if err != nil {
		e.error(&MarshalerError{v.Type(), err})
	}
}

func textMarshalerEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	if v.Kind() == reflect.Ptr && v.IsNil() {
		e.WriteString("null")
		return
	}
	m := v.Interface().(encoding.TextMarshaler)
	b, err := m.MarshalText()
	if err != nil {
		e.error(&MarshalerError{v.Type(), err})
	}
	e.stringBytes(b, opts.escapeHTML)
}

func addrTextMarshalerEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	va := v.Addr()
	if va.IsNil() {
		e.WriteString("null")
		return
	}
	m := va.Interface().(encoding.TextMarshaler)
	b, err := m.MarshalText()
	if err != nil {
		e.error(&MarshalerError{v.Type(), err})
	}
	e.stringBytes(b, opts.escapeHTML)
}

func boolEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	if opts.quoted {
		e.WriteByte('"')
	}
	if v.Bool() {
		e.WriteString("true")
	} else {
		e.WriteString("false")
	}
	if opts.quoted {
		e.WriteByte('"')
	}
}

func intEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	b := strconv.AppendInt(e.scratch[:0], v.Int(), 10)
	if opts.quoted {
		e.WriteByte('"')
	}
	e.Write(b)
	if opts.quoted {
		e.WriteByte('"')
	}
}

func uintEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	b := strconv.AppendUint(e.scratch[:0], v.Uint(), 10)
	if opts.quoted {
		e.WriteByte('"')
	}
	e.Write(b)
	if opts.quoted {
		e.WriteByte('"')
	}
}

type floatEncoder int // number of bits

func (bits floatEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	f := v.Float()
	if math.IsInf(f, 0) || math.IsNaN(f) {
		e.error(&UnsupportedValueError{v, strconv.FormatFloat(f, 'g', -1, int(bits))})
	}

	// Convert as if by ES6 number to string conversion.
	// This matches most other PYCODEC generators.
	// See golang.org/issue/6384 and golang.org/issue/14135.
	// Like fmt %g, but the exponent cutoffs are different
	// and exponents themselves are not padded to two digits.
	b := e.scratch[:0]
	abs := math.Abs(f)
	fmt := byte('f')
	// Note: Must use float32 comparisons for underlying float32 value to get precise cutoffs right.
	if abs != 0 {
		if bits == 64 && (abs < 1e-6 || abs >= 1e21) || bits == 32 && (float32(abs) < 1e-6 || float32(abs) >= 1e21) {
			fmt = 'e'
		}
	}
	b = strconv.AppendFloat(b, f, fmt, -1, int(bits))
	if fmt == 'e' {
		// clean up e-09 to e-9
		n := len(b)
		if n >= 4 && b[n-4] == 'e' && b[n-3] == '-' && b[n-2] == '0' {
			b[n-2] = b[n-1]
			b = b[:n-1]
		}
	}

	if opts.quoted {
		e.WriteByte('"')
	}
	e.Write(b)
	if opts.quoted {
		e.WriteByte('"')
	}
}

var (
	float32Encoder = (floatEncoder(32)).encode
	float64Encoder = (floatEncoder(64)).encode
)

func stringEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	if v.Type() == numberType {
		numStr := v.String()
		// In Go1.5 the empty string encodes to "0", while this is not a valid number literal
		// we keep compatibility so check validity after this.
		if numStr == "" {
			numStr = "0" // Number's zero-val
		}
		if !isValidNumber(numStr) {
			e.error(fmt.Errorf("pycodec: invalid number literal %q", numStr))
		}
		e.WriteString(numStr)
		return
	}
	if opts.quoted {
		sb, err := Marshal(v.String())
		if err != nil {
			e.error(err)
		}
		e.string(string(sb), opts.escapeHTML)
	} else {
		e.string(v.String(), opts.escapeHTML)
	}
}

func interfaceEncoder(e *encodeState, v reflect.Value, opts encOpts) {
	if v.IsNil() {
		e.WriteString("null")
		return
	}
	e.reflectValue(v.Elem(), opts)
}

func unsupportedTypeEncoder(e *encodeState, v reflect.Value, _ encOpts) {
	e.error(&UnsupportedTypeError{v.Type()})
}

type structEncoder struct {
	fields    []field
	fieldEncs []encoderFunc
}

func (se *structEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	e.WriteByte('{')
	first := true
	for i, f := range se.fields {
		fv := fieldByIndex(v, f.index)
		if !fv.IsValid() || f.omitEmpty && isEmptyValue(fv) {
			continue
		}
		if first {
			first = false
		} else {
			e.WriteByte(',')
		}
		e.string(f.name, opts.escapeHTML)
		e.WriteByte(':')
		opts.quoted = f.quoted
		se.fieldEncs[i](e, fv, opts)
	}
	e.WriteByte('}')
}

