Serge Bazanski | cc25bdf | 2018-10-25 14:02:58 +0200 | [diff] [blame] | 1 | // Copyright 2015 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 width |
| 6 | |
| 7 | import ( |
| 8 | "unicode/utf8" |
| 9 | |
| 10 | "golang.org/x/text/transform" |
| 11 | ) |
| 12 | |
| 13 | type foldTransform struct { |
| 14 | transform.NopResetter |
| 15 | } |
| 16 | |
| 17 | func (foldTransform) Span(src []byte, atEOF bool) (n int, err error) { |
| 18 | for n < len(src) { |
| 19 | if src[n] < utf8.RuneSelf { |
| 20 | // ASCII fast path. |
| 21 | for n++; n < len(src) && src[n] < utf8.RuneSelf; n++ { |
| 22 | } |
| 23 | continue |
| 24 | } |
| 25 | v, size := trie.lookup(src[n:]) |
| 26 | if size == 0 { // incomplete UTF-8 encoding |
| 27 | if !atEOF { |
| 28 | err = transform.ErrShortSrc |
| 29 | } else { |
| 30 | n = len(src) |
| 31 | } |
| 32 | break |
| 33 | } |
| 34 | if elem(v)&tagNeedsFold != 0 { |
| 35 | err = transform.ErrEndOfSpan |
| 36 | break |
| 37 | } |
| 38 | n += size |
| 39 | } |
| 40 | return n, err |
| 41 | } |
| 42 | |
| 43 | func (foldTransform) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) { |
| 44 | for nSrc < len(src) { |
| 45 | if src[nSrc] < utf8.RuneSelf { |
| 46 | // ASCII fast path. |
| 47 | start, end := nSrc, len(src) |
| 48 | if d := len(dst) - nDst; d < end-start { |
| 49 | end = nSrc + d |
| 50 | } |
| 51 | for nSrc++; nSrc < end && src[nSrc] < utf8.RuneSelf; nSrc++ { |
| 52 | } |
| 53 | n := copy(dst[nDst:], src[start:nSrc]) |
| 54 | if nDst += n; nDst == len(dst) { |
| 55 | nSrc = start + n |
| 56 | if nSrc == len(src) { |
| 57 | return nDst, nSrc, nil |
| 58 | } |
| 59 | if src[nSrc] < utf8.RuneSelf { |
| 60 | return nDst, nSrc, transform.ErrShortDst |
| 61 | } |
| 62 | } |
| 63 | continue |
| 64 | } |
| 65 | v, size := trie.lookup(src[nSrc:]) |
| 66 | if size == 0 { // incomplete UTF-8 encoding |
| 67 | if !atEOF { |
| 68 | return nDst, nSrc, transform.ErrShortSrc |
| 69 | } |
| 70 | size = 1 // gobble 1 byte |
| 71 | } |
| 72 | if elem(v)&tagNeedsFold == 0 { |
| 73 | if size != copy(dst[nDst:], src[nSrc:nSrc+size]) { |
| 74 | return nDst, nSrc, transform.ErrShortDst |
| 75 | } |
| 76 | nDst += size |
| 77 | } else { |
| 78 | data := inverseData[byte(v)] |
| 79 | if len(dst)-nDst < int(data[0]) { |
| 80 | return nDst, nSrc, transform.ErrShortDst |
| 81 | } |
| 82 | i := 1 |
| 83 | for end := int(data[0]); i < end; i++ { |
| 84 | dst[nDst] = data[i] |
| 85 | nDst++ |
| 86 | } |
| 87 | dst[nDst] = data[i] ^ src[nSrc+size-1] |
| 88 | nDst++ |
| 89 | } |
| 90 | nSrc += size |
| 91 | } |
| 92 | return nDst, nSrc, nil |
| 93 | } |
| 94 | |
| 95 | type narrowTransform struct { |
| 96 | transform.NopResetter |
| 97 | } |
| 98 | |
| 99 | func (narrowTransform) Span(src []byte, atEOF bool) (n int, err error) { |
| 100 | for n < len(src) { |
| 101 | if src[n] < utf8.RuneSelf { |
| 102 | // ASCII fast path. |
| 103 | for n++; n < len(src) && src[n] < utf8.RuneSelf; n++ { |
| 104 | } |
| 105 | continue |
| 106 | } |
| 107 | v, size := trie.lookup(src[n:]) |
| 108 | if size == 0 { // incomplete UTF-8 encoding |
| 109 | if !atEOF { |
| 110 | err = transform.ErrShortSrc |
| 111 | } else { |
| 112 | n = len(src) |
| 113 | } |
| 114 | break |
| 115 | } |
| 116 | if k := elem(v).kind(); byte(v) == 0 || k != EastAsianFullwidth && k != EastAsianWide && k != EastAsianAmbiguous { |
| 117 | } else { |
| 118 | err = transform.ErrEndOfSpan |
| 119 | break |
| 120 | } |
| 121 | n += size |
| 122 | } |
| 123 | return n, err |
| 124 | } |
| 125 | |
| 126 | func (narrowTransform) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) { |
| 127 | for nSrc < len(src) { |
| 128 | if src[nSrc] < utf8.RuneSelf { |
| 129 | // ASCII fast path. |
