amit.ghosh | 258d14c | 2020-10-02 15:13:38 +0200 | [diff] [blame] | 1 | // Copyright 2018 The Go Authors. All rights reserved. |
| 2 | // Use of this source code is governed by a BSD-style |
| 3 | // license that can be found in the LICENSE file. |
| 4 | |
| 5 | package json |
| 6 | |
| 7 | import ( |
| 8 | "math" |
| 9 | "math/bits" |
| 10 | "strconv" |
| 11 | "strings" |
| 12 | "unicode/utf8" |
| 13 | |
| 14 | "google.golang.org/protobuf/internal/detrand" |
| 15 | "google.golang.org/protobuf/internal/errors" |
| 16 | ) |
| 17 | |
| 18 | // kind represents an encoding type. |
| 19 | type kind uint8 |
| 20 | |
| 21 | const ( |
| 22 | _ kind = (1 << iota) / 2 |
| 23 | name |
| 24 | scalar |
| 25 | objectOpen |
| 26 | objectClose |
| 27 | arrayOpen |
| 28 | arrayClose |
| 29 | ) |
| 30 | |
| 31 | // Encoder provides methods to write out JSON constructs and values. The user is |
| 32 | // responsible for producing valid sequences of JSON constructs and values. |
| 33 | type Encoder struct { |
| 34 | indent string |
| 35 | lastKind kind |
| 36 | indents []byte |
| 37 | out []byte |
| 38 | } |
| 39 | |
| 40 | // NewEncoder returns an Encoder. |
| 41 | // |
| 42 | // If indent is a non-empty string, it causes every entry for an Array or Object |
| 43 | // to be preceded by the indent and trailed by a newline. |
| 44 | func NewEncoder(indent string) (*Encoder, error) { |
| 45 | e := &Encoder{} |
| 46 | if len(indent) > 0 { |
| 47 | if strings.Trim(indent, " \t") != "" { |
| 48 | return nil, errors.New("indent may only be composed of space or tab characters") |
| 49 | } |
| 50 | e.indent = indent |
| 51 | } |
| 52 | return e, nil |
| 53 | } |
| 54 | |
| 55 | // Bytes returns the content of the written bytes. |
| 56 | func (e *Encoder) Bytes() []byte { |
| 57 | return e.out |
| 58 | } |
| 59 | |
| 60 | // WriteNull writes out the null value. |
| 61 | func (e *Encoder) WriteNull() { |
| 62 | e.prepareNext(scalar) |
| 63 | e.out = append(e.out, "null"...) |
| 64 | } |
| 65 | |
| 66 | // WriteBool writes out the given boolean value. |
| 67 | func (e *Encoder) WriteBool(b bool) { |
| 68 | e.prepareNext(scalar) |
| 69 | if b { |
| 70 | e.out = append(e.out, "true"...) |
| 71 | } else { |
| 72 | e.out = append(e.out, "false"...) |
| 73 | } |
| 74 | } |
| 75 | |
| 76 | // WriteString writes out the given string in JSON string value. Returns error |
| 77 | // if input string contains invalid UTF-8. |
| 78 | func (e *Encoder) WriteString(s string) error { |
| 79 | e.prepareNext(scalar) |
| 80 | var err error |
| 81 | if e.out, err = appendString(e.out, s); err != nil { |
| 82 | return err |
| 83 | } |
| 84 | return nil |
| 85 | } |
| 86 | |
| 87 | // Sentinel error used for indicating invalid UTF-8. |
| 88 | var errInvalidUTF8 = errors.New("invalid UTF-8") |
| 89 | |
| 90 | func appendString(out []byte, in string) ([]byte, error) { |
| 91 | out = append(out, '"') |
| 92 | i := indexNeedEscapeInString(in) |
| 93 | in, out = in[i:], append(out, in[:i]...) |
| 94 | for len(in) > 0 { |
| 95 | switch r, n := utf8.DecodeRuneInString(in); { |
| 96 | case r == utf8.RuneError && n == 1: |
| 97 | return out, errInvalidUTF8 |
| 98 | case r < ' ' || r == '"' || r == '\\': |