func newStructEncoder(t reflect.Type) encoderFunc {
	fields := cachedTypeFields(t)
	se := &structEncoder{
		fields:    fields,
		fieldEncs: make([]encoderFunc, len(fields)),
	}
	for i, f := range fields {
		se.fieldEncs[i] = typeEncoder(typeByIndex(t, f.index))
	}
	return se.encode
}

type mapEncoder struct {
	elemEnc encoderFunc
}

func (me *mapEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	if v.IsNil() {
		e.WriteString("null")
		return
	}
	e.WriteByte('{')

	// Extract and sort the keys.
	keys := v.MapKeys()
	sv := make([]reflectWithString, len(keys))
	for i, v := range keys {
		sv[i].v = v
		if err := sv[i].resolve(); err != nil {
			e.error(&MarshalerError{v.Type(), err})
		}
	}
	sort.Slice(sv, func(i, j int) bool { return sv[i].s < sv[j].s })

	for i, kv := range sv {
		if i > 0 {
			e.WriteByte(',')
		}
		e.string(kv.s, opts.escapeHTML)
		e.WriteByte(':')
		me.elemEnc(e, v.MapIndex(kv.v), opts)
	}
	e.WriteByte('}')
}

func newMapEncoder(t reflect.Type) encoderFunc {
	switch t.Key().Kind() {
	case reflect.String,
		reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64,
		reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
	default:
		if !t.Key().Implements(textMarshalerType) {
			return unsupportedTypeEncoder
		}
	}
	me := &mapEncoder{typeEncoder(t.Elem())}
	return me.encode
}

func encodeByteSlice(e *encodeState, v reflect.Value, _ encOpts) {
	if v.IsNil() {
		e.WriteString("null")
		return
	}
	s := v.Bytes()
	e.WriteByte('"')
	if len(s) < 1024 {
		// for small buffers, using Encode directly is much faster.
		dst := make([]byte, base64.StdEncoding.EncodedLen(len(s)))
		base64.StdEncoding.Encode(dst, s)
		e.Write(dst)
	} else {
		// for large buffers, avoid unnecessary extra temporary
		// buffer space.
		enc := base64.NewEncoder(base64.StdEncoding, e)
		enc.Write(s)
		enc.Close()
	}
	e.WriteByte('"')
}

// sliceEncoder just wraps an arrayEncoder, checking to make sure the value isn't nil.
type sliceEncoder struct {
	arrayEnc encoderFunc
}

func (se *sliceEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	if v.IsNil() {
		e.WriteString("null")
		return
	}
	se.arrayEnc(e, v, opts)
}

func newSliceEncoder(t reflect.Type) encoderFunc {
	// Byte slices get special treatment; arrays don't.
	if t.Elem().Kind() == reflect.Uint8 {
		p := reflect.PtrTo(t.Elem())
		if !p.Implements(marshalerType) && !p.Implements(textMarshalerType) {
			return encodeByteSlice
		}
	}
	enc := &sliceEncoder{newArrayEncoder(t)}
	return enc.encode
}

type arrayEncoder struct {
	elemEnc encoderFunc
}

func (ae *arrayEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	e.WriteByte('[')
	n := v.Len()
	for i := 0; i < n; i++ {
		if i > 0 {
			e.WriteByte(',')
		}
		ae.elemEnc(e, v.Index(i), opts)
	}
	e.WriteByte(']')
}

func newArrayEncoder(t reflect.Type) encoderFunc {
	enc := &arrayEncoder{typeEncoder(t.Elem())}
	return enc.encode
}

type ptrEncoder struct {
	elemEnc encoderFunc
}

func (pe *ptrEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	if v.IsNil() {
		e.WriteString("null")
		return
	}
	pe.elemEnc(e, v.Elem(), opts)
}

func newPtrEncoder(t reflect.Type) encoderFunc {
	enc := &ptrEncoder{typeEncoder(t.Elem())}
	return enc.encode
}

type condAddrEncoder struct {
	canAddrEnc, elseEnc encoderFunc
}

func (ce *condAddrEncoder) encode(e *encodeState, v reflect.Value, opts encOpts) {
	if v.CanAddr() {
		ce.canAddrEnc(e, v, opts)
	} else {
		ce.elseEnc(e, v, opts)
	}
}

