/src/mozilla-central/dom/base/nsLineBreaker.h

Source (jump to first uncovered line)
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#ifndef NSLINEBREAKER_H_
#define NSLINEBREAKER_H_

#include "mozilla/intl/LineBreaker.h"
#include "nsString.h"
#include "nsTArray.h"

class nsAtom;
class nsHyphenator;

/**
 * A receiver of line break data.
 */
class nsILineBreakSink {
public:
  /**
   * Sets the break data for a substring of the associated text chunk.
   * One or more of these calls will be performed; the union of all substrings
   * will cover the entire text chunk. Substrings may overlap (i.e., we may
   * set the break-before state of a character more than once).
   * @param aBreakBefore the break-before states for the characters in the substring.
   * These are enum values from gfxTextRun::CompressedGlyph:
   *    FLAG_BREAK_TYPE_NONE     - no linebreak is allowed here
   *    FLAG_BREAK_TYPE_NORMAL   - a normal (whitespace) linebreak
   *    FLAG_BREAK_TYPE_HYPHEN   - a hyphenation point
   */
  virtual void SetBreaks(uint32_t aStart, uint32_t aLength, uint8_t* aBreakBefore) = 0;

  /**
   * Indicates which characters should be capitalized. Only called if
   * BREAK_NEED_CAPITALIZATION was requested.
   */
  virtual void SetCapitalization(uint32_t aStart, uint32_t aLength, bool* aCapitalize) = 0;
};

/**
 * A line-breaking state machine. You feed text into it via AppendText calls
 * and it computes the possible line breaks. Because break decisions can
 * require a lot of context, the breaks for a piece of text are sometimes not
 * known until later text has been seen (or all text ends). So breaks are
 * returned via a call to SetBreaks on the nsILineBreakSink object passed
 * with each text chunk, which might happen during the corresponding AppendText
 * call, or might happen during a later AppendText call or even a Reset()
 * call.
 *
 * The linebreak results MUST NOT depend on how the text is broken up
 * into AppendText calls.
 *
 * The current strategy is that we break the overall text into
 * whitespace-delimited "words". Then those words are passed to the LineBreaker
 * service for deeper analysis if they contain a "complex" character as described
 * below.
 *
 * This class also handles detection of which characters should be capitalized
 * for text-transform:capitalize. This is a good place to handle that because
 * we have all the context we need.
 */
class nsLineBreaker {
public:
  nsLineBreaker();
  ~nsLineBreaker();

  static inline bool IsSpace(char16_t u) { return mozilla::intl::NS_IsSpace(u); }

  static inline bool IsComplexASCIIChar(char16_t u)
  {
    return !((0x0030 <= u && u <= 0x0039) ||
             (0x0041 <= u && u <= 0x005A) ||
             (0x0061 <= u && u <= 0x007A) ||
             (0x000a == u));
  }

  static inline bool IsComplexChar(char16_t u)
  {
    return IsComplexASCIIChar(u) ||
           mozilla::intl::NS_NeedsPlatformNativeHandling(u) ||
           (0x1100 <= u && u <= 0x11ff) || // Hangul Jamo
           (0x2000 <= u && u <= 0x21ff) || // Punctuations and Symbols
           (0x2e80 <= u && u <= 0xd7ff) || // several CJK blocks
           (0xf900 <= u && u <= 0xfaff) || // CJK Compatibility Idographs
           (0xff00 <= u && u <= 0xffef);   // Halfwidth and Fullwidth Forms
  }

  // Break opportunities exist at the end of each run of breakable whitespace
  // (see IsSpace above). Break opportunities can also exist between pairs of
  // non-whitespace characters, as determined by mozilla::intl::LineBreaker.
  // We pass a whitespace-
  // delimited word to LineBreaker if it contains at least one character
  // matching IsComplexChar.
  // We provide flags to control on a per-chunk basis where breaks are allowed.
  // At any character boundary, exactly one text chunk governs whether a
  // break is allowed at that boundary.
  //
  // We operate on text after whitespace processing has been applied, so
  // other characters (e.g. tabs and newlines) may have been converted to
  // spaces.

