1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style license that can be 3 // found in the LICENSE file. 4 5 #ifndef BASE_I18N_BREAK_ITERATOR_H_ 6 #define BASE_I18N_BREAK_ITERATOR_H_ 7 8 #include "base/basictypes.h" 9 #include "base/i18n/base_i18n_export.h" 10 #include "base/strings/string16.h" 11 12 // The BreakIterator class iterates through the words, word breaks, and 13 // line breaks in a UTF-16 string. 14 // 15 // It provides several modes, BREAK_WORD, BREAK_LINE, and BREAK_NEWLINE, 16 // which modify how characters are aggregated into the returned string. 17 // 18 // Under BREAK_WORD mode, once a word is encountered any non-word 19 // characters are not included in the returned string (e.g. in the 20 // UTF-16 equivalent of the string " foo bar! ", the word breaks are at 21 // the periods in ". .foo. .bar.!. ."). 22 // Note that Chinese/Japanese/Thai do not use spaces between words so that 23 // boundaries can fall in the middle of a continuous run of non-space / 24 // non-punctuation characters. 25 // 26 // Under BREAK_LINE mode, once a line breaking opportunity is encountered, 27 // any non-word characters are included in the returned string, breaking 28 // only when a space-equivalent character or a line breaking opportunity 29 // is encountered (e.g. in the UTF16-equivalent of the string " foo bar! ", 30 // the breaks are at the periods in ". .foo .bar! ."). 31 // 32 // Note that lines can be broken at any character/syllable/grapheme cluster 33 // boundary in Chinese/Japanese/Korean and at word boundaries in Thai 34 // (Thai does not use spaces between words). Therefore, this is NOT the same 35 // as breaking only at space-equivalent characters where its former 36 // name (BREAK_SPACE) implied. 37 // 38 // Under BREAK_NEWLINE mode, all characters are included in the returned 39 // string, breaking only when a newline-equivalent character is encountered 40 // (eg. in the UTF-16 equivalent of the string "foo\nbar!\n\n", the line 41 // breaks are at the periods in ".foo\n.bar\n.\n."). 42 // 43 // To extract the words from a string, move a BREAK_WORD BreakIterator 44 // through the string and test whether IsWord() is true. E.g., 45 // BreakIterator iter(str, BreakIterator::BREAK_WORD); 46 // if (!iter.Init()) 47 // return false; 48 // while (iter.Advance()) { 49 // if (iter.IsWord()) { 50 // // Region [iter.prev(), iter.pos()) contains a word. 51 // VLOG(1) << "word: " << iter.GetString(); 52 // } 53 // } 54 55 namespace base { 56 namespace i18n { 57 58 class BASE_I18N_EXPORT BreakIterator { 59 public: 60 enum BreakType { 61 BREAK_WORD, 62 BREAK_LINE, 63 // TODO(jshin): Remove this after reviewing call sites. 64 // If call sites really need break only on space-like characters 65 // implement it separately. 66 BREAK_SPACE = BREAK_LINE, 67 BREAK_NEWLINE, 68 BREAK_CHARACTER, 69 // But don't remove this one! 70 RULE_BASED, 71 }; 72 73 // Requires |str| to live as long as the BreakIterator does. 74 BreakIterator(const string16& str, BreakType break_type); 75 // Make a rule-based iterator. BreakType == RULE_BASED is implied. 76 // TODO(andrewhayden): This signature could easily be misinterpreted as 77 // "(const string16& str, const string16& locale)". We should do something 78 // better. 79 BreakIterator(const string16& str, const string16& rules); 80 ~BreakIterator(); 81 82 // Init() must be called before any of the iterators are valid. 83 // Returns false if ICU failed to initialize. 84 bool Init(); 85 86 // Advance to the next break. Returns false if we've run past the end of 87 // the string. (Note that the very last "break" is after the final 88 // character in the string, and when we advance to that position it's the 89 // last time Advance() returns true.) 90 bool Advance(); 91 92 // Updates the text used by the iterator, resetting the iterator as if 93 // if Init() had been called again. Any old state is lost. Returns true 94 // unless there is an error setting the text. 95 bool SetText(const base::char16* text, const size_t length); 96 97 // Under BREAK_WORD mode, returns true if the break we just hit is the 98 // end of a word. (Otherwise, the break iterator just skipped over e.g. 99 // whitespace or punctuation.) Under BREAK_LINE and BREAK_NEWLINE modes, 100 // this distinction doesn't apply and it always returns false. 101 bool IsWord() const; 102 103 // Under BREAK_WORD mode, returns true if |position| is at the end of word or 104 // at the start of word. It always returns false under BREAK_LINE and 105 // BREAK_NEWLINE modes. 106 bool IsEndOfWord(size_t position) const; 107 bool IsStartOfWord(size_t position) const; 108 109 // Returns the string between prev() and pos(). 110 // Advance() must have been called successfully at least once for pos() to 111 // have advanced to somewhere useful. 112 string16 GetString() const; 113 114 // Returns the value of pos() returned before Advance() was last called. prev()115 size_t prev() const { return prev_; } 116 117 // Returns the current break position within the string, 118 // or BreakIterator::npos when done. pos()119 size_t pos() const { return pos_; } 120 121 private: 122 // ICU iterator, avoiding ICU ubrk.h dependence. 123 // This is actually an ICU UBreakiterator* type, which turns out to be 124 // a typedef for a void* in the ICU headers. Using void* directly prevents 125 // callers from needing access to the ICU public headers directory. 126 void* iter_; 127 128 // The string we're iterating over. Can be changed with SetText(...) 129 const string16& string_; 130 131 // Rules for our iterator. Mutually exclusive with break_type_. 132 const string16 rules_; 133 134 // The breaking style (word/space/newline). Mutually exclusive with rules_ 135 BreakType break_type_; 136 137 // Previous and current iterator positions. 138 size_t prev_, pos_; 139 140 DISALLOW_COPY_AND_ASSIGN(BreakIterator); 141 }; 142 143 } // namespace i18n 144 } // namespace base 145 146 #endif // BASE_I18N_BREAK_ITERATOR_H_ 147