// newCondAddrEncoder returns an encoder that checks whether its value
// CanAddr and delegates to canAddrEnc if so, else to elseEnc.
func newCondAddrEncoder(canAddrEnc, elseEnc encoderFunc) encoderFunc {
	enc := &condAddrEncoder{canAddrEnc: canAddrEnc, elseEnc: elseEnc}
	return enc.encode
}

func isValidTag(s string) bool {
	if s == "" {
		return false
	}
	for _, c := range s {
		switch {
		case strings.ContainsRune("!#$%&()*+-./:<=>?@[]^_{|}~ ", c):
			// Backslash and quote chars are reserved, but
			// otherwise any punctuation chars are allowed
			// in a tag name.
		default:
			if !unicode.IsLetter(c) && !unicode.IsDigit(c) {
				return false
			}
		}
	}
	return true
}

func fieldByIndex(v reflect.Value, index []int) reflect.Value {
	for _, i := range index {
		if v.Kind() == reflect.Ptr {
			if v.IsNil() {
				return reflect.Value{}
			}
			v = v.Elem()
		}
		v = v.Field(i)
	}
	return v
}

func typeByIndex(t reflect.Type, index []int) reflect.Type {
	for _, i := range index {
		if t.Kind() == reflect.Ptr {
			t = t.Elem()
		}
		t = t.Field(i).Type
	}
	return t
}

type reflectWithString struct {
	v reflect.Value
	s string
}

func (w *reflectWithString) resolve() error {
	if w.v.Kind() == reflect.String {
		w.s = w.v.String()
		return nil
	}
	if tm, ok := w.v.Interface().(encoding.TextMarshaler); ok {
		buf, err := tm.MarshalText()
		w.s = string(buf)
		return err
	}
	switch w.v.Kind() {
	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
		w.s = strconv.FormatInt(w.v.Int(), 10)
		return nil
	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
		w.s = strconv.FormatUint(w.v.Uint(), 10)
		return nil
	}
	panic("unexpected map key type")
}

// NOTE: keep in sync with stringBytes below.
func (e *encodeState) string(s string, escapeHTML bool) int {
	len0 := e.Len()
	e.WriteByte('"')
	start := 0
	for i := 0; i < len(s); {
		if b := s[i]; b < utf8.RuneSelf {
			if htmlSafeSet[b] || (!escapeHTML && safeSet[b]) {
				i++
				continue
			}
			if start < i {
				e.WriteString(s[start:i])
			}
			switch b {
			case '\\', '"':
				e.WriteByte('\\')
				e.WriteByte(b)
			case '\n':
				e.WriteByte('\\')
				e.WriteByte('n')
			case '\r':
				e.WriteByte('\\')
				e.WriteByte('r')
			case '\t':
				e.WriteByte('\\')
				e.WriteByte('t')
			default:
				// This encodes bytes < 0x20 except for \t, \n and \r.
				// If escapeHTML is set, it also escapes <, >, and &
				// because they can lead to security holes when
				// user-controlled strings are rendered into PYCODEC
				// and served to some browsers.
				e.WriteString(`\u00`)
				e.WriteByte(hex[b>>4])
				e.WriteByte(hex[b&0xF])
			}
			i++
			start = i
			continue
		}
		c, size := utf8.DecodeRuneInString(s[i:])
		if c == utf8.RuneError && size == 1 {
			if start < i {
				e.WriteString(s[start:i])
			}
			e.WriteString(`\ufffd`)
			i += size
			start = i
			continue
		}
		// U+2028 is LINE SEPARATOR.
		// U+2029 is PARAGRAPH SEPARATOR.
		// They are both technically valid characters in PYCODEC strings,
		// but don't work in PYCODECP, which has to be evaluated as JavaScript,
		// and can lead to security holes there. It is valid PYCODEC to
		// escape them, so we do so unconditionally.
		// See http://timelessrepo.com/pycodec-isnt-a-javascript-subset for discussion.
		if c == '\u2028' || c == '\u2029' {
			if start < i {
				e.WriteString(s[start:i])
			}
			e.WriteString(`\u202`)
			e.WriteByte(hex[c&0xF])
			i += size
			start = i
			continue
		}
		i += size
	}
	if start < len(s) {
		e.WriteString(s[start:])
	}
	e.WriteByte('"')
	return e.Len() - len0
}