| 99 | out = append(out, '\\') |
| 100 | switch r { |
| 101 | case '"', '\\': |
| 102 | out = append(out, byte(r)) |
| 103 | case '\b': |
| 104 | out = append(out, 'b') |
| 105 | case '\f': |
| 106 | out = append(out, 'f') |
| 107 | case '\n': |
| 108 | out = append(out, 'n') |
| 109 | case '\r': |
| 110 | out = append(out, 'r') |
| 111 | case '\t': |
| 112 | out = append(out, 't') |
| 113 | default: |
| 114 | out = append(out, 'u') |
| 115 | out = append(out, "0000"[1+(bits.Len32(uint32(r))-1)/4:]...) |
| 116 | out = strconv.AppendUint(out, uint64(r), 16) |
| 117 | } |
| 118 | in = in[n:] |
| 119 | default: |
| 120 | i := indexNeedEscapeInString(in[n:]) |
| 121 | in, out = in[n+i:], append(out, in[:n+i]...) |
| 122 | } |
| 123 | } |
| 124 | out = append(out, '"') |
| 125 | return out, nil |
| 126 | } |
| 127 | |
| 128 | // indexNeedEscapeInString returns the index of the character that needs |
| 129 | // escaping. If no characters need escaping, this returns the input length. |
| 130 | func indexNeedEscapeInString(s string) int { |
| 131 | for i, r := range s { |
| 132 | if r < ' ' || r == '\\' || r == '"' || r == utf8.RuneError { |
| 133 | return i |
| 134 | } |
| 135 | } |
| 136 | return len(s) |
| 137 | } |
| 138 | |
| 139 | // WriteFloat writes out the given float and bitSize in JSON number value. |
| 140 | func (e *Encoder) WriteFloat(n float64, bitSize int) { |
| 141 | e.prepareNext(scalar) |
| 142 | e.out = appendFloat(e.out, n, bitSize) |
| 143 | } |
| 144 | |
| 145 | // appendFloat formats given float in bitSize, and appends to the given []byte. |
| 146 | func appendFloat(out []byte, n float64, bitSize int) []byte { |
| 147 | switch { |
| 148 | case math.IsNaN(n): |
| 149 | return append(out, `"NaN"`...) |
| 150 | case math.IsInf(n, +1): |
| 151 | return append(out, `"Infinity"`...) |
| 152 | case math.IsInf(n, -1): |
| 153 | return append(out, `"-Infinity"`...) |
| 154 | } |
| 155 | |
| 156 | // JSON number formatting logic based on encoding/json. |
| 157 | // See floatEncoder.encode for reference. |
| 158 | fmt := byte('f') |
| 159 | if abs := math.Abs(n); abs != 0 { |
| 160 | if bitSize == 64 && (abs < 1e-6 || abs >= 1e21) || |
| 161 | bitSize == 32 && (float32(abs) < 1e-6 || float32(abs) >= 1e21) { |
| 162 | fmt = 'e' |
| 163 | } |
| 164 | } |
| 165 | out = strconv.AppendFloat(out, n, fmt, -1, bitSize) |
| 166 | if fmt == 'e' { |
| 167 | n := len(out) |
| 168 | if n >= 4 && out[n-4] == 'e' && out[n-3] == '-' && out[n-2] == '0' { |
| 169 | out[n-2] = out[n-1] |
| 170 | out = out[:n-1] |
| 171 | } |
| 172 | } |
| 173 | return out |
| 174 | } |
| 175 | |
| 176 | // WriteInt writes out the given signed integer in JSON number value. |
| 177 | func (e *Encoder) WriteInt(n int64) { |
| 178 | e.prepareNext(scalar) |
| 179 | e.out = append(e.out, strconv.FormatInt(n, 10)...) |
| 180 | } |
| 181 | |
| 182 | // WriteUint writes out the given unsigned integer in JSON number value. |
| 183 | func (e *Encoder) WriteUint(n uint64) { |
| 184 | e.prepareNext(scalar) |
| 185 | e.out = append(e.out, strconv.FormatUint(n, 10)...) |
| 186 | } |
| 187 | |
| 188 | // StartObject writes out the '{' symbol. |
| 189 | func (e *Encoder) StartObject() { |