  /**
   * Flags passed with each chunk of text.
   */
  enum {
    /*
     * Do not introduce a break opportunity at the start of this chunk of text.
     */
    BREAK_SUPPRESS_INITIAL = 0x01,
    /**
     * Do not introduce a break opportunity in the interior of this chunk of text.
     * Also, whitespace in this chunk is treated as non-breakable.
     */
    BREAK_SUPPRESS_INSIDE = 0x02,
    /**
     * The sink currently is already set up to have no breaks in it;
     * if no breaks are possible, nsLineBreaker does not need to call
     * SetBreaks on it. This is useful when handling large quantities of
     * preformatted text; the textruns will never have any breaks set on them,
     * and there is no need to ever actually scan the text for breaks, except
     * at the end of textruns in case context is needed for following breakable
     * text.
     */
    BREAK_SKIP_SETTING_NO_BREAKS = 0x04,
    /**
     * We need to be notified of characters that should be capitalized
     * (as in text-transform:capitalize) in this chunk of text.
     */
    BREAK_NEED_CAPITALIZATION = 0x08,
    /**
     * Auto-hyphenation is enabled, so we need to get a hyphenator
     * (if available) and use it to find breakpoints.
     */
    BREAK_USE_AUTO_HYPHENATION = 0x10
  };

  /**
   * Append "invisible whitespace". This acts like whitespace, but there is
   * no actual text associated with it. Only the BREAK_SUPPRESS_INSIDE flag
   * is relevant here.
   */
  nsresult AppendInvisibleWhitespace(uint32_t aFlags);

  /**
   * Feed Unicode text into the linebreaker for analysis. aLength must be
   * nonzero.
   * @param aSink can be null if the breaks are not actually needed (we may
   * still be setting up state for later breaks)
   */
  nsresult AppendText(nsAtom* aHyphenationLanguage, const char16_t* aText, uint32_t aLength,
                      uint32_t aFlags, nsILineBreakSink* aSink);
  /**
   * Feed 8-bit text into the linebreaker for analysis. aLength must be nonzero.
   * @param aSink can be null if the breaks are not actually needed (we may
   * still be setting up state for later breaks)
   */
  nsresult AppendText(nsAtom* aHyphenationLanguage, const uint8_t* aText, uint32_t aLength,
                      uint32_t aFlags, nsILineBreakSink* aSink);
  /**
   * Reset all state. This means the current run has ended; any outstanding
   * calls through nsILineBreakSink are made, and all outstanding references to
   * nsILineBreakSink objects are dropped.
   * After this call, this linebreaker can be reused.
   * This must be called at least once between any call to AppendText() and
   * destroying the object.
   * @param aTrailingBreak this is set to true when there is a break opportunity
   * at the end of the text. This will normally only be declared true when there
   * is breakable whitespace at the end.
   */
  nsresult Reset(bool* aTrailingBreak);

  /*
   * Set word-break mode for linebreaker.  This is set by word-break property.
   * @param aMode is LineBreaker::kWordBreak_* value.
   */
  void SetWordBreak(uint8_t aMode) { mWordBreak = aMode; }

private:
  // This is a list of text sources that make up the "current word" (i.e.,
  // run of text which does not contain any whitespace). All the mLengths
  // are are nonzero, these cannot overlap.
  struct TextItem {
    TextItem(nsILineBreakSink* aSink, uint32_t aSinkOffset, uint32_t aLength,
             uint32_t aFlags)
      : mSink(aSink), mSinkOffset(aSinkOffset), mLength(aLength), mFlags(aFlags) {}

    nsILineBreakSink* mSink;
    uint32_t          mSinkOffset;
    uint32_t          mLength;
    uint32_t          mFlags;
  };

  // State for the nonwhitespace "word" that started in previous text and hasn't
  // finished yet.

  // When the current word ends, this computes the linebreak opportunities
  // *inside* the word (excluding either end) and sets them through the
  // appropriate sink(s). Then we clear the current word state.
  nsresult FlushCurrentWord();

  void UpdateCurrentWordLanguage(nsAtom *aHyphenationLanguage);

  void FindHyphenationPoints(nsHyphenator *aHyphenator,
                             const char16_t *aTextStart,
                             const char16_t *aTextLimit,
                             uint8_t *aBreakState);

  AutoTArray<char16_t,100> mCurrentWord;
  // All the items that contribute to mCurrentWord
  AutoTArray<TextItem,2>    mTextItems;
  nsAtom*                    mCurrentWordLanguage;
  bool                        mCurrentWordContainsMixedLang;
  bool                        mCurrentWordContainsComplexChar;

  // True if the previous character was breakable whitespace
  bool                        mAfterBreakableSpace;
  // True if a break must be allowed at the current position because
  // a run of breakable whitespace ends here
  bool                        mBreakHere;
  // line break mode by "word-break" style
  uint8_t                     mWordBreak;
};

