David K. Bainbridge | bd6b288 | 2021-08-26 13:31:02 +0000 | [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 | "bytes" |
| 9 | "strconv" |
| 10 | ) |
| 11 | |
| 12 | // parseNumber reads the given []byte for a valid JSON number. If it is valid, |
| 13 | // it returns the number of bytes. Parsing logic follows the definition in |
| 14 | // https://tools.ietf.org/html/rfc7159#section-6, and is based off |
| 15 | // encoding/json.isValidNumber function. |
| 16 | func parseNumber(input []byte) (int, bool) { |
| 17 | var n int |
| 18 | |
| 19 | s := input |
| 20 | if len(s) == 0 { |
| 21 | return 0, false |
| 22 | } |
| 23 | |
| 24 | // Optional - |
| 25 | if s[0] == '-' { |
| 26 | s = s[1:] |
| 27 | n++ |
| 28 | if len(s) == 0 { |
| 29 | return 0, false |
| 30 | } |
| 31 | } |
| 32 | |
| 33 | // Digits |
| 34 | switch { |
| 35 | case s[0] == '0': |
| 36 | s = s[1:] |
| 37 | n++ |
| 38 | |
| 39 | case '1' <= s[0] && s[0] <= '9': |
| 40 | s = s[1:] |
| 41 | n++ |
| 42 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 43 | s = s[1:] |
| 44 | n++ |
| 45 | } |
| 46 | |
| 47 | default: |
| 48 | return 0, false |
| 49 | } |
| 50 | |
| 51 | // . followed by 1 or more digits. |
| 52 | if len(s) >= 2 && s[0] == '.' && '0' <= s[1] && s[1] <= '9' { |
| 53 | s = s[2:] |
| 54 | n += 2 |
| 55 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 56 | s = s[1:] |
| 57 | n++ |
| 58 | } |
| 59 | } |
| 60 | |
| 61 | // e or E followed by an optional - or + and |
| 62 | // 1 or more digits. |
| 63 | if len(s) >= 2 && (s[0] == 'e' || s[0] == 'E') { |
| 64 | s = s[1:] |
| 65 | n++ |
| 66 | if s[0] == '+' || s[0] == '-' { |
| 67 | s = s[1:] |
| 68 | n++ |
| 69 | if len(s) == 0 { |
| 70 | return 0, false |
| 71 | } |
| 72 | } |
| 73 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 74 | s = s[1:] |
| 75 | n++ |
| 76 | } |
| 77 | } |
| 78 | |
| 79 | // Check that next byte is a delimiter or it is at the end. |
| 80 | if n < len(input) && isNotDelim(input[n]) { |
| 81 | return 0, false |
| 82 | } |
| 83 | |
| 84 | return n, true |
| 85 | } |
| 86 | |
| 87 | // numberParts is the result of parsing out a valid JSON number. It contains |
| 88 | // the parts of a number. The parts are used for integer conversion. |
| 89 | type numberParts struct { |
| 90 | neg bool |
| 91 | intp []byte |
| 92 | frac []byte |
| 93 | exp []byte |
| 94 | } |
| 95 | |
| 96 | // parseNumber constructs numberParts from given []byte. The logic here is |
| 97 | // similar to consumeNumber above with the difference of having to construct |
| 98 | // numberParts. The slice fields in numberParts are subslices of the input. |
| 99 | func parseNumberParts(input []byte) (numberParts, bool) { |
| 100 | var neg bool |
| 101 | var intp []byte |
| 102 | var frac []byte |
| 103 | var exp []byte |
| 104 | |
| 105 | s := input |
| 106 | if len(s) == 0 { |
| 107 | return numberParts{}, false |
| 108 | } |
| 109 | |
| 110 | // Optional - |
| 111 | if s[0] == '-' { |
| 112 | neg = true |
| 113 | s = s[1:] |
| 114 | if len(s) == 0 { |
| 115 | return numberParts{}, false |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | // Digits |
| 120 | switch { |
| 121 | case s[0] == '0': |
| 122 | // Skip first 0 and no need to store. |
| 123 | s = s[1:] |
| 124 | |
| 125 | case '1' <= s[0] && s[0] <= '9': |
| 126 | intp = s |
| 127 | n := 1 |
| 128 | s = s[1:] |
| 129 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 130 | s = s[1:] |
| 131 | n++ |
| 132 | } |
| 133 | intp = intp[:n] |
| 134 | |
| 135 | default: |
| 136 | return numberParts{}, false |
| 137 | } |
| 138 | |
| 139 | // . followed by 1 or more digits. |