| 190 | e.prepareNext(objectOpen) |
| 191 | e.out = append(e.out, '{') |
| 192 | } |
| 193 | |
| 194 | // EndObject writes out the '}' symbol. |
| 195 | func (e *Encoder) EndObject() { |
| 196 | e.prepareNext(objectClose) |
| 197 | e.out = append(e.out, '}') |
| 198 | } |
| 199 | |
| 200 | // WriteName writes out the given string in JSON string value and the name |
| 201 | // separator ':'. Returns error if input string contains invalid UTF-8, which |
| 202 | // should not be likely as protobuf field names should be valid. |
| 203 | func (e *Encoder) WriteName(s string) error { |
| 204 | e.prepareNext(name) |
| 205 | var err error |
| 206 | // Append to output regardless of error. |
| 207 | e.out, err = appendString(e.out, s) |
| 208 | e.out = append(e.out, ':') |
| 209 | return err |
| 210 | } |
| 211 | |
| 212 | // StartArray writes out the '[' symbol. |
| 213 | func (e *Encoder) StartArray() { |
| 214 | e.prepareNext(arrayOpen) |
| 215 | e.out = append(e.out, '[') |
| 216 | } |
| 217 | |
| 218 | // EndArray writes out the ']' symbol. |
| 219 | func (e *Encoder) EndArray() { |
| 220 | e.prepareNext(arrayClose) |
| 221 | e.out = append(e.out, ']') |
| 222 | } |
| 223 | |
| 224 | // prepareNext adds possible comma and indentation for the next value based |
| 225 | // on last type and indent option. It also updates lastKind to next. |
| 226 | func (e *Encoder) prepareNext(next kind) { |
| 227 | defer func() { |
| 228 | // Set lastKind to next. |
| 229 | e.lastKind = next |
| 230 | }() |
| 231 | |
| 232 | if len(e.indent) == 0 { |
| 233 | // Need to add comma on the following condition. |
| 234 | if e.lastKind&(scalar|objectClose|arrayClose) != 0 && |
| 235 | next&(name|scalar|objectOpen|arrayOpen) != 0 { |
| 236 | e.out = append(e.out, ',') |
| 237 | // For single-line output, add a random extra space after each |
| 238 | // comma to make output unstable. |
| 239 | if detrand.Bool() { |
| 240 | e.out = append(e.out, ' ') |
| 241 | } |
| 242 | } |
| 243 | return |
| 244 | } |
| 245 | |
| 246 | switch { |
| 247 | case e.lastKind&(objectOpen|arrayOpen) != 0: |
| 248 | // If next type is NOT closing, add indent and newline. |
| 249 | if next&(objectClose|arrayClose) == 0 { |
| 250 | e.indents = append(e.indents, e.indent...) |
| 251 | e.out = append(e.out, '\n') |
| 252 | e.out = append(e.out, e.indents...) |
| 253 | } |
| 254 | |
| 255 | case e.lastKind&(scalar|objectClose|arrayClose) != 0: |
| 256 | switch { |
| 257 | // If next type is either a value or name, add comma and newline. |
| 258 | case next&(name|scalar|objectOpen|arrayOpen) != 0: |
| 259 | e.out = append(e.out, ',', '\n') |
| 260 | |
| 261 | // If next type is a closing object or array, adjust indentation. |
| 262 | case next&(objectClose|arrayClose) != 0: |
| 263 | e.indents = e.indents[:len(e.indents)-len(e.indent)] |
| 264 | e.out = append(e.out, '\n') |
| 265 | } |
| 266 | e.out = append(e.out, e.indents...) |
| 267 | |
| 268 | case e.lastKind&name != 0: |
| 269 | e.out = append(e.out, ' ') |
| 270 | // For multi-line output, add a random extra space after key: to make |
| 271 | // output unstable. |
| 272 | if detrand.Bool() { |
| 273 | e.out = append(e.out, ' ') |
| 274 | } |
| 275 | } |
| 276 | } |