| 130 | start, end := nSrc, len(src) |
| 131 | if d := len(dst) - nDst; d < end-start { |
| 132 | end = nSrc + d |
| 133 | } |
| 134 | for nSrc++; nSrc < end && src[nSrc] < utf8.RuneSelf; nSrc++ { |
| 135 | } |
| 136 | n := copy(dst[nDst:], src[start:nSrc]) |
| 137 | if nDst += n; nDst == len(dst) { |
| 138 | nSrc = start + n |
| 139 | if nSrc == len(src) { |
| 140 | return nDst, nSrc, nil |
| 141 | } |
| 142 | if src[nSrc] < utf8.RuneSelf { |
| 143 | return nDst, nSrc, transform.ErrShortDst |
| 144 | } |
| 145 | } |
| 146 | continue |
| 147 | } |
| 148 | v, size := trie.lookup(src[nSrc:]) |
| 149 | if size == 0 { // incomplete UTF-8 encoding |
| 150 | if !atEOF { |
| 151 | return nDst, nSrc, transform.ErrShortSrc |
| 152 | } |
| 153 | size = 1 // gobble 1 byte |
| 154 | } |
| 155 | if k := elem(v).kind(); byte(v) == 0 || k != EastAsianFullwidth && k != EastAsianWide && k != EastAsianAmbiguous { |
| 156 | if size != copy(dst[nDst:], src[nSrc:nSrc+size]) { |
| 157 | return nDst, nSrc, transform.ErrShortDst |
| 158 | } |
| 159 | nDst += size |
| 160 | } else { |
| 161 | data := inverseData[byte(v)] |
| 162 | if len(dst)-nDst < int(data[0]) { |
| 163 | return nDst, nSrc, transform.ErrShortDst |
| 164 | } |
| 165 | i := 1 |
| 166 | for end := int(data[0]); i < end; i++ { |
| 167 | dst[nDst] = data[i] |
| 168 | nDst++ |
| 169 | } |
| 170 | dst[nDst] = data[i] ^ src[nSrc+size-1] |
| 171 | nDst++ |
| 172 | } |
| 173 | nSrc += size |
| 174 | } |
| 175 | return nDst, nSrc, nil |
| 176 | } |
| 177 | |
| 178 | type wideTransform struct { |
| 179 | transform.NopResetter |
| 180 | } |
| 181 | |
| 182 | func (wideTransform) Span(src []byte, atEOF bool) (n int, err error) { |
| 183 | for n < len(src) { |
| 184 | // TODO: Consider ASCII fast path. Special-casing ASCII handling can |
| 185 | // reduce the ns/op of BenchmarkWideASCII by about 30%. This is probably |
| 186 | // not enough to warrant the extra code and complexity. |
| 187 | v, size := trie.lookup(src[n:]) |
| 188 | if size == 0 { // incomplete UTF-8 encoding |
| 189 | if !atEOF { |
| 190 | err = transform.ErrShortSrc |
| 191 | } else { |
| 192 | n = len(src) |
| 193 | } |
| 194 | break |
| 195 | } |
| 196 | if k := elem(v).kind(); byte(v) == 0 || k != EastAsianHalfwidth && k != EastAsianNarrow { |
| 197 | } else { |
| 198 | err = transform.ErrEndOfSpan |
| 199 | break |
| 200 | } |
| 201 | n += size |
| 202 | } |
| 203 | return n, err |
| 204 | } |
| 205 | |
| 206 | func (wideTransform) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) { |
| 207 | for nSrc < len(src) { |
| 208 | // TODO: Consider ASCII fast path. Special-casing ASCII handling can |
| 209 | // reduce the ns/op of BenchmarkWideASCII by about 30%. This is probably |
| 210 | // not enough to warrant the extra code and complexity. |
| 211 | v, size := trie.lookup(src[nSrc:]) |
| 212 | if size == 0 { // incomplete UTF-8 encoding |
| 213 | if !atEOF { |
| 214 | return nDst, nSrc, transform.ErrShortSrc |
| 215 | } |
| 216 | size = 1 // gobble 1 byte |
| 217 | } |
| 218 | if k := elem(v).kind(); byte(v) == 0 || k != EastAsianHalfwidth && k != EastAsianNarrow { |
| 219 | if size != copy(dst[nDst:], src[nSrc:nSrc+size]) { |
| 220 | return nDst, nSrc, transform.ErrShortDst |
| 221 | } |
| 222 | nDst += size |
| 223 | } else { |
| 224 | data := inverseData[byte(v)] |
| 225 | if len(dst)-nDst < int(data[0]) { |
| 226 | return nDst, nSrc, transform.ErrShortDst |
| 227 | } |
| 228 | i := 1 |
| 229 | for end := int(data[0]); i < end; i++ { |
| 230 | dst[nDst] = data[i] |
| 231 | nDst++ |
| 232 | } |
| 233 | dst[nDst] = data[i] ^ src[nSrc+size-1] |
| 234 | nDst++ |
| 235 | } |
| 236 | nSrc += size |
| 237 | } |
| 238 | return nDst, nSrc, nil |
| 239 | } |