Line data Source code
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 "nsString.h"
11 : #include "nsTArray.h"
12 : #include "nsILineBreaker.h"
13 :
14 : class nsIAtom;
15 : class nsHyphenator;
16 :
17 : /**
18 : * A receiver of line break data.
19 : */
20 21 : 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 nsILineBreaker
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 :
69 301 : static inline bool IsSpace(char16_t u) { return NS_IsSpace(u); }
70 :
71 67 : static inline bool IsComplexASCIIChar(char16_t u)
72 : {
73 133 : return !((0x0030 <= u && u <= 0x0039) ||
74 177 : (0x0041 <= u && u <= 0x005A) ||
75 110 : (0x0061 <= u && u <= 0x007A) ||
76 67 : (0x000a == u));
77 : }
78 :
79 0 : static inline bool IsComplexChar(char16_t u)
80 : {
81 0 : return IsComplexASCIIChar(u) ||
82 0 : 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 : }
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 nsILineBreaker. We pass a whitespace-
93 : // delimited word to nsILineBreaker if it contains at least one character
94 : // matching IsComplexChar.
95 : // We provide flags to control on a per-chunk basis where breaks are allowed.
96 : // At any character boundary, exactly one text chunk governs whether a
97 : // break is allowed at that boundary.
98 : //
99 : // We operate on text after whitespace processing has been applied, so
100 : // other characters (e.g. tabs and newlines) may have been converted to
101 : // spaces.
102 :
103 : /**
104 : * Flags passed with each chunk of text.
105 : */
106 : enum {
107 : /*
108 : * Do not introduce a break opportunity at the start of this chunk of text.
109 : */
110 : BREAK_SUPPRESS_INITIAL = 0x01,
111 : /**
112 : * Do not introduce a break opportunity in the interior of this chunk of text.
113 : * Also, whitespace in this chunk is treated as non-breakable.
114 : */
115 : BREAK_SUPPRESS_INSIDE = 0x02,
116 : /**
117 : * The sink currently is already set up to have no breaks in it;
118 : * if no breaks are possible, nsLineBreaker does not need to call
119 : * SetBreaks on it. This is useful when handling large quantities of
120 : * preformatted text; the textruns will never have any breaks set on them,
121 : * and there is no need to ever actually scan the text for breaks, except
122 : * at the end of textruns in case context is needed for following breakable
123 : * text.
124 : */
125 : BREAK_SKIP_SETTING_NO_BREAKS = 0x04,
126 : /**
127 : * We need to be notified of characters that should be capitalized
128 : * (as in text-transform:capitalize) in this chunk of text.
129 : */
130 : BREAK_NEED_CAPITALIZATION = 0x08,
131 : /**
132 : * Auto-hyphenation is enabled, so we need to get a hyphenator
133 : * (if available) and use it to find breakpoints.
134 : */
135 : BREAK_USE_AUTO_HYPHENATION = 0x10
136 : };
137 :
138 : /**
139 : * Append "invisible whitespace". This acts like whitespace, but there is
140 : * no actual text associated with it. Only the BREAK_SUPPRESS_INSIDE flag
141 : * is relevant here.
142 : */
143 : nsresult AppendInvisibleWhitespace(uint32_t aFlags);
144 :
145 : /**
146 : * Feed Unicode text into the linebreaker for analysis. aLength must be
147 : * nonzero.
148 : * @param aSink can be null if the breaks are not actually needed (we may
149 : * still be setting up state for later breaks)
150 : */
151 : nsresult AppendText(nsIAtom* aHyphenationLanguage, const char16_t* aText, uint32_t aLength,
152 : uint32_t aFlags, nsILineBreakSink* aSink);
153 : /**
154 : * Feed 8-bit text into the linebreaker for analysis. aLength must be nonzero.
155 : * @param aSink can be null if the breaks are not actually needed (we may
156 : * still be setting up state for later breaks)
157 : */
158 : nsresult AppendText(nsIAtom* aHyphenationLanguage, const uint8_t* aText, uint32_t aLength,
159 : uint32_t aFlags, nsILineBreakSink* aSink);
160 : /**
161 : * Reset all state. This means the current run has ended; any outstanding
162 : * calls through nsILineBreakSink are made, and all outstanding references to
163 : * nsILineBreakSink objects are dropped.
164 : * After this call, this linebreaker can be reused.
165 : * This must be called at least once between any call to AppendText() and
166 : * destroying the object.
167 : * @param aTrailingBreak this is set to true when there is a break opportunity
168 : * at the end of the text. This will normally only be declared true when there
169 : * is breakable whitespace at the end.
170 : */
171 : nsresult Reset(bool* aTrailingBreak);
172 :
173 : /*
174 : * Set word-break mode for linebreaker. This is set by word-break property.
175 : * @param aMode is nsILineBreaker::kWordBreak_* value.
176 : */
177 21 : void SetWordBreak(uint8_t aMode) { mWordBreak = aMode; }
178 :
179 : private:
180 : // This is a list of text sources that make up the "current word" (i.e.,
181 : // run of text which does not contain any whitespace). All the mLengths
182 : // are are nonzero, these cannot overlap.
183 : struct TextItem {
184 9 : TextItem(nsILineBreakSink* aSink, uint32_t aSinkOffset, uint32_t aLength,
185 : uint32_t aFlags)
186 9 : : mSink(aSink), mSinkOffset(aSinkOffset), mLength(aLength), mFlags(aFlags) {}
187 :
188 : nsILineBreakSink* mSink;
189 : uint32_t mSinkOffset;
190 : uint32_t mLength;
191 : uint32_t mFlags;
192 : };
193 :
194 : // State for the nonwhitespace "word" that started in previous text and hasn't
195 : // finished yet.
196 :
197 : // When the current word ends, this computes the linebreak opportunities
198 : // *inside* the word (excluding either end) and sets them through the
199 : // appropriate sink(s). Then we clear the current word state.
200 : nsresult FlushCurrentWord();
201 :
202 : void UpdateCurrentWordLanguage(nsIAtom *aHyphenationLanguage);
203 :
204 : void FindHyphenationPoints(nsHyphenator *aHyphenator,
205 : const char16_t *aTextStart,
206 : const char16_t *aTextLimit,
207 : uint8_t *aBreakState);
208 :
209 : AutoTArray<char16_t,100> mCurrentWord;
210 : // All the items that contribute to mCurrentWord
211 : AutoTArray<TextItem,2> mTextItems;
212 : nsIAtom* mCurrentWordLanguage;
213 : bool mCurrentWordContainsMixedLang;
214 : bool mCurrentWordContainsComplexChar;
215 :
216 : // True if the previous character was breakable whitespace
217 : bool mAfterBreakableSpace;
218 : // True if a break must be allowed at the current position because
219 : // a run of breakable whitespace ends here
220 : bool mBreakHere;
221 : // line break mode by "word-break" style
222 : uint8_t mWordBreak;
223 : };
224 :
225 : #endif /*NSLINEBREAKER_H_*/
|
