15821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Copyright (c) 2011 The Chromium Authors. All rights reserved. 25821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Use of this source code is governed by a BSD-style license that can be 35821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// found in the LICENSE file. 45821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 55821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)#ifndef BASE_I18N_BREAK_ITERATOR_H_ 65821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)#define BASE_I18N_BREAK_ITERATOR_H_ 75821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 85821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)#include "base/basictypes.h" 95821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)#include "base/i18n/base_i18n_export.h" 10868fa2fe829687343ffae624259930155e16dbd8Torne (Richard Coles)#include "base/strings/string16.h" 115821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 125821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// The BreakIterator class iterates through the words, word breaks, and 135821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// line breaks in a UTF-16 string. 145821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 155821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// It provides several modes, BREAK_WORD, BREAK_LINE, and BREAK_NEWLINE, 165821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// which modify how characters are aggregated into the returned string. 175821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 185821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Under BREAK_WORD mode, once a word is encountered any non-word 195821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// characters are not included in the returned string (e.g. in the 205821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// UTF-16 equivalent of the string " foo bar! ", the word breaks are at 215821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// the periods in ". .foo. .bar.!. ."). 225821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Note that Chinese/Japanese/Thai do not use spaces between words so that 235821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// boundaries can fall in the middle of a continuous run of non-space / 245821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// non-punctuation characters. 255821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 265821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Under BREAK_LINE mode, once a line breaking opportunity is encountered, 275821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// any non-word characters are included in the returned string, breaking 285821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// only when a space-equivalent character or a line breaking opportunity 295821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// is encountered (e.g. in the UTF16-equivalent of the string " foo bar! ", 305821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// the breaks are at the periods in ". .foo .bar! ."). 315821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 325821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Note that lines can be broken at any character/syllable/grapheme cluster 335821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// boundary in Chinese/Japanese/Korean and at word boundaries in Thai 345821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// (Thai does not use spaces between words). Therefore, this is NOT the same 355821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// as breaking only at space-equivalent characters where its former 365821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// name (BREAK_SPACE) implied. 375821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 385821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// Under BREAK_NEWLINE mode, all characters are included in the returned 394e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)// string, breaking only when a newline-equivalent character is encountered 405821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// (eg. in the UTF-16 equivalent of the string "foo\nbar!\n\n", the line 415821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// breaks are at the periods in ".foo\n.bar\n.\n."). 425821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// 435821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// To extract the words from a string, move a BREAK_WORD BreakIterator 445821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// through the string and test whether IsWord() is true. E.g., 455821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// BreakIterator iter(str, BreakIterator::BREAK_WORD); 465821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// if (!iter.Init()) 475821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// return false; 485821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// while (iter.Advance()) { 495821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// if (iter.IsWord()) { 505821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// // Region [iter.prev(), iter.pos()) contains a word. 515821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// VLOG(1) << "word: " << iter.GetString(); 525821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// } 535821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)// } 545821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 555821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)namespace base { 565821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)namespace i18n { 575821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 585821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)class BASE_I18N_EXPORT BreakIterator { 595821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) public: 605821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) enum BreakType { 615821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BREAK_WORD, 625821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BREAK_LINE, 635821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // TODO(jshin): Remove this after reviewing call sites. 645821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // If call sites really need break only on space-like characters 655821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // implement it separately. 665821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BREAK_SPACE = BREAK_LINE, 675821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BREAK_NEWLINE, 685821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BREAK_CHARACTER, 69f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // But don't remove this one! 70f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) RULE_BASED, 715821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) }; 725821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 735821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Requires |str| to live as long as the BreakIterator does. 745821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BreakIterator(const string16& str, BreakType break_type); 75f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // Make a rule-based iterator. BreakType == RULE_BASED is implied. 76f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // TODO(andrewhayden): This signature could easily be misinterpreted as 77f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // "(const string16& str, const string16& locale)". We should do something 78f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // better. 79f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) BreakIterator(const string16& str, const string16& rules); 805821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) ~BreakIterator(); 815821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 825821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Init() must be called before any of the iterators are valid. 835821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Returns false if ICU failed to initialize. 845821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) bool Init(); 855821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 865821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Advance to the next break. Returns false if we've run past the end of 875821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // the string. (Note that the very last "break" is after the final 885821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // character in the string, and when we advance to that position it's the 895821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // last time Advance() returns true.) 905821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) bool Advance(); 915821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 92f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // Updates the text used by the iterator, resetting the iterator as if 93f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // if Init() had been called again. Any old state is lost. Returns true 94f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // unless there is an error setting the text. 95f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) bool SetText(const base::char16* text, const size_t length); 96f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) 975821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Under BREAK_WORD mode, returns true if the break we just hit is the 985821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // end of a word. (Otherwise, the break iterator just skipped over e.g. 995821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // whitespace or punctuation.) Under BREAK_LINE and BREAK_NEWLINE modes, 1004e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles) // this distinction doesn't apply and it always returns false. 1015821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) bool IsWord() const; 1025821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1035821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Under BREAK_WORD mode, returns true if |position| is at the end of word or 1044e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles) // at the start of word. It always returns false under BREAK_LINE and 1055821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // BREAK_NEWLINE modes. 1065821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) bool IsEndOfWord(size_t position) const; 1075821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) bool IsStartOfWord(size_t position) const; 1085821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1095f1c94371a64b3196d4be9466099bb892df9b88eTorne (Richard Coles) // Under BREAK_CHARACTER mode, returns whether |position| is a Unicode 1105f1c94371a64b3196d4be9466099bb892df9b88eTorne (Richard Coles) // grapheme boundary. 1115f1c94371a64b3196d4be9466099bb892df9b88eTorne (Richard Coles) bool IsGraphemeBoundary(size_t position) const; 1125f1c94371a64b3196d4be9466099bb892df9b88eTorne (Richard Coles) 1135821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Returns the string between prev() and pos(). 1145821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Advance() must have been called successfully at least once for pos() to 1155821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // have advanced to somewhere useful. 1165821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) string16 GetString() const; 1175821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1185821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Returns the value of pos() returned before Advance() was last called. 1195821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) size_t prev() const { return prev_; } 1205821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1215821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Returns the current break position within the string, 1225821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // or BreakIterator::npos when done. 1235821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) size_t pos() const { return pos_; } 1245821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1255821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) private: 1265821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // ICU iterator, avoiding ICU ubrk.h dependence. 1275821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // This is actually an ICU UBreakiterator* type, which turns out to be 1285821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // a typedef for a void* in the ICU headers. Using void* directly prevents 1295821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // callers from needing access to the ICU public headers directory. 1305821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) void* iter_; 1315821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 132f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // The string we're iterating over. Can be changed with SetText(...) 1335821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) const string16& string_; 1345821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 135f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // Rules for our iterator. Mutually exclusive with break_type_. 136f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) const string16 rules_; 137f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) 138f8ee788a64d60abd8f2d742a5fdedde054ecd910Torne (Richard Coles) // The breaking style (word/space/newline). Mutually exclusive with rules_ 1395821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) BreakType break_type_; 1405821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1415821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) // Previous and current iterator positions. 1425821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) size_t prev_, pos_; 1435821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1445821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) DISALLOW_COPY_AND_ASSIGN(BreakIterator); 1455821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)}; 1465821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1475821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)} // namespace i18n 1485821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)} // namespace base 1495821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles) 1505821806d5e7f356e8fa4b058a389a808ea183019Torne (Richard Coles)#endif // BASE_I18N_BREAK_ITERATOR_H_ 151