| 140 | if len(s) >= 2 && s[0] == '.' && '0' <= s[1] && s[1] <= '9' { |
| 141 | frac = s[1:] |
| 142 | n := 1 |
| 143 | s = s[2:] |
| 144 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 145 | s = s[1:] |
| 146 | n++ |
| 147 | } |
| 148 | frac = frac[:n] |
| 149 | } |
| 150 | |
| 151 | // e or E followed by an optional - or + and |
| 152 | // 1 or more digits. |
| 153 | if len(s) >= 2 && (s[0] == 'e' || s[0] == 'E') { |
| 154 | s = s[1:] |
| 155 | exp = s |
| 156 | n := 0 |
| 157 | if s[0] == '+' || s[0] == '-' { |
| 158 | s = s[1:] |
| 159 | n++ |
| 160 | if len(s) == 0 { |
| 161 | return numberParts{}, false |
| 162 | } |
| 163 | } |
| 164 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 165 | s = s[1:] |
| 166 | n++ |
| 167 | } |
| 168 | exp = exp[:n] |
| 169 | } |
| 170 | |
| 171 | return numberParts{ |
| 172 | neg: neg, |
| 173 | intp: intp, |
| 174 | frac: bytes.TrimRight(frac, "0"), // Remove unnecessary 0s to the right. |
| 175 | exp: exp, |
| 176 | }, true |
| 177 | } |
| 178 | |
| 179 | // normalizeToIntString returns an integer string in normal form without the |
| 180 | // E-notation for given numberParts. It will return false if it is not an |
| 181 | // integer or if the exponent exceeds than max/min int value. |
| 182 | func normalizeToIntString(n numberParts) (string, bool) { |
| 183 | intpSize := len(n.intp) |
| 184 | fracSize := len(n.frac) |
| 185 | |
| 186 | if intpSize == 0 && fracSize == 0 { |
| 187 | return "0", true |
| 188 | } |
| 189 | |
| 190 | var exp int |
| 191 | if len(n.exp) > 0 { |
| 192 | i, err := strconv.ParseInt(string(n.exp), 10, 32) |
| 193 | if err != nil { |
| 194 | return "", false |
| 195 | } |
| 196 | exp = int(i) |
| 197 | } |
| 198 | |
| 199 | var num []byte |
| 200 | if exp >= 0 { |
| 201 | // For positive E, shift fraction digits into integer part and also pad |
| 202 | // with zeroes as needed. |
| 203 | |
| 204 | // If there are more digits in fraction than the E value, then the |
| 205 | // number is not an integer. |
| 206 | if fracSize > exp { |
| 207 | return "", false |
| 208 | } |
| 209 | |
| 210 | // Make sure resulting digits are within max value limit to avoid |
| 211 | // unnecessarily constructing a large byte slice that may simply fail |
| 212 | // later on. |
| 213 | const maxDigits = 20 // Max uint64 value has 20 decimal digits. |
| 214 | if intpSize+exp > maxDigits { |
| 215 | return "", false |
| 216 | } |
| 217 | |
| 218 | // Set cap to make a copy of integer part when appended. |
| 219 | num = n.intp[:len(n.intp):len(n.intp)] |
| 220 | num = append(num, n.frac...) |
| 221 | for i := 0; i < exp-fracSize; i++ { |
| 222 | num = append(num, '0') |
| 223 | } |
| 224 | } else { |
| 225 | // For negative E, shift digits in integer part out. |
| 226 | |
| 227 | // If there are fractions, then the number is not an integer. |
| 228 | if fracSize > 0 { |
| 229 | return "", false |
| 230 | } |
| 231 | |
| 232 | // index is where the decimal point will be after adjusting for negative |
| 233 | // exponent. |
| 234 | index := intpSize + exp |
| 235 | if index < 0 { |
| 236 | return "", false |
| 237 | } |
| 238 | |
| 239 | num = n.intp |
| 240 | // If any of the digits being shifted to the right of the decimal point |
| 241 | // is non-zero, then the number is not an integer. |
| 242 | for i := index; i < intpSize; i++ { |
| 243 | if num[i] != '0' { |
| 244 | return "", false |
| 245 | } |
| 246 | } |
| 247 | num = num[:index] |
| 248 | } |
| 249 | |
| 250 | if n.neg { |
| 251 | return "-" + string(num), true |
| 252 | } |
| 253 | return string(num), true |
| 254 | } |