William Kurkian | 6ea97f8 | 2019-03-13 15:51:55 -0400 | [diff] [blame^] | 1 | // Copyright 2012 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 colltab |
| 6 | |
| 7 | import "unicode/utf8" |
| 8 | |
| 9 | // For a description of ContractTrieSet, see text/collate/build/contract.go. |
| 10 | |
| 11 | type ContractTrieSet []struct{ L, H, N, I uint8 } |
| 12 | |
| 13 | // ctScanner is used to match a trie to an input sequence. |
| 14 | // A contraction may match a non-contiguous sequence of bytes in an input string. |
| 15 | // For example, if there is a contraction for <a, combining_ring>, it should match |
| 16 | // the sequence <a, combining_cedilla, combining_ring>, as combining_cedilla does |
| 17 | // not block combining_ring. |
| 18 | // ctScanner does not automatically skip over non-blocking non-starters, but rather |
| 19 | // retains the state of the last match and leaves it up to the user to continue |
| 20 | // the match at the appropriate points. |
| 21 | type ctScanner struct { |
| 22 | states ContractTrieSet |
| 23 | s []byte |
| 24 | n int |
| 25 | index int |
| 26 | pindex int |
| 27 | done bool |
| 28 | } |
| 29 | |
| 30 | type ctScannerString struct { |
| 31 | states ContractTrieSet |
| 32 | s string |
| 33 | n int |
| 34 | index int |
| 35 | pindex int |
| 36 | done bool |
| 37 | } |
| 38 | |
| 39 | func (t ContractTrieSet) scanner(index, n int, b []byte) ctScanner { |
| 40 | return ctScanner{s: b, states: t[index:], n: n} |
| 41 | } |
| 42 | |
| 43 | func (t ContractTrieSet) scannerString(index, n int, str string) ctScannerString { |
| 44 | return ctScannerString{s: str, states: t[index:], n: n} |
| 45 | } |
| 46 | |
| 47 | // result returns the offset i and bytes consumed p so far. If no suffix |
| 48 | // matched, i and p will be 0. |
| 49 | func (s *ctScanner) result() (i, p int) { |
| 50 | return s.index, s.pindex |
| 51 | } |
| 52 | |
| 53 | func (s *ctScannerString) result() (i, p int) { |
| 54 | return s.index, s.pindex |
| 55 | } |
| 56 | |
| 57 | const ( |
| 58 | final = 0 |
| 59 | noIndex = 0xFF |
| 60 | ) |
| 61 | |
| 62 | // scan matches the longest suffix at the current location in the input |
| 63 | // and returns the number of bytes consumed. |
| 64 | func (s *ctScanner) scan(p int) int { |
| 65 | pr := p // the p at the rune start |
| 66 | str := s.s |
| 67 | states, n := s.states, s.n |
| 68 | for i := 0; i < n && p < len(str); { |
| 69 | e := states[i] |
| 70 | c := str[p] |
| 71 | // TODO: a significant number of contractions are of a form that |
| 72 | // cannot match discontiguous UTF-8 in a normalized string. We could let |
| 73 | // a negative value of e.n mean that we can set s.done = true and avoid |
| 74 | // the need for additional matches. |
| 75 | if c >= e.L { |
| 76 | if e.L == c { |
| 77 | p++ |
| 78 | if e.I != noIndex { |
| 79 | s.index = int(e.I) |
| 80 | s.pindex = p |
| 81 | } |
| 82 | if e.N != final { |
| 83 | i, states, n = 0, states[int(e.H)+n:], int(e.N) |
| 84 | if p >= len(str) || utf8.RuneStart(str[p]) { |
| 85 | s.states, s.n, pr = states, n, p |
| 86 | } |
| 87 | } else { |
| 88 | s.done = true |
| 89 | return p |
| 90 | } |
| 91 | continue |
| 92 | } else if e.N == final && c <= e.H { |
| 93 | p++ |
| 94 | s.done = true |
| 95 | s.index = int(c-e.L) + int(e.I) |
| 96 | s.pindex = p |
| 97 | return p |
| 98 | } |
| 99 | } |
| 100 | i++ |
| 101 | } |
| 102 | return pr |
| 103 | } |
| 104 | |
| 105 | // scan is a verbatim copy of ctScanner.scan. |
| 106 | func (s *ctScannerString) scan(p int) int { |
| 107 | pr := p // the p at the rune start |
| 108 | str := s.s |
| 109 | states, n := s.states, s.n |
| 110 | for i := 0; i < n && p < len(str); { |
| 111 | e := states[i] |
| 112 | c := str[p] |
| 113 | // TODO: a significant number of contractions are of a form that |
| 114 | // cannot match discontiguous UTF-8 in a normalized string. We could let |
| 115 | // a negative value of e.n mean that we can set s.done = true and avoid |
| 116 | // the need for additional matches. |
| 117 | if c >= e.L { |
| 118 | if e.L == c { |
| 119 | p++ |
| 120 | if e.I != noIndex { |
| 121 | s.index = int(e.I) |
| 122 | s.pindex = p |
| 123 | } |
| 124 | if e.N != final { |
| 125 | i, states, n = 0, states[int(e.H)+n:], int(e.N) |
| 126 | if p >= len(str) || utf8.RuneStart(str[p]) { |
| 127 | s.states, s.n, pr = states, n, p |
| 128 | } |
| 129 | } else { |
| 130 | s.done = true |
| 131 | return p |
| 132 | } |
| 133 | continue |
| 134 | } else if e.N == final && c <= e.H { |
| 135 | p++ |
| 136 | s.done = true |
| 137 | s.index = int(c-e.L) + int(e.I) |
| 138 | s.pindex = p |
| 139 | return p |
| 140 | } |
| 141 | } |
| 142 | i++ |
| 143 | } |
| 144 | return pr |
| 145 | } |