nvidia-container-toolkit/vendor/google.golang.org/protobuf/internal/encoding/text/encode.go
Evan Lezar 7bf5c25831 Update go vendoring
Signed-off-by: Evan Lezar <elezar@nvidia.com>
2021-10-21 12:07:42 +02:00

266 lines
6.9 KiB
Go

// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package text
import (
"math"
"math/bits"
"strconv"
"strings"
"unicode/utf8"
"google.golang.org/protobuf/internal/detrand"
"google.golang.org/protobuf/internal/errors"
)
// encType represents an encoding type.
type encType uint8
const (
_ encType = (1 << iota) / 2
name
scalar
messageOpen
messageClose
)
// Encoder provides methods to write out textproto constructs and values. The user is
// responsible for producing valid sequences of constructs and values.
type Encoder struct {
encoderState
indent string
delims [2]byte
outputASCII bool
}
type encoderState struct {
lastType encType
indents []byte
out []byte
}
// NewEncoder returns an Encoder.
//
// If indent is a non-empty string, it causes every entry in a List or Message
// to be preceded by the indent and trailed by a newline.
//
// If delims is not the zero value, it controls the delimiter characters used
// for messages (e.g., "{}" vs "<>").
//
// If outputASCII is true, strings will be serialized in such a way that
// multi-byte UTF-8 sequences are escaped. This property ensures that the
// overall output is ASCII (as opposed to UTF-8).
func NewEncoder(indent string, delims [2]byte, outputASCII bool) (*Encoder, error) {
e := &Encoder{}
if len(indent) > 0 {
if strings.Trim(indent, " \t") != "" {
return nil, errors.New("indent may only be composed of space and tab characters")
}
e.indent = indent
}
switch delims {
case [2]byte{0, 0}:
e.delims = [2]byte{'{', '}'}
case [2]byte{'{', '}'}, [2]byte{'<', '>'}:
e.delims = delims
default:
return nil, errors.New("delimiters may only be \"{}\" or \"<>\"")
}
e.outputASCII = outputASCII
return e, nil
}
// Bytes returns the content of the written bytes.
func (e *Encoder) Bytes() []byte {
return e.out
}
// StartMessage writes out the '{' or '<' symbol.
func (e *Encoder) StartMessage() {
e.prepareNext(messageOpen)
e.out = append(e.out, e.delims[0])
}
// EndMessage writes out the '}' or '>' symbol.
func (e *Encoder) EndMessage() {
e.prepareNext(messageClose)
e.out = append(e.out, e.delims[1])
}
// WriteName writes out the field name and the separator ':'.
func (e *Encoder) WriteName(s string) {
e.prepareNext(name)
e.out = append(e.out, s...)
e.out = append(e.out, ':')
}
// WriteBool writes out the given boolean value.
func (e *Encoder) WriteBool(b bool) {
if b {
e.WriteLiteral("true")
} else {
e.WriteLiteral("false")
}
}
// WriteString writes out the given string value.
func (e *Encoder) WriteString(s string) {
e.prepareNext(scalar)
e.out = appendString(e.out, s, e.outputASCII)
}
func appendString(out []byte, in string, outputASCII bool) []byte {
out = append(out, '"')
i := indexNeedEscapeInString(in)
in, out = in[i:], append(out, in[:i]...)
for len(in) > 0 {
switch r, n := utf8.DecodeRuneInString(in); {
case r == utf8.RuneError && n == 1:
// We do not report invalid UTF-8 because strings in the text format
// are used to represent both the proto string and bytes type.
r = rune(in[0])
fallthrough
case r < ' ' || r == '"' || r == '\\' || r == 0x7f:
out = append(out, '\\')
switch r {
case '"', '\\':
out = append(out, byte(r))
case '\n':
out = append(out, 'n')
case '\r':
out = append(out, 'r')
case '\t':
out = append(out, 't')
default:
out = append(out, 'x')
out = append(out, "00"[1+(bits.Len32(uint32(r))-1)/4:]...)
