khenaidoo | 59ce9dd | 2019-11-11 13:05:32 -0500 | [diff] [blame] | 1 | package jsoniter |
| 2 | |
| 3 | import ( |
| 4 | "encoding/json" |
| 5 | "io" |
| 6 | "math/big" |
| 7 | "strconv" |
| 8 | "strings" |
| 9 | "unsafe" |
| 10 | ) |
| 11 | |
| 12 | var floatDigits []int8 |
| 13 | |
| 14 | const invalidCharForNumber = int8(-1) |
| 15 | const endOfNumber = int8(-2) |
| 16 | const dotInNumber = int8(-3) |
| 17 | |
| 18 | func init() { |
| 19 | floatDigits = make([]int8, 256) |
| 20 | for i := 0; i < len(floatDigits); i++ { |
| 21 | floatDigits[i] = invalidCharForNumber |
| 22 | } |
| 23 | for i := int8('0'); i <= int8('9'); i++ { |
| 24 | floatDigits[i] = i - int8('0') |
| 25 | } |
| 26 | floatDigits[','] = endOfNumber |
| 27 | floatDigits[']'] = endOfNumber |
| 28 | floatDigits['}'] = endOfNumber |
| 29 | floatDigits[' '] = endOfNumber |
| 30 | floatDigits['\t'] = endOfNumber |
| 31 | floatDigits['\n'] = endOfNumber |
| 32 | floatDigits['.'] = dotInNumber |
| 33 | } |
| 34 | |
| 35 | // ReadBigFloat read big.Float |
| 36 | func (iter *Iterator) ReadBigFloat() (ret *big.Float) { |
| 37 | str := iter.readNumberAsString() |
| 38 | if iter.Error != nil && iter.Error != io.EOF { |
| 39 | return nil |
| 40 | } |
| 41 | prec := 64 |
| 42 | if len(str) > prec { |
| 43 | prec = len(str) |
| 44 | } |
| 45 | val, _, err := big.ParseFloat(str, 10, uint(prec), big.ToZero) |
| 46 | if err != nil { |
| 47 | iter.Error = err |
| 48 | return nil |
| 49 | } |
| 50 | return val |
| 51 | } |
| 52 | |
| 53 | // ReadBigInt read big.Int |
| 54 | func (iter *Iterator) ReadBigInt() (ret *big.Int) { |
| 55 | str := iter.readNumberAsString() |
| 56 | if iter.Error != nil && iter.Error != io.EOF { |
| 57 | return nil |
| 58 | } |
| 59 | ret = big.NewInt(0) |
| 60 | var success bool |
| 61 | ret, success = ret.SetString(str, 10) |
| 62 | if !success { |
| 63 | iter.ReportError("ReadBigInt", "invalid big int") |
| 64 | return nil |
| 65 | } |
| 66 | return ret |
| 67 | } |
| 68 | |
| 69 | //ReadFloat32 read float32 |
| 70 | func (iter *Iterator) ReadFloat32() (ret float32) { |
| 71 | c := iter.nextToken() |
| 72 | if c == '-' { |
| 73 | return -iter.readPositiveFloat32() |
| 74 | } |
| 75 | iter.unreadByte() |
| 76 | return iter.readPositiveFloat32() |
| 77 | } |
| 78 | |
| 79 | func (iter *Iterator) readPositiveFloat32() (ret float32) { |
| 80 | i := iter.head |
| 81 | // first char |
| 82 | if i == iter.tail { |
| 83 | return iter.readFloat32SlowPath() |
| 84 | } |
| 85 | c := iter.buf[i] |
| 86 | i++ |
| 87 | ind := floatDigits[c] |
| 88 | switch ind { |
| 89 | case invalidCharForNumber: |
| 90 | return iter.readFloat32SlowPath() |
| 91 | case endOfNumber: |
| 92 | iter.ReportError("readFloat32", "empty number") |
| 93 | return |
| 94 | case dotInNumber: |
| 95 | iter.ReportError("readFloat32", "leading dot is invalid") |
| 96 | return |
| 97 | case 0: |
| 98 | if i == iter.tail { |
| 99 | return iter.readFloat32SlowPath() |
| 100 | } |
| 101 | c = iter.buf[i] |
| 102 | switch c { |
| 103 | case '0', '1', '2', '3', '4', '5', '6', '7', '8', '9': |
| 104 | iter.ReportError("readFloat32", "leading zero is invalid") |
| 105 | return |
| 106 | } |
| 107 | } |
| 108 | value := uint64(ind) |
| 109 | // chars before dot |
| 110 | non_decimal_loop: |
| 111 | for ; i < iter.tail; i++ { |
| 112 | c = iter.buf[i] |
| 113 | ind := floatDigits[c] |
| 114 | switch ind { |
| 115 | case invalidCharForNumber: |
| 116 | return iter.readFloat32SlowPath() |
| 117 | case endOfNumber: |
| 118 | iter.head = i |