#endif /*NSLINEBREAKER_H_*/

Coverage Report

Created: 2018-09-25 14:53

Line	Count	Source (jump to first uncovered line)
1		/* -- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -- */
2		/* vim: set ts=8 sts=2 et sw=2 tw=80: */
3		/* This Source Code Form is subject to the terms of the Mozilla Public
4		* License, v. 2.0. If a copy of the MPL was not distributed with this
5		* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
6
7		#ifndef NSLINEBREAKER_H_
8		#define NSLINEBREAKER_H_
9
10		#include "mozilla/intl/LineBreaker.h"
11		#include "nsString.h"
12		#include "nsTArray.h"
13
14		class nsAtom;
15		class nsHyphenator;
16
17		/**
18		* A receiver of line break data.
19		*/
20		class nsILineBreakSink {
21		public:
22		/**
23		* Sets the break data for a substring of the associated text chunk.
24		* One or more of these calls will be performed; the union of all substrings
25		* will cover the entire text chunk. Substrings may overlap (i.e., we may
26		* set the break-before state of a character more than once).
27		* @param aBreakBefore the break-before states for the characters in the substring.
28		* These are enum values from gfxTextRun::CompressedGlyph:
29		* FLAG_BREAK_TYPE_NONE - no linebreak is allowed here
30		* FLAG_BREAK_TYPE_NORMAL - a normal (whitespace) linebreak
31		* FLAG_BREAK_TYPE_HYPHEN - a hyphenation point
32		*/
33		virtual void SetBreaks(uint32_t aStart, uint32_t aLength, uint8_t* aBreakBefore) = 0;
34
35		/**
36		* Indicates which characters should be capitalized. Only called if
37		* BREAK_NEED_CAPITALIZATION was requested.
38		*/
39		virtual void SetCapitalization(uint32_t aStart, uint32_t aLength, bool* aCapitalize) = 0;
40		};
41
42		/**
43		* A line-breaking state machine. You feed text into it via AppendText calls
44		* and it computes the possible line breaks. Because break decisions can
45		* require a lot of context, the breaks for a piece of text are sometimes not
46		* known until later text has been seen (or all text ends). So breaks are
47		* returned via a call to SetBreaks on the nsILineBreakSink object passed
48		* with each text chunk, which might happen during the corresponding AppendText
49		* call, or might happen during a later AppendText call or even a Reset()
50		* call.
51		*
52		* The linebreak results MUST NOT depend on how the text is broken up
53		* into AppendText calls.
54		*
55		* The current strategy is that we break the overall text into
56		* whitespace-delimited "words". Then those words are passed to the LineBreaker
57		* service for deeper analysis if they contain a "complex" character as described
58		* below.
59		*
60		* This class also handles detection of which characters should be capitalized
61		* for text-transform:capitalize. This is a good place to handle that because
62		* we have all the context we need.
63		*/
64		class nsLineBreaker {
65		public:
66		nsLineBreaker();
67		~nsLineBreaker();
68
69	0	static inline bool IsSpace(char16_t u) { return mozilla::intl::NS_IsSpace(u); }
70
71		static inline bool IsComplexASCIIChar(char16_t u)
72	0	{
73	0	return !((0x0030 <= u && u <= 0x0039) \|\|
74	0	(0x0041 <= u && u <= 0x005A) \|\|
75	0	(0x0061 <= u && u <= 0x007A) \|\|
76	0	(0x000a == u));
77	0	}
78
79		static inline bool IsComplexChar(char16_t u)
80	0	{
81	0	return IsComplexASCIIChar(u) \|\|
82	0	mozilla::intl::NS_NeedsPlatformNativeHandling(u) \|\|
83	0	(0x1100 <= u && u <= 0x11ff) \|\| // Hangul Jamo
84	0	(0x2000 <= u && u <= 0x21ff) \|\| // Punctuations and Symbols
85	0	(0x2e80 <= u && u <= 0xd7ff) \|\| // several CJK blocks
86	0	(0xf900 <= u && u <= 0xfaff) \|\| // CJK Compatibility Idographs
87	0	(0xff00 <= u && u <= 0xffef); // Halfwidth and Fullwidth Forms
88	0	}
89
90		// Break opportunities exist at the end of each run of breakable whitespace
91		// (see IsSpace above). Break opportunities can also exist between pairs of
92		// non-whitespace characters, as determined by mozilla::intl::LineBreaker.
93		// We pass a whitespace-
94		// delimited word to LineBreaker if it contains at least one character
95		// matching IsComplexChar.
96		// We provide flags to control on a per-chunk basis where breaks are allowed.
97		// At any character boundary, exactly one text chunk governs whether a
98		// break is allowed at that boundary.
99		//
100		// We operate on text after whitespace processing has been applied, so
101		// other characters (e.g. tabs and newlines) may have been converted to
102		// spaces.
103
104		/**
105		* Flags passed with each chunk of text.
106		*/
107		enum {
108		/*
109		* Do not introduce a break opportunity at the start of this chunk of text.
110		*/
111		BREAK_SUPPRESS_INITIAL = 0x01,
112		/**