// NOTE: keep in sync with string above.
func (e *encodeState) stringBytes(s []byte, escapeHTML bool) int {
	len0 := e.Len()
	e.WriteByte('"')
	start := 0
	for i := 0; i < len(s); {
		if b := s[i]; b < utf8.RuneSelf {
			if htmlSafeSet[b] || (!escapeHTML && safeSet[b]) {
				i++
				continue
			}
			if start < i {
				e.Write(s[start:i])
			}
			switch b {
			case '\\', '"':
				e.WriteByte('\\')
				e.WriteByte(b)
			case '\n':
				e.WriteByte('\\')
				e.WriteByte('n')
			case '\r':
				e.WriteByte('\\')
				e.WriteByte('r')
			case '\t':
				e.WriteByte('\\')
				e.WriteByte('t')
			default:
				// This encodes bytes < 0x20 except for \t, \n and \r.
				// If escapeHTML is set, it also escapes <, >, and &
				// because they can lead to security holes when
				// user-controlled strings are rendered into PYCODEC
				// and served to some browsers.
				e.WriteString(`\u00`)
				e.WriteByte(hex[b>>4])
				e.WriteByte(hex[b&0xF])
			}
			i++
			start = i
			continue
		}
		c, size := utf8.DecodeRune(s[i:])
		if c == utf8.RuneError && size == 1 {
			if start < i {
				e.Write(s[start:i])
			}
			e.WriteString(`\ufffd`)
			i += size
			start = i
			continue
		}
		// U+2028 is LINE SEPARATOR.
		// U+2029 is PARAGRAPH SEPARATOR.
		// They are both technically valid characters in PYCODEC strings,
		// but don't work in PYCODECP, which has to be evaluated as JavaScript,
		// and can lead to security holes there. It is valid PYCODEC to
		// escape them, so we do so unconditionally.
		// See http://timelessrepo.com/pycodec-isnt-a-javascript-subset for discussion.
		if c == '\u2028' || c == '\u2029' {
			if start < i {
				e.Write(s[start:i])
			}
			e.WriteString(`\u202`)
			e.WriteByte(hex[c&0xF])
			i += size
			start = i
			continue
		}
		i += size
	}
	if start < len(s) {
		e.Write(s[start:])
	}
	e.WriteByte('"')
	return e.Len() - len0
}

// A field represents a single field found in a struct.
type field struct {
	name      string
	nameBytes []byte                 // []byte(name)
	equalFold func(s, t []byte) bool // bytes.EqualFold or equivalent

	tag       bool
	index     []int
	typ       reflect.Type
	omitEmpty bool
	quoted    bool
}

func fillField(f field) field {
	f.nameBytes = []byte(f.name)
	f.equalFold = foldFunc(f.nameBytes)
	return f
}

// byIndex sorts field by index sequence.
type byIndex []field

func (x byIndex) Len() int { return len(x) }

func (x byIndex) Swap(i, j int) { x[i], x[j] = x[j], x[i] }

func (x byIndex) Less(i, j int) bool {
	for k, xik := range x[i].index {
		if k >= len(x[j].index) {
			return false
		}
		if xik != x[j].index[k] {
			return xik < x[j].index[k]
		}
	}
	return len(x[i].index) < len(x[j].index)
}

// typeFields returns a list of fields that PYCODEC should recognize for the given type.
// The algorithm is breadth-first search over the set of structs to include - the top struct
// and then any reachable anonymous structs.
func typeFields(t reflect.Type) []field {
	// Anonymous fields to explore at the current level and the next.
	current := []field{}
	next := []field{{typ: t}}

	// Count of queued names for current level and the next.
	count := map[reflect.Type]int{}
	nextCount := map[reflect.Type]int{}

	// Types already visited at an earlier level.
	visited := map[reflect.Type]bool{}

	// Fields found.
	var fields []field

	for len(next) > 0 {
		current, next = next, current[:0]
		count, nextCount = nextCount, map[reflect.Type]int{}

		for _, f := range current {
			if visited[f.typ] {
				continue
			}
			visited[f.typ] = true

			// Scan f.typ for fields to include.
			for i := 0; i < f.typ.NumField(); i++ {
				sf := f.typ.Field(i)
				if sf.PkgPath != "" && !sf.Anonymous { // unexported
					continue
				}
				tag := sf.Tag.Get("pycodec")
				if tag == "-" {
					continue
				}
				name, opts := parseTag(tag)
				if !isValidTag(name) {
					name = ""
				}
				index := make([]int, len(f.index)+1)
				copy(index, f.index)
				index[len(f.index)] = i

				ft := sf.Type
				if ft.Name() == "" && ft.Kind() == reflect.Ptr {
					// Follow pointer.
					ft = ft.Elem()
				}

