Home | History | Annotate | Download | only in i18n
      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 #pragma once
      8 
      9 #include "base/basictypes.h"
     10 #include "base/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, breking 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()) return false;
     47 //   while (iter.Advance()) {
     48 //     if (iter.IsWord()) {
     49 //       // region [iter.prev(),iter.pos()) contains a word.
     50 //       VLOG(1) << "word: " << iter.GetString();
     51 //     }
     52 //   }
     53 
     54 namespace base {
     55 
     56 class BreakIterator {
     57  public:
     58   enum BreakType {
     59     BREAK_WORD,
     60     BREAK_LINE,
     61     // TODO(jshin): Remove this after reviewing call sites.
     62     // If call sites really need break only on space-like characters
     63     // implement it separately.
     64     BREAK_SPACE = BREAK_LINE,
     65     BREAK_NEWLINE,
     66   };
     67 
     68   // Requires |str| to live as long as the BreakIterator does.
     69   BreakIterator(const string16* str, BreakType break_type);
     70   ~BreakIterator();
     71 
     72   // Init() must be called before any of the iterators are valid.
     73   // Returns false if ICU failed to initialize.
     74   bool Init();
     75 
     76   // Return the current break position within the string,
     77   // or BreakIterator::npos when done.
     78   size_t pos() const { return pos_; }
     79 
     80   // Return the value of pos() returned before Advance() was last called.
     81   size_t prev() const { return prev_; }
     82 
     83   // Advance to the next break.  Returns false if we've run past the end of
     84   // the string.  (Note that the very last "break" is after the final
     85   // character in the string, and when we advance to that position it's the
     86   // last time Advance() returns true.)
     87   bool Advance();
     88 
     89   // Under BREAK_WORD mode, returns true if the break we just hit is the
     90   // end of a word. (Otherwise, the break iterator just skipped over e.g.
     91   // whitespace or punctuation.)  Under BREAK_LINE and BREAK_NEWLINE modes,
     92   // this distinction doesn't apply and it always retuns false.
     93   bool IsWord() const;
     94 
     95   // Return the string between prev() and pos().
     96   // Advance() must have been called successfully at least once
     97   // for pos() to have advanced to somewhere useful.
     98   string16 GetString() const;
     99 
    100  private:
    101   // ICU iterator, avoiding ICU ubrk.h dependence.
    102   // This is actually an ICU UBreakiterator* type, which turns out to be
    103   // a typedef for a void* in the ICU headers. Using void* directly prevents
    104   // callers from needing access to the ICU public headers directory.
    105   void* iter_;
    106 
    107   // The string we're iterating over.
    108   const string16* string_;
    109 
    110   // The breaking style (word/space/newline).
    111   BreakType break_type_;
    112 
    113   // Previous and current iterator positions.
    114   size_t prev_, pos_;
    115 
    116   DISALLOW_COPY_AND_ASSIGN(BreakIterator);
    117 };
    118 
    119 }  // namespace base
    120 
    121 #endif  // BASE_I18N_BREAK_ITERATOR_H__
    122