out = strconv.AppendUint(out, uint64(r), 16)
}
in = in[n:]
case r >= utf8.RuneSelf && (outputASCII || r <= 0x009f):
out = append(out, '\\')
if r <= math.MaxUint16 {
out = append(out, 'u')
out = append(out, "0000"[1+(bits.Len32(uint32(r))-1)/4:]...)
out = strconv.AppendUint(out, uint64(r), 16)
} else {
out = append(out, 'U')
out = append(out, "00000000"[1+(bits.Len32(uint32(r))-1)/4:]...)
out = strconv.AppendUint(out, uint64(r), 16)
}
in = in[n:]
default:
i := indexNeedEscapeInString(in[n:])
in, out = in[n+i:], append(out, in[:n+i]...)
}
}
out = append(out, '"')
return out
}
// indexNeedEscapeInString returns the index of the character that needs
// escaping. If no characters need escaping, this returns the input length.
func indexNeedEscapeInString(s string) int {
for i := 0; i < len(s); i++ {
if c := s[i]; c < ' ' || c == '"' || c == '\'' || c == '\\' || c >= 0x7f {
return i
}
}
return len(s)
}
// WriteFloat writes out the given float value for given bitSize.
func (e *Encoder) WriteFloat(n float64, bitSize int) {
e.prepareNext(scalar)
e.out = appendFloat(e.out, n, bitSize)
}
func appendFloat(out []byte, n float64, bitSize int) []byte {
switch {
case math.IsNaN(n):
return append(out, "nan"...)
case math.IsInf(n, +1):
return append(out, "inf"...)
case math.IsInf(n, -1):
return append(out, "-inf"...)
default:
return strconv.AppendFloat(out, n, 'g', -1, bitSize)
}
}
// WriteInt writes out the given signed integer value.
func (e *Encoder) WriteInt(n int64) {
e.prepareNext(scalar)
e.out = append(e.out, strconv.FormatInt(n, 10)...)
}
// WriteUint writes out the given unsigned integer value.
func (e *Encoder) WriteUint(n uint64) {
e.prepareNext(scalar)
e.out = append(e.out, strconv.FormatUint(n, 10)...)
}
// WriteLiteral writes out the given string as a literal value without quotes.
// This is used for writing enum literal strings.
func (e *Encoder) WriteLiteral(s string) {
e.prepareNext(scalar)
e.out = append(e.out, s...)
}
// prepareNext adds possible space and indentation for the next value based
// on last encType and indent option. It also updates e.lastType to next.
func (e *Encoder) prepareNext(next encType) {
defer func() {
e.lastType = next
}()
// Single line.
if len(e.indent) == 0 {
// Add space after each field before the next one.
if e.lastType&(scalar|messageClose) != 0 && next == name {
e.out = append(e.out, ' ')
// Add a random extra space to make output unstable.
if detrand.Bool() {
e.out = append(e.out, ' ')
}
}
return
}
// Multi-line.
switch {
case e.lastType == name:
e.out = append(e.out, ' ')
// Add a random extra space after name: to make output unstable.
if detrand.Bool() {
e.out = append(e.out, ' ')
}
case e.lastType == messageOpen && next != messageClose:
e.indents = append(e.indents, e.indent...)
e.out = append(e.out, '\n')
e.out = append(e.out, e.indents...)
case e.lastType&(scalar|messageClose) != 0:
if next == messageClose {
e.indents = e.indents[:len(e.indents)-len(e.indent)]
}
e.out = append(e.out, '\n')
e.out = append(e.out, e.indents...)
}
}
// Snapshot returns the current snapshot for use in Reset.
func (e *Encoder) Snapshot() encoderState {
return e.encoderState
}
// Reset resets the Encoder to the given encoderState from a Snapshot.
func (e *Encoder) Reset(es encoderState) {
e.encoderState = es
}