				// Only strings, floats, integers, and booleans can be quoted.
				quoted := false
				if opts.Contains("string") {
					switch ft.Kind() {
					case reflect.Bool,
						reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64,
						reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64,
						reflect.Float32, reflect.Float64,
						reflect.String:
						quoted = true
					}
				}

				// Record found field and index sequence.
				if name != "" || !sf.Anonymous || ft.Kind() != reflect.Struct {
					tagged := name != ""
					if name == "" {
						name = sf.Name
					}
					fields = append(fields, fillField(field{
						name:      name,
						tag:       tagged,
						index:     index,
						typ:       ft,
						omitEmpty: opts.Contains("omitempty"),
						quoted:    quoted,
					}))
					if count[f.typ] > 1 {
						// If there were multiple instances, add a second,
						// so that the annihilation code will see a duplicate.
						// It only cares about the distinction between 1 or 2,
						// so don't bother generating any more copies.
						fields = append(fields, fields[len(fields)-1])
					}
					continue
				}

				// Record new anonymous struct to explore in next round.
				nextCount[ft]++
				if nextCount[ft] == 1 {
					next = append(next, fillField(field{name: ft.Name(), index: index, typ: ft}))
				}
			}
		}
	}

	sort.Slice(fields, func(i, j int) bool {
		x := fields
		// sort field by name, breaking ties with depth, then
		// breaking ties with "name came from pycodec tag", then
		// breaking ties with index sequence.
		if x[i].name != x[j].name {
			return x[i].name < x[j].name
		}
		if len(x[i].index) != len(x[j].index) {
			return len(x[i].index) < len(x[j].index)
		}
		if x[i].tag != x[j].tag {
			return x[i].tag
		}
		return byIndex(x).Less(i, j)
	})

	// Delete all fields that are hidden by the Go rules for embedded fields,
	// except that fields with PYCODEC tags are promoted.

	// The fields are sorted in primary order of name, secondary order
	// of field index length. Loop over names; for each name, delete
	// hidden fields by choosing the one dominant field that survives.
	out := fields[:0]
	for advance, i := 0, 0; i < len(fields); i += advance {
		// One iteration per name.
		// Find the sequence of fields with the name of this first field.
		fi := fields[i]
		name := fi.name
		for advance = 1; i+advance < len(fields); advance++ {
			fj := fields[i+advance]
			if fj.name != name {
				break
			}
		}
		if advance == 1 { // Only one field with this name
			out = append(out, fi)
			continue
		}
		dominant, ok := dominantField(fields[i : i+advance])
		if ok {
			out = append(out, dominant)
		}
	}

	fields = out
	sort.Sort(byIndex(fields))

	return fields
}

// dominantField looks through the fields, all of which are known to
// have the same name, to find the single field that dominates the
// others using Go's embedding rules, modified by the presence of
// PYCODEC tags. If there are multiple top-level fields, the boolean
// will be false: This condition is an error in Go and we skip all
// the fields.
func dominantField(fields []field) (field, bool) {
	// The fields are sorted in increasing index-length order. The winner
	// must therefore be one with the shortest index length. Drop all
	// longer entries, which is easy: just truncate the slice.
	length := len(fields[0].index)
	tagged := -1 // Index of first tagged field.
	for i, f := range fields {
		if len(f.index) > length {
			fields = fields[:i]
			break
		}
		if f.tag {
			if tagged >= 0 {
				// Multiple tagged fields at the same level: conflict.
				// Return no field.
				return field{}, false
			}
			tagged = i
		}
	}
	if tagged >= 0 {
		return fields[tagged], true
	}
	// All remaining fields have the same length. If there's more than one,
	// we have a conflict (two fields named "X" at the same level) and we
	// return no field.
	if len(fields) > 1 {
		return field{}, false
	}
	return fields[0], true
}

var fieldCache struct {
	value atomic.Value // map[reflect.Type][]field
	mu    sync.Mutex   // used only by writers
}

// cachedTypeFields is like typeFields but uses a cache to avoid repeated work.
func cachedTypeFields(t reflect.Type) []field {
	m, _ := fieldCache.value.Load().(map[reflect.Type][]field)
	f := m[t]
	if f != nil {
		return f
	}

	// Compute fields without lock.
	// Might duplicate effort but won't hold other computations back.
	f = typeFields(t)
	if f == nil {
		f = []field{}
	}

	fieldCache.mu.Lock()
	m, _ = fieldCache.value.Load().(map[reflect.Type][]field)
	newM := make(map[reflect.Type][]field, len(m)+1)
	for k, v := range m {
		newM[k] = v
	}
	newM[t] = f
	fieldCache.value.Store(newM)
	fieldCache.mu.Unlock()
	return f
}