113		* Do not introduce a break opportunity in the interior of this chunk of text.
114		* Also, whitespace in this chunk is treated as non-breakable.
115		*/
116		BREAK_SUPPRESS_INSIDE = 0x02,
117		/**
118		* The sink currently is already set up to have no breaks in it;
119		* if no breaks are possible, nsLineBreaker does not need to call
120		* SetBreaks on it. This is useful when handling large quantities of
121		* preformatted text; the textruns will never have any breaks set on them,
122		* and there is no need to ever actually scan the text for breaks, except
123		* at the end of textruns in case context is needed for following breakable
124		* text.
125		*/
126		BREAK_SKIP_SETTING_NO_BREAKS = 0x04,
127		/**
128		* We need to be notified of characters that should be capitalized
129		* (as in text-transform:capitalize) in this chunk of text.
130		*/
131		BREAK_NEED_CAPITALIZATION = 0x08,
132		/**
133		* Auto-hyphenation is enabled, so we need to get a hyphenator
134		* (if available) and use it to find breakpoints.
135		*/
136		BREAK_USE_AUTO_HYPHENATION = 0x10
137		};
138
139		/**
140		* Append "invisible whitespace". This acts like whitespace, but there is
141		* no actual text associated with it. Only the BREAK_SUPPRESS_INSIDE flag
142		* is relevant here.
143		*/
144		nsresult AppendInvisibleWhitespace(uint32_t aFlags);
145
146		/**
147		* Feed Unicode text into the linebreaker for analysis. aLength must be
148		* nonzero.
149		* @param aSink can be null if the breaks are not actually needed (we may
150		* still be setting up state for later breaks)
151		*/
152		nsresult AppendText(nsAtom* aHyphenationLanguage, const char16_t* aText, uint32_t aLength,
153		uint32_t aFlags, nsILineBreakSink* aSink);
154		/**
155		* Feed 8-bit text into the linebreaker for analysis. aLength must be nonzero.
156		* @param aSink can be null if the breaks are not actually needed (we may
157		* still be setting up state for later breaks)
158		*/
159		nsresult AppendText(nsAtom* aHyphenationLanguage, const uint8_t* aText, uint32_t aLength,
160		uint32_t aFlags, nsILineBreakSink* aSink);
161		/**
162		* Reset all state. This means the current run has ended; any outstanding
163		* calls through nsILineBreakSink are made, and all outstanding references to
164		* nsILineBreakSink objects are dropped.
165		* After this call, this linebreaker can be reused.
166		* This must be called at least once between any call to AppendText() and
167		* destroying the object.
168		* @param aTrailingBreak this is set to true when there is a break opportunity
169		* at the end of the text. This will normally only be declared true when there
170		* is breakable whitespace at the end.
171		*/
172		nsresult Reset(bool* aTrailingBreak);
173
174		/*
175		* Set word-break mode for linebreaker. This is set by word-break property.
176		* @param aMode is LineBreaker::kWordBreak_* value.
177		*/
178	0	void SetWordBreak(uint8_t aMode) { mWordBreak = aMode; }
179
180		private:
181		// This is a list of text sources that make up the "current word" (i.e.,
182		// run of text which does not contain any whitespace). All the mLengths
183		// are are nonzero, these cannot overlap.
184		struct TextItem {
185		TextItem(nsILineBreakSink* aSink, uint32_t aSinkOffset, uint32_t aLength,
186		uint32_t aFlags)
187	0	: mSink(aSink), mSinkOffset(aSinkOffset), mLength(aLength), mFlags(aFlags) {}
188
189		nsILineBreakSink* mSink;
190		uint32_t mSinkOffset;
191		uint32_t mLength;
192		uint32_t mFlags;
193		};
194
195		// State for the nonwhitespace "word" that started in previous text and hasn't
196		// finished yet.
197
198		// When the current word ends, this computes the linebreak opportunities
199		// inside the word (excluding either end) and sets them through the
200		// appropriate sink(s). Then we clear the current word state.
201		nsresult FlushCurrentWord();
202
203		void UpdateCurrentWordLanguage(nsAtom *aHyphenationLanguage);
204
205		void FindHyphenationPoints(nsHyphenator *aHyphenator,
206		const char16_t *aTextStart,
207		const char16_t *aTextLimit,
208		uint8_t *aBreakState);
209
210		AutoTArray<char16_t,100> mCurrentWord;
211		// All the items that contribute to mCurrentWord
212		AutoTArray<TextItem,2> mTextItems;
213		nsAtom* mCurrentWordLanguage;
214		bool mCurrentWordContainsMixedLang;
215		bool mCurrentWordContainsComplexChar;
216
217		// True if the previous character was breakable whitespace
218		bool mAfterBreakableSpace;
219		// True if a break must be allowed at the current position because
220		// a run of breakable whitespace ends here
221		bool mBreakHere;
222		// line break mode by "word-break" style
223		uint8_t mWordBreak;
224		};
225
226		#endif /NSLINEBREAKER_H_/