David K. Bainbridge | e05cf0c | 2021-08-19 03:16:50 +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 text |
| 6 | |
| 7 | // parseNumberValue parses a number from the input and returns a Token object. |
| 8 | func (d *Decoder) parseNumberValue() (Token, bool) { |
| 9 | in := d.in |
| 10 | num := parseNumber(in) |
| 11 | if num.size == 0 { |
| 12 | return Token{}, false |
| 13 | } |
| 14 | numAttrs := num.kind |
| 15 | if num.neg { |
| 16 | numAttrs |= isNegative |
| 17 | } |
| 18 | strSize := num.size |
| 19 | last := num.size - 1 |
| 20 | if num.kind == numFloat && (d.in[last] == 'f' || d.in[last] == 'F') { |
| 21 | strSize = last |
| 22 | } |
| 23 | tok := Token{ |
| 24 | kind: Scalar, |
| 25 | attrs: numberValue, |
| 26 | pos: len(d.orig) - len(d.in), |
| 27 | raw: d.in[:num.size], |
| 28 | str: string(d.in[:strSize]), |
| 29 | numAttrs: numAttrs, |
| 30 | } |
| 31 | d.consume(num.size) |
| 32 | return tok, true |
| 33 | } |
| 34 | |
| 35 | const ( |
| 36 | numDec uint8 = (1 << iota) / 2 |
| 37 | numHex |
| 38 | numOct |
| 39 | numFloat |
| 40 | ) |
| 41 | |
| 42 | // number is the result of parsing out a valid number from parseNumber. It |
| 43 | // contains data for doing float or integer conversion via the strconv package |
| 44 | // in conjunction with the input bytes. |
| 45 | type number struct { |
| 46 | kind uint8 |
| 47 | neg bool |
| 48 | size int |
| 49 | } |
| 50 | |
| 51 | // parseNumber constructs a number object from given input. It allows for the |
| 52 | // following patterns: |
| 53 | // integer: ^-?([1-9][0-9]*|0[xX][0-9a-fA-F]+|0[0-7]*) |
| 54 | // float: ^-?((0|[1-9][0-9]*)?([.][0-9]*)?([eE][+-]?[0-9]+)?[fF]?) |
| 55 | // It also returns the number of parsed bytes for the given number, 0 if it is |
| 56 | // not a number. |
| 57 | func parseNumber(input []byte) number { |
| 58 | kind := numDec |
| 59 | var size int |
| 60 | var neg bool |
| 61 | |
| 62 | s := input |
| 63 | if len(s) == 0 { |
| 64 | return number{} |
| 65 | } |
| 66 | |
| 67 | // Optional - |
| 68 | if s[0] == '-' { |
| 69 | neg = true |
| 70 | s = s[1:] |
| 71 | size++ |
| 72 | if len(s) == 0 { |
| 73 | return number{} |
| 74 | } |
| 75 | } |
| 76 | |
| 77 | // C++ allows for whitespace and comments in between the negative sign and |
| 78 | // the rest of the number. This logic currently does not but is consistent |
| 79 | // with v1. |
| 80 | |
| 81 | switch { |
| 82 | case s[0] == '0': |
| 83 | if len(s) > 1 { |
| 84 | switch { |
| 85 | case s[1] == 'x' || s[1] == 'X': |
| 86 | // Parse as hex number. |
| 87 | kind = numHex |
| 88 | n := 2 |
| 89 | s = s[2:] |
| 90 | for len(s) > 0 && (('0' <= s[0] && s[0] <= '9') || |
| 91 | ('a' <= s[0] && s[0] <= 'f') || |
| 92 | ('A' <= s[0] && s[0] <= 'F')) { |
| 93 | s = s[1:] |
| 94 | n++ |
| 95 | } |
| 96 | if n == 2 { |
| 97 | return number{} |
| 98 | } |
| 99 | size += n |
| 100 | |
| 101 | case '0' <= s[1] && s[1] <= '7': |
| 102 | // Parse as octal number. |
| 103 | kind = numOct |
| 104 | n := 2 |
| 105 | s = s[2:] |
| 106 | for len(s) > 0 && '0' <= s[0] && s[0] <= '7' { |
| 107 | s = s[1:] |
| 108 | n++ |
| 109 | } |
| 110 | size += n |
| 111 | } |
| 112 | |
| 113 | if kind&(numHex|numOct) > 0 { |
| 114 | if len(s) > 0 && !isDelim(s[0]) { |
| 115 | return number{} |
| 116 | } |
| 117 | return number{kind: kind, neg: neg, size: size} |
| 118 | } |
| 119 | } |
| 120 | s = s[1:] |
| 121 | size++ |
| 122 | |
| 123 | case '1' <= s[0] && s[0] <= '9': |
| 124 | n := 1 |
| 125 | s = s[1:] |
| 126 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 127 | s = s[1:] |
| 128 | n++ |
| 129 | } |
| 130 | size += n |
| 131 | |
| 132 | case s[0] == '.': |
| 133 | // Set kind to numFloat to signify the intent to parse as float. And |
| 134 | // that it needs to have other digits after '.'. |
| 135 | kind = numFloat |
| 136 | |
| 137 | default: |
| 138 | return number{} |
| 139 | } |
| 140 | |
| 141 | // . followed by 0 or more digits. |
| 142 | if len(s) > 0 && s[0] == '.' { |
| 143 | n := 1 |
| 144 | s = s[1:] |
| 145 | // If decimal point was before any digits, it should be followed by |
| 146 | // other digits. |
| 147 | if len(s) == 0 && kind == numFloat { |
| 148 | return number{} |
| 149 | } |
| 150 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 151 | s = s[1:] |
| 152 | n++ |
| 153 | } |
| 154 | size += n |
| 155 | kind = numFloat |
| 156 | } |
| 157 | |
| 158 | // e or E followed by an optional - or + and 1 or more digits. |
| 159 | if len(s) >= 2 && (s[0] == 'e' || s[0] == 'E') { |
| 160 | kind = numFloat |
| 161 | s = s[1:] |
| 162 | n := 1 |
| 163 | if s[0] == '+' || s[0] == '-' { |
| 164 | s = s[1:] |
| 165 | n++ |
| 166 | if len(s) == 0 { |
| 167 | return number{} |
| 168 | } |
| 169 | } |
| 170 | for len(s) > 0 && '0' <= s[0] && s[0] <= '9' { |
| 171 | s = s[1:] |
| 172 | n++ |
| 173 | } |
| 174 | size += n |
| 175 | } |
| 176 | |
| 177 | // Optional suffix f or F for floats. |
| 178 | if len(s) > 0 && (s[0] == 'f' || s[0] == 'F') { |
| 179 | kind = numFloat |
| 180 | s = s[1:] |
| 181 | size++ |
| 182 | } |
| 183 | |
| 184 | // Check that next byte is a delimiter or it is at the end. |
| 185 | if len(s) > 0 && !isDelim(s[0]) { |
| 186 | return number{} |
| 187 | } |
| 188 | |
| 189 | return number{kind: kind, neg: neg, size: size} |
| 190 | } |