| 119 | return float32(value) |
| 120 | case dotInNumber: |
| 121 | break non_decimal_loop |
| 122 | } |
| 123 | if value > uint64SafeToMultiple10 { |
| 124 | return iter.readFloat32SlowPath() |
| 125 | } |
| 126 | value = (value << 3) + (value << 1) + uint64(ind) // value = value * 10 + ind; |
| 127 | } |
| 128 | // chars after dot |
| 129 | if c == '.' { |
| 130 | i++ |
| 131 | decimalPlaces := 0 |
| 132 | if i == iter.tail { |
| 133 | return iter.readFloat32SlowPath() |
| 134 | } |
| 135 | for ; i < iter.tail; i++ { |
| 136 | c = iter.buf[i] |
| 137 | ind := floatDigits[c] |
| 138 | switch ind { |
| 139 | case endOfNumber: |
| 140 | if decimalPlaces > 0 && decimalPlaces < len(pow10) { |
| 141 | iter.head = i |
| 142 | return float32(float64(value) / float64(pow10[decimalPlaces])) |
| 143 | } |
| 144 | // too many decimal places |
| 145 | return iter.readFloat32SlowPath() |
| 146 | case invalidCharForNumber, dotInNumber: |
| 147 | return iter.readFloat32SlowPath() |
| 148 | } |
| 149 | decimalPlaces++ |
| 150 | if value > uint64SafeToMultiple10 { |
| 151 | return iter.readFloat32SlowPath() |
| 152 | } |
| 153 | value = (value << 3) + (value << 1) + uint64(ind) |
| 154 | } |
| 155 | } |
| 156 | return iter.readFloat32SlowPath() |
| 157 | } |
| 158 | |
| 159 | func (iter *Iterator) readNumberAsString() (ret string) { |
| 160 | strBuf := [16]byte{} |
| 161 | str := strBuf[0:0] |
| 162 | load_loop: |
| 163 | for { |
| 164 | for i := iter.head; i < iter.tail; i++ { |
| 165 | c := iter.buf[i] |
| 166 | switch c { |
| 167 | case '+', '-', '.', 'e', 'E', '0', '1', '2', '3', '4', '5', '6', '7', '8', '9': |
| 168 | str = append(str, c) |
| 169 | continue |
| 170 | default: |
| 171 | iter.head = i |
| 172 | break load_loop |
| 173 | } |
| 174 | } |
| 175 | if !iter.loadMore() { |
| 176 | break |
| 177 | } |
| 178 | } |
| 179 | if iter.Error != nil && iter.Error != io.EOF { |
| 180 | return |
| 181 | } |
| 182 | if len(str) == 0 { |
| 183 | iter.ReportError("readNumberAsString", "invalid number") |
| 184 | } |
| 185 | return *(*string)(unsafe.Pointer(&str)) |
| 186 | } |
| 187 | |
| 188 | func (iter *Iterator) readFloat32SlowPath() (ret float32) { |
| 189 | str := iter.readNumberAsString() |
| 190 | if iter.Error != nil && iter.Error != io.EOF { |
| 191 | return |
| 192 | } |
| 193 | errMsg := validateFloat(str) |
| 194 | if errMsg != "" { |
| 195 | iter.ReportError("readFloat32SlowPath", errMsg) |
| 196 | return |
| 197 | } |
| 198 | val, err := strconv.ParseFloat(str, 32) |
| 199 | if err != nil { |
| 200 | iter.Error = err |
| 201 | return |
| 202 | } |
| 203 | return float32(val) |
| 204 | } |
| 205 | |
| 206 | // ReadFloat64 read float64 |
| 207 | func (iter *Iterator) ReadFloat64() (ret float64) { |
| 208 | c := iter.nextToken() |
| 209 | if c == '-' { |
| 210 | return -iter.readPositiveFloat64() |
| 211 | } |
| 212 | iter.unreadByte() |
| 213 | return iter.readPositiveFloat64() |
| 214 | } |
| 215 | |
| 216 | func (iter *Iterator) readPositiveFloat64() (ret float64) { |
| 217 | i := iter.head |
| 218 | // first char |
| 219 | if i == iter.tail { |
| 220 | return iter.readFloat64SlowPath() |
| 221 | } |
| 222 | c := iter.buf[i] |
| 223 | i++ |
| 224 | ind := floatDigits[c] |
| 225 | switch ind { |
| 226 | case invalidCharForNumber: |
| 227 | return iter.readFloat64SlowPath() |
| 228 | case endOfNumber: |
| 229 | iter.ReportError("readFloat64", "empty number") |
| 230 | return |
| 231 | case dotInNumber: |
| 232 | iter.ReportError("readFloat64", "leading dot is invalid") |
| 233 | return |
| 234 | case 0: |
| 235 | if i == iter.tail { |
| 236 | return iter.readFloat64SlowPath() |
| 237 | } |
| 238 | c = iter.buf[i] |
| 239 | switch c { |
| 240 | case '0', '1', '2', '3', '4', '5', '6', '7', '8', '9': |
| 241 | iter.ReportError("readFloat64", "leading zero is invalid") |
| 242 | return |
| 243 | } |
| 244 | } |
| 245 | value := uint64(ind) |
| 246 | // chars before dot |
| 247 | non_decimal_loop: |
| 248 | for ; i < iter.tail; i++ { |
| 249 | c = iter.buf[i] |
| 250 | ind := floatDigits[c] |
| 251 | switch ind { |
| 252 | case invalidCharForNumber: |
| 253 | return iter.readFloat64SlowPath() |
| 254 | case endOfNumber: |
| 255 | iter.head = i |
| 256 | return float64(value) |
| 257 | case dotInNumber: |
| 258 | break non_decimal_loop |
| 259 | } |
| 260 | if value > uint64SafeToMultiple10 { |
| 261 | return iter.readFloat64SlowPath() |
| 262 | } |
| 263 | value = (value << 3) + (value << 1) + uint64(ind) // value = value * 10 + ind; |
| 264 | } |
| 265 | // chars after dot |
| 266 | if c == '.' { |
| 267 | i++ |
| 268 | decimalPlaces := 0 |
| 269 | if i == iter.tail { |
| 270 | return iter.readFloat64SlowPath() |
| 271 | } |
| 272 | for ; i < iter.tail; i++ { |
| 273 | c = iter.buf[i] |
| 274 | ind := floatDigits[c] |
| 275 | switch ind { |
| 276 | case endOfNumber: |
| 277 | if decimalPlaces > 0 && decimalPlaces < len(pow10) { |
| 278 | iter.head = i |
| 279 | return float64(value) / float64(pow10[decimalPlaces]) |
| 280 | } |
| 281 | // too many decimal places |
| 282 | return iter.readFloat64SlowPath() |
| 283 | case invalidCharForNumber, dotInNumber: |
| 284 | return iter.readFloat64SlowPath() |
| 285 | } |
| 286 | decimalPlaces++ |
| 287 | if value > uint64SafeToMultiple10 { |
| 288 | return iter.readFloat64SlowPath() |
| 289 | } |
| 290 | value = (value << 3) + (value << 1) + uint64(ind) |
khenaidoo | 2672188 | 2021-08-11 17:42:52 -0400 | [diff] [blame] | 291 | if value > maxFloat64 { |
| 292 | return iter.readFloat64SlowPath() |
| 293 | } |
khenaidoo | 59ce9dd | 2019-11-11 13:05:32 -0500 | [diff] [blame] | 294 | } |
| 295 | } |
| 296 | return iter.readFloat64SlowPath() |
| 297 | } |
| 298 | |
| 299 | func (iter *Iterator) readFloat64SlowPath() (ret float64) { |
| 300 | str := iter.readNumberAsString() |
| 301 | if iter.Error != nil && iter.Error != io.EOF { |
| 302 | return |
| 303 | } |
| 304 | errMsg := validateFloat(str) |
| 305 | if errMsg != "" { |
| 306 | iter.ReportError("readFloat64SlowPath", errMsg) |
| 307 | return |
| 308 | } |
| 309 | val, err := strconv.ParseFloat(str, 64) |
| 310 | if err != nil { |
| 311 | iter.Error = err |
| 312 | return |
| 313 | } |
| 314 | return val |
| 315 | } |
| 316 | |
| 317 | func validateFloat(str string) string { |
| 318 | // strconv.ParseFloat is not validating `1.` or `1.e1` |
| 319 | if len(str) == 0 { |
| 320 | return "empty number" |
| 321 | } |
| 322 | if str[0] == '-' { |
| 323 | return "-- is not valid" |
| 324 | } |
| 325 | dotPos := strings.IndexByte(str, '.') |
| 326 | if dotPos != -1 { |
| 327 | if dotPos == len(str)-1 { |
| 328 | return "dot can not be last character" |
| 329 | } |
| 330 | switch str[dotPos+1] { |
| 331 | case '0', '1', '2', '3', '4', '5', '6', '7', '8', '9': |
| 332 | default: |
| 333 | return "missing digit after dot" |
| 334 | } |
| 335 | } |
| 336 | return "" |
| 337 | } |
| 338 | |
| 339 | // ReadNumber read json.Number |
| 340 | func (iter *Iterator) ReadNumber() (ret json.Number) { |
| 341 | return json.Number(iter.readNumberAsString()) |
| 342 | } |